2 * Copyright (c) 2000-2006 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"
28 #include "xfs_mount.h"
29 #include "xfs_da_format.h"
30 #include "xfs_da_btree.h"
32 #include "xfs_inode.h"
33 #include "xfs_btree.h"
34 #include "xfs_trans.h"
35 #include "xfs_inode_item.h"
36 #include "xfs_extfree_item.h"
37 #include "xfs_alloc.h"
39 #include "xfs_bmap_util.h"
40 #include "xfs_bmap_btree.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_quota.h"
44 #include "xfs_trans_space.h"
45 #include "xfs_buf_item.h"
46 #include "xfs_trace.h"
47 #include "xfs_symlink.h"
48 #include "xfs_attr_leaf.h"
49 #include "xfs_dinode.h"
50 #include "xfs_filestream.h"
53 kmem_zone_t
*xfs_bmap_free_item_zone
;
56 * Miscellaneous helper functions
60 * Compute and fill in the value of the maximum depth of a bmap btree
61 * in this filesystem. Done once, during mount.
64 xfs_bmap_compute_maxlevels(
65 xfs_mount_t
*mp
, /* file system mount structure */
66 int whichfork
) /* data or attr fork */
68 int level
; /* btree level */
69 uint maxblocks
; /* max blocks at this level */
70 uint maxleafents
; /* max leaf entries possible */
71 int maxrootrecs
; /* max records in root block */
72 int minleafrecs
; /* min records in leaf block */
73 int minnoderecs
; /* min records in node block */
74 int sz
; /* root block size */
77 * The maximum number of extents in a file, hence the maximum
78 * number of leaf entries, is controlled by the type of di_nextents
79 * (a signed 32-bit number, xfs_extnum_t), or by di_anextents
80 * (a signed 16-bit number, xfs_aextnum_t).
82 * Note that we can no longer assume that if we are in ATTR1 that
83 * the fork offset of all the inodes will be
84 * (xfs_default_attroffset(ip) >> 3) because we could have mounted
85 * with ATTR2 and then mounted back with ATTR1, keeping the
86 * di_forkoff's fixed but probably at various positions. Therefore,
87 * for both ATTR1 and ATTR2 we have to assume the worst case scenario
88 * of a minimum size available.
90 if (whichfork
== XFS_DATA_FORK
) {
91 maxleafents
= MAXEXTNUM
;
92 sz
= XFS_BMDR_SPACE_CALC(MINDBTPTRS
);
94 maxleafents
= MAXAEXTNUM
;
95 sz
= XFS_BMDR_SPACE_CALC(MINABTPTRS
);
97 maxrootrecs
= xfs_bmdr_maxrecs(sz
, 0);
98 minleafrecs
= mp
->m_bmap_dmnr
[0];
99 minnoderecs
= mp
->m_bmap_dmnr
[1];
100 maxblocks
= (maxleafents
+ minleafrecs
- 1) / minleafrecs
;
101 for (level
= 1; maxblocks
> 1; level
++) {
102 if (maxblocks
<= maxrootrecs
)
105 maxblocks
= (maxblocks
+ minnoderecs
- 1) / minnoderecs
;
107 mp
->m_bm_maxlevels
[whichfork
] = level
;
110 STATIC
int /* error */
112 struct xfs_btree_cur
*cur
,
116 int *stat
) /* success/failure */
118 cur
->bc_rec
.b
.br_startoff
= off
;
119 cur
->bc_rec
.b
.br_startblock
= bno
;
120 cur
->bc_rec
.b
.br_blockcount
= len
;
121 return xfs_btree_lookup(cur
, XFS_LOOKUP_EQ
, stat
);
124 STATIC
int /* error */
126 struct xfs_btree_cur
*cur
,
130 int *stat
) /* success/failure */
132 cur
->bc_rec
.b
.br_startoff
= off
;
133 cur
->bc_rec
.b
.br_startblock
= bno
;
134 cur
->bc_rec
.b
.br_blockcount
= len
;
135 return xfs_btree_lookup(cur
, XFS_LOOKUP_GE
, stat
);
139 * Check if the inode needs to be converted to btree format.
141 static inline bool xfs_bmap_needs_btree(struct xfs_inode
*ip
, int whichfork
)
143 return XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_EXTENTS
&&
144 XFS_IFORK_NEXTENTS(ip
, whichfork
) >
145 XFS_IFORK_MAXEXT(ip
, whichfork
);
149 * Check if the inode should be converted to extent format.
151 static inline bool xfs_bmap_wants_extents(struct xfs_inode
*ip
, int whichfork
)
153 return XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_BTREE
&&
154 XFS_IFORK_NEXTENTS(ip
, whichfork
) <=
155 XFS_IFORK_MAXEXT(ip
, whichfork
);
159 * Update the record referred to by cur to the value given
160 * by [off, bno, len, state].
161 * This either works (return 0) or gets an EFSCORRUPTED error.
165 struct xfs_btree_cur
*cur
,
171 union xfs_btree_rec rec
;
173 xfs_bmbt_disk_set_allf(&rec
.bmbt
, off
, bno
, len
, state
);
174 return xfs_btree_update(cur
, &rec
);
178 * Compute the worst-case number of indirect blocks that will be used
179 * for ip's delayed extent of length "len".
182 xfs_bmap_worst_indlen(
183 xfs_inode_t
*ip
, /* incore inode pointer */
184 xfs_filblks_t len
) /* delayed extent length */
186 int level
; /* btree level number */
187 int maxrecs
; /* maximum record count at this level */
188 xfs_mount_t
*mp
; /* mount structure */
189 xfs_filblks_t rval
; /* return value */
192 maxrecs
= mp
->m_bmap_dmxr
[0];
193 for (level
= 0, rval
= 0;
194 level
< XFS_BM_MAXLEVELS(mp
, XFS_DATA_FORK
);
197 do_div(len
, maxrecs
);
200 return rval
+ XFS_BM_MAXLEVELS(mp
, XFS_DATA_FORK
) -
203 maxrecs
= mp
->m_bmap_dmxr
[1];
209 * Calculate the default attribute fork offset for newly created inodes.
212 xfs_default_attroffset(
213 struct xfs_inode
*ip
)
215 struct xfs_mount
*mp
= ip
->i_mount
;
218 if (mp
->m_sb
.sb_inodesize
== 256) {
219 offset
= XFS_LITINO(mp
, ip
->i_d
.di_version
) -
220 XFS_BMDR_SPACE_CALC(MINABTPTRS
);
222 offset
= XFS_BMDR_SPACE_CALC(6 * MINABTPTRS
);
225 ASSERT(offset
< XFS_LITINO(mp
, ip
->i_d
.di_version
));
230 * Helper routine to reset inode di_forkoff field when switching
231 * attribute fork from local to extent format - we reset it where
232 * possible to make space available for inline data fork extents.
235 xfs_bmap_forkoff_reset(
239 if (whichfork
== XFS_ATTR_FORK
&&
240 ip
->i_d
.di_format
!= XFS_DINODE_FMT_DEV
&&
241 ip
->i_d
.di_format
!= XFS_DINODE_FMT_UUID
&&
242 ip
->i_d
.di_format
!= XFS_DINODE_FMT_BTREE
) {
243 uint dfl_forkoff
= xfs_default_attroffset(ip
) >> 3;
245 if (dfl_forkoff
> ip
->i_d
.di_forkoff
)
246 ip
->i_d
.di_forkoff
= dfl_forkoff
;
251 * Debug/sanity checking code
255 xfs_bmap_sanity_check(
256 struct xfs_mount
*mp
,
260 struct xfs_btree_block
*block
= XFS_BUF_TO_BLOCK(bp
);
262 if (block
->bb_magic
!= cpu_to_be32(XFS_BMAP_CRC_MAGIC
) &&
263 block
->bb_magic
!= cpu_to_be32(XFS_BMAP_MAGIC
))
266 if (be16_to_cpu(block
->bb_level
) != level
||
267 be16_to_cpu(block
->bb_numrecs
) == 0 ||
268 be16_to_cpu(block
->bb_numrecs
) > mp
->m_bmap_dmxr
[level
!= 0])
275 STATIC
struct xfs_buf
*
277 struct xfs_btree_cur
*cur
,
280 struct xfs_log_item_desc
*lidp
;
286 for (i
= 0; i
< XFS_BTREE_MAXLEVELS
; i
++) {
287 if (!cur
->bc_bufs
[i
])
289 if (XFS_BUF_ADDR(cur
->bc_bufs
[i
]) == bno
)
290 return cur
->bc_bufs
[i
];
293 /* Chase down all the log items to see if the bp is there */
294 list_for_each_entry(lidp
, &cur
->bc_tp
->t_items
, lid_trans
) {
295 struct xfs_buf_log_item
*bip
;
296 bip
= (struct xfs_buf_log_item
*)lidp
->lid_item
;
297 if (bip
->bli_item
.li_type
== XFS_LI_BUF
&&
298 XFS_BUF_ADDR(bip
->bli_buf
) == bno
)
307 struct xfs_btree_block
*block
,
313 __be64
*pp
, *thispa
; /* pointer to block address */
314 xfs_bmbt_key_t
*prevp
, *keyp
;
316 ASSERT(be16_to_cpu(block
->bb_level
) > 0);
319 for( i
= 1; i
<= xfs_btree_get_numrecs(block
); i
++) {
320 dmxr
= mp
->m_bmap_dmxr
[0];
321 keyp
= XFS_BMBT_KEY_ADDR(mp
, block
, i
);
324 ASSERT(be64_to_cpu(prevp
->br_startoff
) <
325 be64_to_cpu(keyp
->br_startoff
));
330 * Compare the block numbers to see if there are dups.
333 pp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, block
, i
, sz
);
335 pp
= XFS_BMBT_PTR_ADDR(mp
, block
, i
, dmxr
);
337 for (j
= i
+1; j
<= be16_to_cpu(block
->bb_numrecs
); j
++) {
339 thispa
= XFS_BMAP_BROOT_PTR_ADDR(mp
, block
, j
, sz
);
341 thispa
= XFS_BMBT_PTR_ADDR(mp
, block
, j
, dmxr
);
342 if (*thispa
== *pp
) {
343 xfs_warn(mp
, "%s: thispa(%d) == pp(%d) %Ld",
345 (unsigned long long)be64_to_cpu(*thispa
));
346 panic("%s: ptrs are equal in node\n",
354 * Check that the extents for the inode ip are in the right order in all
359 xfs_bmap_check_leaf_extents(
360 xfs_btree_cur_t
*cur
, /* btree cursor or null */
361 xfs_inode_t
*ip
, /* incore inode pointer */
362 int whichfork
) /* data or attr fork */
364 struct xfs_btree_block
*block
; /* current btree block */
365 xfs_fsblock_t bno
; /* block # of "block" */
366 xfs_buf_t
*bp
; /* buffer for "block" */
367 int error
; /* error return value */
368 xfs_extnum_t i
=0, j
; /* index into the extents list */
369 xfs_ifork_t
*ifp
; /* fork structure */
370 int level
; /* btree level, for checking */
371 xfs_mount_t
*mp
; /* file system mount structure */
372 __be64
*pp
; /* pointer to block address */
373 xfs_bmbt_rec_t
*ep
; /* pointer to current extent */
374 xfs_bmbt_rec_t last
= {0, 0}; /* last extent in prev block */
375 xfs_bmbt_rec_t
*nextp
; /* pointer to next extent */
378 if (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
) {
384 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
385 block
= ifp
->if_broot
;
387 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
389 level
= be16_to_cpu(block
->bb_level
);
391 xfs_check_block(block
, mp
, 1, ifp
->if_broot_bytes
);
392 pp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, block
, 1, ifp
->if_broot_bytes
);
393 bno
= be64_to_cpu(*pp
);
395 ASSERT(bno
!= NULLFSBLOCK
);
396 ASSERT(XFS_FSB_TO_AGNO(mp
, bno
) < mp
->m_sb
.sb_agcount
);
397 ASSERT(XFS_FSB_TO_AGBNO(mp
, bno
) < mp
->m_sb
.sb_agblocks
);
400 * Go down the tree until leaf level is reached, following the first
401 * pointer (leftmost) at each level.
403 while (level
-- > 0) {
404 /* See if buf is in cur first */
406 bp
= xfs_bmap_get_bp(cur
, XFS_FSB_TO_DADDR(mp
, bno
));
409 error
= xfs_btree_read_bufl(mp
, NULL
, bno
, 0, &bp
,
415 block
= XFS_BUF_TO_BLOCK(bp
);
416 XFS_WANT_CORRUPTED_GOTO(
417 xfs_bmap_sanity_check(mp
, bp
, level
),
423 * Check this block for basic sanity (increasing keys and
424 * no duplicate blocks).
427 xfs_check_block(block
, mp
, 0, 0);
428 pp
= XFS_BMBT_PTR_ADDR(mp
, block
, 1, mp
->m_bmap_dmxr
[1]);
429 bno
= be64_to_cpu(*pp
);
430 XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp
, bno
), error0
);
433 xfs_trans_brelse(NULL
, bp
);
438 * Here with bp and block set to the leftmost leaf node in the tree.
443 * Loop over all leaf nodes checking that all extents are in the right order.
446 xfs_fsblock_t nextbno
;
447 xfs_extnum_t num_recs
;
450 num_recs
= xfs_btree_get_numrecs(block
);
453 * Read-ahead the next leaf block, if any.
456 nextbno
= be64_to_cpu(block
->bb_u
.l
.bb_rightsib
);
459 * Check all the extents to make sure they are OK.
460 * If we had a previous block, the last entry should
461 * conform with the first entry in this one.
464 ep
= XFS_BMBT_REC_ADDR(mp
, block
, 1);
466 ASSERT(xfs_bmbt_disk_get_startoff(&last
) +
467 xfs_bmbt_disk_get_blockcount(&last
) <=
468 xfs_bmbt_disk_get_startoff(ep
));
470 for (j
= 1; j
< num_recs
; j
++) {
471 nextp
= XFS_BMBT_REC_ADDR(mp
, block
, j
+ 1);
472 ASSERT(xfs_bmbt_disk_get_startoff(ep
) +
473 xfs_bmbt_disk_get_blockcount(ep
) <=
474 xfs_bmbt_disk_get_startoff(nextp
));
482 xfs_trans_brelse(NULL
, bp
);
486 * If we've reached the end, stop.
488 if (bno
== NULLFSBLOCK
)
492 bp
= xfs_bmap_get_bp(cur
, XFS_FSB_TO_DADDR(mp
, bno
));
495 error
= xfs_btree_read_bufl(mp
, NULL
, bno
, 0, &bp
,
501 block
= XFS_BUF_TO_BLOCK(bp
);
505 xfs_trans_brelse(NULL
, bp
);
510 xfs_warn(mp
, "%s: at error0", __func__
);
512 xfs_trans_brelse(NULL
, bp
);
514 xfs_warn(mp
, "%s: BAD after btree leaves for %d extents",
516 panic("%s: CORRUPTED BTREE OR SOMETHING", __func__
);
521 * Add bmap trace insert entries for all the contents of the extent records.
524 xfs_bmap_trace_exlist(
525 xfs_inode_t
*ip
, /* incore inode pointer */
526 xfs_extnum_t cnt
, /* count of entries in the list */
527 int whichfork
, /* data or attr fork */
528 unsigned long caller_ip
)
530 xfs_extnum_t idx
; /* extent record index */
531 xfs_ifork_t
*ifp
; /* inode fork pointer */
534 if (whichfork
== XFS_ATTR_FORK
)
535 state
|= BMAP_ATTRFORK
;
537 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
538 ASSERT(cnt
== (ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
)));
539 for (idx
= 0; idx
< cnt
; idx
++)
540 trace_xfs_extlist(ip
, idx
, whichfork
, caller_ip
);
544 * Validate that the bmbt_irecs being returned from bmapi are valid
545 * given the caller's original parameters. Specifically check the
546 * ranges of the returned irecs to ensure that they only extend beyond
547 * the given parameters if the XFS_BMAPI_ENTIRE flag was set.
550 xfs_bmap_validate_ret(
554 xfs_bmbt_irec_t
*mval
,
558 int i
; /* index to map values */
560 ASSERT(ret_nmap
<= nmap
);
562 for (i
= 0; i
< ret_nmap
; i
++) {
563 ASSERT(mval
[i
].br_blockcount
> 0);
564 if (!(flags
& XFS_BMAPI_ENTIRE
)) {
565 ASSERT(mval
[i
].br_startoff
>= bno
);
566 ASSERT(mval
[i
].br_blockcount
<= len
);
567 ASSERT(mval
[i
].br_startoff
+ mval
[i
].br_blockcount
<=
570 ASSERT(mval
[i
].br_startoff
< bno
+ len
);
571 ASSERT(mval
[i
].br_startoff
+ mval
[i
].br_blockcount
>
575 mval
[i
- 1].br_startoff
+ mval
[i
- 1].br_blockcount
==
576 mval
[i
].br_startoff
);
577 ASSERT(mval
[i
].br_startblock
!= DELAYSTARTBLOCK
&&
578 mval
[i
].br_startblock
!= HOLESTARTBLOCK
);
579 ASSERT(mval
[i
].br_state
== XFS_EXT_NORM
||
580 mval
[i
].br_state
== XFS_EXT_UNWRITTEN
);
585 #define xfs_bmap_check_leaf_extents(cur, ip, whichfork) do { } while (0)
586 #define xfs_bmap_validate_ret(bno,len,flags,mval,onmap,nmap)
590 * bmap free list manipulation functions
594 * Add the extent to the list of extents to be free at transaction end.
595 * The list is maintained sorted (by block number).
599 xfs_fsblock_t bno
, /* fs block number of extent */
600 xfs_filblks_t len
, /* length of extent */
601 xfs_bmap_free_t
*flist
, /* list of extents */
602 xfs_mount_t
*mp
) /* mount point structure */
604 xfs_bmap_free_item_t
*cur
; /* current (next) element */
605 xfs_bmap_free_item_t
*new; /* new element */
606 xfs_bmap_free_item_t
*prev
; /* previous element */
611 ASSERT(bno
!= NULLFSBLOCK
);
613 ASSERT(len
<= MAXEXTLEN
);
614 ASSERT(!isnullstartblock(bno
));
615 agno
= XFS_FSB_TO_AGNO(mp
, bno
);
616 agbno
= XFS_FSB_TO_AGBNO(mp
, bno
);
617 ASSERT(agno
< mp
->m_sb
.sb_agcount
);
618 ASSERT(agbno
< mp
->m_sb
.sb_agblocks
);
619 ASSERT(len
< mp
->m_sb
.sb_agblocks
);
620 ASSERT(agbno
+ len
<= mp
->m_sb
.sb_agblocks
);
622 ASSERT(xfs_bmap_free_item_zone
!= NULL
);
623 new = kmem_zone_alloc(xfs_bmap_free_item_zone
, KM_SLEEP
);
624 new->xbfi_startblock
= bno
;
625 new->xbfi_blockcount
= (xfs_extlen_t
)len
;
626 for (prev
= NULL
, cur
= flist
->xbf_first
;
628 prev
= cur
, cur
= cur
->xbfi_next
) {
629 if (cur
->xbfi_startblock
>= bno
)
633 prev
->xbfi_next
= new;
635 flist
->xbf_first
= new;
636 new->xbfi_next
= cur
;
641 * Remove the entry "free" from the free item list. Prev points to the
642 * previous entry, unless "free" is the head of the list.
646 xfs_bmap_free_t
*flist
, /* free item list header */
647 xfs_bmap_free_item_t
*prev
, /* previous item on list, if any */
648 xfs_bmap_free_item_t
*free
) /* list item to be freed */
651 prev
->xbfi_next
= free
->xbfi_next
;
653 flist
->xbf_first
= free
->xbfi_next
;
655 kmem_zone_free(xfs_bmap_free_item_zone
, free
);
659 * Free up any items left in the list.
663 xfs_bmap_free_t
*flist
) /* list of bmap_free_items */
665 xfs_bmap_free_item_t
*free
; /* free list item */
666 xfs_bmap_free_item_t
*next
;
668 if (flist
->xbf_count
== 0)
670 ASSERT(flist
->xbf_first
!= NULL
);
671 for (free
= flist
->xbf_first
; free
; free
= next
) {
672 next
= free
->xbfi_next
;
673 xfs_bmap_del_free(flist
, NULL
, free
);
675 ASSERT(flist
->xbf_count
== 0);
679 * Inode fork format manipulation functions
683 * Transform a btree format file with only one leaf node, where the
684 * extents list will fit in the inode, into an extents format file.
685 * Since the file extents are already in-core, all we have to do is
686 * give up the space for the btree root and pitch the leaf block.
688 STATIC
int /* error */
689 xfs_bmap_btree_to_extents(
690 xfs_trans_t
*tp
, /* transaction pointer */
691 xfs_inode_t
*ip
, /* incore inode pointer */
692 xfs_btree_cur_t
*cur
, /* btree cursor */
693 int *logflagsp
, /* inode logging flags */
694 int whichfork
) /* data or attr fork */
697 struct xfs_btree_block
*cblock
;/* child btree block */
698 xfs_fsblock_t cbno
; /* child block number */
699 xfs_buf_t
*cbp
; /* child block's buffer */
700 int error
; /* error return value */
701 xfs_ifork_t
*ifp
; /* inode fork data */
702 xfs_mount_t
*mp
; /* mount point structure */
703 __be64
*pp
; /* ptr to block address */
704 struct xfs_btree_block
*rblock
;/* root btree block */
707 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
708 ASSERT(ifp
->if_flags
& XFS_IFEXTENTS
);
709 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_BTREE
);
710 rblock
= ifp
->if_broot
;
711 ASSERT(be16_to_cpu(rblock
->bb_level
) == 1);
712 ASSERT(be16_to_cpu(rblock
->bb_numrecs
) == 1);
713 ASSERT(xfs_bmbt_maxrecs(mp
, ifp
->if_broot_bytes
, 0) == 1);
714 pp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, rblock
, 1, ifp
->if_broot_bytes
);
715 cbno
= be64_to_cpu(*pp
);
718 if ((error
= xfs_btree_check_lptr(cur
, cbno
, 1)))
721 error
= xfs_btree_read_bufl(mp
, tp
, cbno
, 0, &cbp
, XFS_BMAP_BTREE_REF
,
725 cblock
= XFS_BUF_TO_BLOCK(cbp
);
726 if ((error
= xfs_btree_check_block(cur
, cblock
, 0, cbp
)))
728 xfs_bmap_add_free(cbno
, 1, cur
->bc_private
.b
.flist
, mp
);
729 ip
->i_d
.di_nblocks
--;
730 xfs_trans_mod_dquot_byino(tp
, ip
, XFS_TRANS_DQ_BCOUNT
, -1L);
731 xfs_trans_binval(tp
, cbp
);
732 if (cur
->bc_bufs
[0] == cbp
)
733 cur
->bc_bufs
[0] = NULL
;
734 xfs_iroot_realloc(ip
, -1, whichfork
);
735 ASSERT(ifp
->if_broot
== NULL
);
736 ASSERT((ifp
->if_flags
& XFS_IFBROOT
) == 0);
737 XFS_IFORK_FMT_SET(ip
, whichfork
, XFS_DINODE_FMT_EXTENTS
);
738 *logflagsp
= XFS_ILOG_CORE
| xfs_ilog_fext(whichfork
);
743 * Convert an extents-format file into a btree-format file.
744 * The new file will have a root block (in the inode) and a single child block.
746 STATIC
int /* error */
747 xfs_bmap_extents_to_btree(
748 xfs_trans_t
*tp
, /* transaction pointer */
749 xfs_inode_t
*ip
, /* incore inode pointer */
750 xfs_fsblock_t
*firstblock
, /* first-block-allocated */
751 xfs_bmap_free_t
*flist
, /* blocks freed in xaction */
752 xfs_btree_cur_t
**curp
, /* cursor returned to caller */
753 int wasdel
, /* converting a delayed alloc */
754 int *logflagsp
, /* inode logging flags */
755 int whichfork
) /* data or attr fork */
757 struct xfs_btree_block
*ablock
; /* allocated (child) bt block */
758 xfs_buf_t
*abp
; /* buffer for ablock */
759 xfs_alloc_arg_t args
; /* allocation arguments */
760 xfs_bmbt_rec_t
*arp
; /* child record pointer */
761 struct xfs_btree_block
*block
; /* btree root block */
762 xfs_btree_cur_t
*cur
; /* bmap btree cursor */
763 xfs_bmbt_rec_host_t
*ep
; /* extent record pointer */
764 int error
; /* error return value */
765 xfs_extnum_t i
, cnt
; /* extent record index */
766 xfs_ifork_t
*ifp
; /* inode fork pointer */
767 xfs_bmbt_key_t
*kp
; /* root block key pointer */
768 xfs_mount_t
*mp
; /* mount structure */
769 xfs_extnum_t nextents
; /* number of file extents */
770 xfs_bmbt_ptr_t
*pp
; /* root block address pointer */
773 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
774 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_EXTENTS
);
777 * Make space in the inode incore.
779 xfs_iroot_realloc(ip
, 1, whichfork
);
780 ifp
->if_flags
|= XFS_IFBROOT
;
785 block
= ifp
->if_broot
;
786 if (xfs_sb_version_hascrc(&mp
->m_sb
))
787 xfs_btree_init_block_int(mp
, block
, XFS_BUF_DADDR_NULL
,
788 XFS_BMAP_CRC_MAGIC
, 1, 1, ip
->i_ino
,
789 XFS_BTREE_LONG_PTRS
| XFS_BTREE_CRC_BLOCKS
);
791 xfs_btree_init_block_int(mp
, block
, XFS_BUF_DADDR_NULL
,
792 XFS_BMAP_MAGIC
, 1, 1, ip
->i_ino
,
793 XFS_BTREE_LONG_PTRS
);
796 * Need a cursor. Can't allocate until bb_level is filled in.
798 cur
= xfs_bmbt_init_cursor(mp
, tp
, ip
, whichfork
);
799 cur
->bc_private
.b
.firstblock
= *firstblock
;
800 cur
->bc_private
.b
.flist
= flist
;
801 cur
->bc_private
.b
.flags
= wasdel
? XFS_BTCUR_BPRV_WASDEL
: 0;
803 * Convert to a btree with two levels, one record in root.
805 XFS_IFORK_FMT_SET(ip
, whichfork
, XFS_DINODE_FMT_BTREE
);
806 memset(&args
, 0, sizeof(args
));
809 args
.firstblock
= *firstblock
;
810 if (*firstblock
== NULLFSBLOCK
) {
811 args
.type
= XFS_ALLOCTYPE_START_BNO
;
812 args
.fsbno
= XFS_INO_TO_FSB(mp
, ip
->i_ino
);
813 } else if (flist
->xbf_low
) {
814 args
.type
= XFS_ALLOCTYPE_START_BNO
;
815 args
.fsbno
= *firstblock
;
817 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
818 args
.fsbno
= *firstblock
;
820 args
.minlen
= args
.maxlen
= args
.prod
= 1;
821 args
.wasdel
= wasdel
;
823 if ((error
= xfs_alloc_vextent(&args
))) {
824 xfs_iroot_realloc(ip
, -1, whichfork
);
825 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
829 * Allocation can't fail, the space was reserved.
831 ASSERT(args
.fsbno
!= NULLFSBLOCK
);
832 ASSERT(*firstblock
== NULLFSBLOCK
||
833 args
.agno
== XFS_FSB_TO_AGNO(mp
, *firstblock
) ||
835 args
.agno
> XFS_FSB_TO_AGNO(mp
, *firstblock
)));
836 *firstblock
= cur
->bc_private
.b
.firstblock
= args
.fsbno
;
837 cur
->bc_private
.b
.allocated
++;
838 ip
->i_d
.di_nblocks
++;
839 xfs_trans_mod_dquot_byino(tp
, ip
, XFS_TRANS_DQ_BCOUNT
, 1L);
840 abp
= xfs_btree_get_bufl(mp
, tp
, args
.fsbno
, 0);
842 * Fill in the child block.
844 abp
->b_ops
= &xfs_bmbt_buf_ops
;
845 ablock
= XFS_BUF_TO_BLOCK(abp
);
846 if (xfs_sb_version_hascrc(&mp
->m_sb
))
847 xfs_btree_init_block_int(mp
, ablock
, abp
->b_bn
,
848 XFS_BMAP_CRC_MAGIC
, 0, 0, ip
->i_ino
,
849 XFS_BTREE_LONG_PTRS
| XFS_BTREE_CRC_BLOCKS
);
851 xfs_btree_init_block_int(mp
, ablock
, abp
->b_bn
,
852 XFS_BMAP_MAGIC
, 0, 0, ip
->i_ino
,
853 XFS_BTREE_LONG_PTRS
);
855 arp
= XFS_BMBT_REC_ADDR(mp
, ablock
, 1);
856 nextents
= ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
);
857 for (cnt
= i
= 0; i
< nextents
; i
++) {
858 ep
= xfs_iext_get_ext(ifp
, i
);
859 if (!isnullstartblock(xfs_bmbt_get_startblock(ep
))) {
860 arp
->l0
= cpu_to_be64(ep
->l0
);
861 arp
->l1
= cpu_to_be64(ep
->l1
);
865 ASSERT(cnt
== XFS_IFORK_NEXTENTS(ip
, whichfork
));
866 xfs_btree_set_numrecs(ablock
, cnt
);
869 * Fill in the root key and pointer.
871 kp
= XFS_BMBT_KEY_ADDR(mp
, block
, 1);
872 arp
= XFS_BMBT_REC_ADDR(mp
, ablock
, 1);
873 kp
->br_startoff
= cpu_to_be64(xfs_bmbt_disk_get_startoff(arp
));
874 pp
= XFS_BMBT_PTR_ADDR(mp
, block
, 1, xfs_bmbt_get_maxrecs(cur
,
875 be16_to_cpu(block
->bb_level
)));
876 *pp
= cpu_to_be64(args
.fsbno
);
879 * Do all this logging at the end so that
880 * the root is at the right level.
882 xfs_btree_log_block(cur
, abp
, XFS_BB_ALL_BITS
);
883 xfs_btree_log_recs(cur
, abp
, 1, be16_to_cpu(ablock
->bb_numrecs
));
884 ASSERT(*curp
== NULL
);
886 *logflagsp
= XFS_ILOG_CORE
| xfs_ilog_fbroot(whichfork
);
891 * Convert a local file to an extents file.
892 * This code is out of bounds for data forks of regular files,
893 * since the file data needs to get logged so things will stay consistent.
894 * (The bmap-level manipulations are ok, though).
897 xfs_bmap_local_to_extents_empty(
898 struct xfs_inode
*ip
,
901 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(ip
, whichfork
);
903 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
);
904 ASSERT(ifp
->if_bytes
== 0);
905 ASSERT(XFS_IFORK_NEXTENTS(ip
, whichfork
) == 0);
907 xfs_bmap_forkoff_reset(ip
, whichfork
);
908 ifp
->if_flags
&= ~XFS_IFINLINE
;
909 ifp
->if_flags
|= XFS_IFEXTENTS
;
910 XFS_IFORK_FMT_SET(ip
, whichfork
, XFS_DINODE_FMT_EXTENTS
);
914 STATIC
int /* error */
915 xfs_bmap_local_to_extents(
916 xfs_trans_t
*tp
, /* transaction pointer */
917 xfs_inode_t
*ip
, /* incore inode pointer */
918 xfs_fsblock_t
*firstblock
, /* first block allocated in xaction */
919 xfs_extlen_t total
, /* total blocks needed by transaction */
920 int *logflagsp
, /* inode logging flags */
922 void (*init_fn
)(struct xfs_trans
*tp
,
924 struct xfs_inode
*ip
,
925 struct xfs_ifork
*ifp
))
928 int flags
; /* logging flags returned */
929 xfs_ifork_t
*ifp
; /* inode fork pointer */
930 xfs_alloc_arg_t args
; /* allocation arguments */
931 xfs_buf_t
*bp
; /* buffer for extent block */
932 xfs_bmbt_rec_host_t
*ep
; /* extent record pointer */
935 * We don't want to deal with the case of keeping inode data inline yet.
936 * So sending the data fork of a regular inode is invalid.
938 ASSERT(!(S_ISREG(ip
->i_d
.di_mode
) && whichfork
== XFS_DATA_FORK
));
939 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
940 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
);
942 if (!ifp
->if_bytes
) {
943 xfs_bmap_local_to_extents_empty(ip
, whichfork
);
944 flags
= XFS_ILOG_CORE
;
950 ASSERT((ifp
->if_flags
& (XFS_IFINLINE
|XFS_IFEXTENTS
|XFS_IFEXTIREC
)) ==
952 memset(&args
, 0, sizeof(args
));
954 args
.mp
= ip
->i_mount
;
955 args
.firstblock
= *firstblock
;
957 * Allocate a block. We know we need only one, since the
958 * file currently fits in an inode.
960 if (*firstblock
== NULLFSBLOCK
) {
961 args
.fsbno
= XFS_INO_TO_FSB(args
.mp
, ip
->i_ino
);
962 args
.type
= XFS_ALLOCTYPE_START_BNO
;
964 args
.fsbno
= *firstblock
;
965 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
968 args
.minlen
= args
.maxlen
= args
.prod
= 1;
969 error
= xfs_alloc_vextent(&args
);
973 /* Can't fail, the space was reserved. */
974 ASSERT(args
.fsbno
!= NULLFSBLOCK
);
975 ASSERT(args
.len
== 1);
976 *firstblock
= args
.fsbno
;
977 bp
= xfs_btree_get_bufl(args
.mp
, tp
, args
.fsbno
, 0);
979 /* initialise the block and copy the data */
980 init_fn(tp
, bp
, ip
, ifp
);
982 /* account for the change in fork size and log everything */
983 xfs_trans_log_buf(tp
, bp
, 0, ifp
->if_bytes
- 1);
984 xfs_idata_realloc(ip
, -ifp
->if_bytes
, whichfork
);
985 xfs_bmap_local_to_extents_empty(ip
, whichfork
);
986 flags
|= XFS_ILOG_CORE
;
988 xfs_iext_add(ifp
, 0, 1);
989 ep
= xfs_iext_get_ext(ifp
, 0);
990 xfs_bmbt_set_allf(ep
, 0, args
.fsbno
, 1, XFS_EXT_NORM
);
991 trace_xfs_bmap_post_update(ip
, 0,
992 whichfork
== XFS_ATTR_FORK
? BMAP_ATTRFORK
: 0,
994 XFS_IFORK_NEXT_SET(ip
, whichfork
, 1);
995 ip
->i_d
.di_nblocks
= 1;
996 xfs_trans_mod_dquot_byino(tp
, ip
,
997 XFS_TRANS_DQ_BCOUNT
, 1L);
998 flags
|= xfs_ilog_fext(whichfork
);
1006 * Called from xfs_bmap_add_attrfork to handle btree format files.
1008 STATIC
int /* error */
1009 xfs_bmap_add_attrfork_btree(
1010 xfs_trans_t
*tp
, /* transaction pointer */
1011 xfs_inode_t
*ip
, /* incore inode pointer */
1012 xfs_fsblock_t
*firstblock
, /* first block allocated */
1013 xfs_bmap_free_t
*flist
, /* blocks to free at commit */
1014 int *flags
) /* inode logging flags */
1016 xfs_btree_cur_t
*cur
; /* btree cursor */
1017 int error
; /* error return value */
1018 xfs_mount_t
*mp
; /* file system mount struct */
1019 int stat
; /* newroot status */
1022 if (ip
->i_df
.if_broot_bytes
<= XFS_IFORK_DSIZE(ip
))
1023 *flags
|= XFS_ILOG_DBROOT
;
1025 cur
= xfs_bmbt_init_cursor(mp
, tp
, ip
, XFS_DATA_FORK
);
1026 cur
->bc_private
.b
.flist
= flist
;
1027 cur
->bc_private
.b
.firstblock
= *firstblock
;
1028 if ((error
= xfs_bmbt_lookup_ge(cur
, 0, 0, 0, &stat
)))
1030 /* must be at least one entry */
1031 XFS_WANT_CORRUPTED_GOTO(stat
== 1, error0
);
1032 if ((error
= xfs_btree_new_iroot(cur
, flags
, &stat
)))
1035 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
1038 *firstblock
= cur
->bc_private
.b
.firstblock
;
1039 cur
->bc_private
.b
.allocated
= 0;
1040 xfs_btree_del_cursor(cur
, XFS_BTREE_NOERROR
);
1044 xfs_btree_del_cursor(cur
, XFS_BTREE_ERROR
);
1049 * Called from xfs_bmap_add_attrfork to handle extents format files.
1051 STATIC
int /* error */
1052 xfs_bmap_add_attrfork_extents(
1053 xfs_trans_t
*tp
, /* transaction pointer */
1054 xfs_inode_t
*ip
, /* incore inode pointer */
1055 xfs_fsblock_t
*firstblock
, /* first block allocated */
1056 xfs_bmap_free_t
*flist
, /* blocks to free at commit */
1057 int *flags
) /* inode logging flags */
1059 xfs_btree_cur_t
*cur
; /* bmap btree cursor */
1060 int error
; /* error return value */
1062 if (ip
->i_d
.di_nextents
* sizeof(xfs_bmbt_rec_t
) <= XFS_IFORK_DSIZE(ip
))
1065 error
= xfs_bmap_extents_to_btree(tp
, ip
, firstblock
, flist
, &cur
, 0,
1066 flags
, XFS_DATA_FORK
);
1068 cur
->bc_private
.b
.allocated
= 0;
1069 xfs_btree_del_cursor(cur
,
1070 error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);
1076 * Called from xfs_bmap_add_attrfork to handle local format files. Each
1077 * different data fork content type needs a different callout to do the
1078 * conversion. Some are basic and only require special block initialisation
1079 * callouts for the data formating, others (directories) are so specialised they
1080 * handle everything themselves.
1082 * XXX (dgc): investigate whether directory conversion can use the generic
1083 * formatting callout. It should be possible - it's just a very complex
1086 STATIC
int /* error */
1087 xfs_bmap_add_attrfork_local(
1088 xfs_trans_t
*tp
, /* transaction pointer */
1089 xfs_inode_t
*ip
, /* incore inode pointer */
1090 xfs_fsblock_t
*firstblock
, /* first block allocated */
1091 xfs_bmap_free_t
*flist
, /* blocks to free at commit */
1092 int *flags
) /* inode logging flags */
1094 xfs_da_args_t dargs
; /* args for dir/attr code */
1096 if (ip
->i_df
.if_bytes
<= XFS_IFORK_DSIZE(ip
))
1099 if (S_ISDIR(ip
->i_d
.di_mode
)) {
1100 memset(&dargs
, 0, sizeof(dargs
));
1101 dargs
.geo
= ip
->i_mount
->m_dir_geo
;
1103 dargs
.firstblock
= firstblock
;
1104 dargs
.flist
= flist
;
1105 dargs
.total
= dargs
.geo
->fsbcount
;
1106 dargs
.whichfork
= XFS_DATA_FORK
;
1108 return xfs_dir2_sf_to_block(&dargs
);
1111 if (S_ISLNK(ip
->i_d
.di_mode
))
1112 return xfs_bmap_local_to_extents(tp
, ip
, firstblock
, 1,
1113 flags
, XFS_DATA_FORK
,
1114 xfs_symlink_local_to_remote
);
1116 /* should only be called for types that support local format data */
1118 return -EFSCORRUPTED
;
1122 * Convert inode from non-attributed to attributed.
1123 * Must not be in a transaction, ip must not be locked.
1125 int /* error code */
1126 xfs_bmap_add_attrfork(
1127 xfs_inode_t
*ip
, /* incore inode pointer */
1128 int size
, /* space new attribute needs */
1129 int rsvd
) /* xact may use reserved blks */
1131 xfs_fsblock_t firstblock
; /* 1st block/ag allocated */
1132 xfs_bmap_free_t flist
; /* freed extent records */
1133 xfs_mount_t
*mp
; /* mount structure */
1134 xfs_trans_t
*tp
; /* transaction pointer */
1135 int blks
; /* space reservation */
1136 int version
= 1; /* superblock attr version */
1137 int committed
; /* xaction was committed */
1138 int logflags
; /* logging flags */
1139 int error
; /* error return value */
1140 int cancel_flags
= 0;
1142 ASSERT(XFS_IFORK_Q(ip
) == 0);
1145 ASSERT(!XFS_NOT_DQATTACHED(mp
, ip
));
1146 tp
= xfs_trans_alloc(mp
, XFS_TRANS_ADDAFORK
);
1147 blks
= XFS_ADDAFORK_SPACE_RES(mp
);
1149 tp
->t_flags
|= XFS_TRANS_RESERVE
;
1150 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_addafork
, blks
, 0);
1152 xfs_trans_cancel(tp
, 0);
1155 cancel_flags
= XFS_TRANS_RELEASE_LOG_RES
;
1156 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
1157 error
= xfs_trans_reserve_quota_nblks(tp
, ip
, blks
, 0, rsvd
?
1158 XFS_QMOPT_RES_REGBLKS
| XFS_QMOPT_FORCE_RES
:
1159 XFS_QMOPT_RES_REGBLKS
);
1162 cancel_flags
|= XFS_TRANS_ABORT
;
1163 if (XFS_IFORK_Q(ip
))
1165 if (ip
->i_d
.di_aformat
!= XFS_DINODE_FMT_EXTENTS
) {
1167 * For inodes coming from pre-6.2 filesystems.
1169 ASSERT(ip
->i_d
.di_aformat
== 0);
1170 ip
->i_d
.di_aformat
= XFS_DINODE_FMT_EXTENTS
;
1172 ASSERT(ip
->i_d
.di_anextents
== 0);
1174 xfs_trans_ijoin(tp
, ip
, 0);
1175 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
1177 switch (ip
->i_d
.di_format
) {
1178 case XFS_DINODE_FMT_DEV
:
1179 ip
->i_d
.di_forkoff
= roundup(sizeof(xfs_dev_t
), 8) >> 3;
1181 case XFS_DINODE_FMT_UUID
:
1182 ip
->i_d
.di_forkoff
= roundup(sizeof(uuid_t
), 8) >> 3;
1184 case XFS_DINODE_FMT_LOCAL
:
1185 case XFS_DINODE_FMT_EXTENTS
:
1186 case XFS_DINODE_FMT_BTREE
:
1187 ip
->i_d
.di_forkoff
= xfs_attr_shortform_bytesfit(ip
, size
);
1188 if (!ip
->i_d
.di_forkoff
)
1189 ip
->i_d
.di_forkoff
= xfs_default_attroffset(ip
) >> 3;
1190 else if (mp
->m_flags
& XFS_MOUNT_ATTR2
)
1199 ASSERT(ip
->i_afp
== NULL
);
1200 ip
->i_afp
= kmem_zone_zalloc(xfs_ifork_zone
, KM_SLEEP
);
1201 ip
->i_afp
->if_flags
= XFS_IFEXTENTS
;
1203 xfs_bmap_init(&flist
, &firstblock
);
1204 switch (ip
->i_d
.di_format
) {
1205 case XFS_DINODE_FMT_LOCAL
:
1206 error
= xfs_bmap_add_attrfork_local(tp
, ip
, &firstblock
, &flist
,
1209 case XFS_DINODE_FMT_EXTENTS
:
1210 error
= xfs_bmap_add_attrfork_extents(tp
, ip
, &firstblock
,
1213 case XFS_DINODE_FMT_BTREE
:
1214 error
= xfs_bmap_add_attrfork_btree(tp
, ip
, &firstblock
, &flist
,
1222 xfs_trans_log_inode(tp
, ip
, logflags
);
1225 if (!xfs_sb_version_hasattr(&mp
->m_sb
) ||
1226 (!xfs_sb_version_hasattr2(&mp
->m_sb
) && version
== 2)) {
1227 __int64_t sbfields
= 0;
1229 spin_lock(&mp
->m_sb_lock
);
1230 if (!xfs_sb_version_hasattr(&mp
->m_sb
)) {
1231 xfs_sb_version_addattr(&mp
->m_sb
);
1232 sbfields
|= XFS_SB_VERSIONNUM
;
1234 if (!xfs_sb_version_hasattr2(&mp
->m_sb
) && version
== 2) {
1235 xfs_sb_version_addattr2(&mp
->m_sb
);
1236 sbfields
|= (XFS_SB_VERSIONNUM
| XFS_SB_FEATURES2
);
1239 spin_unlock(&mp
->m_sb_lock
);
1240 xfs_mod_sb(tp
, sbfields
);
1242 spin_unlock(&mp
->m_sb_lock
);
1245 error
= xfs_bmap_finish(&tp
, &flist
, &committed
);
1248 error
= xfs_trans_commit(tp
, XFS_TRANS_RELEASE_LOG_RES
);
1249 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
1253 xfs_bmap_cancel(&flist
);
1255 xfs_trans_cancel(tp
, cancel_flags
);
1256 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
1261 * Internal and external extent tree search functions.
1265 * Read in the extents to if_extents.
1266 * All inode fields are set up by caller, we just traverse the btree
1267 * and copy the records in. If the file system cannot contain unwritten
1268 * extents, the records are checked for no "state" flags.
1271 xfs_bmap_read_extents(
1272 xfs_trans_t
*tp
, /* transaction pointer */
1273 xfs_inode_t
*ip
, /* incore inode */
1274 int whichfork
) /* data or attr fork */
1276 struct xfs_btree_block
*block
; /* current btree block */
1277 xfs_fsblock_t bno
; /* block # of "block" */
1278 xfs_buf_t
*bp
; /* buffer for "block" */
1279 int error
; /* error return value */
1280 xfs_exntfmt_t exntf
; /* XFS_EXTFMT_NOSTATE, if checking */
1281 xfs_extnum_t i
, j
; /* index into the extents list */
1282 xfs_ifork_t
*ifp
; /* fork structure */
1283 int level
; /* btree level, for checking */
1284 xfs_mount_t
*mp
; /* file system mount structure */
1285 __be64
*pp
; /* pointer to block address */
1287 xfs_extnum_t room
; /* number of entries there's room for */
1291 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
1292 exntf
= (whichfork
!= XFS_DATA_FORK
) ? XFS_EXTFMT_NOSTATE
:
1293 XFS_EXTFMT_INODE(ip
);
1294 block
= ifp
->if_broot
;
1296 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
1298 level
= be16_to_cpu(block
->bb_level
);
1300 pp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, block
, 1, ifp
->if_broot_bytes
);
1301 bno
= be64_to_cpu(*pp
);
1302 ASSERT(bno
!= NULLFSBLOCK
);
1303 ASSERT(XFS_FSB_TO_AGNO(mp
, bno
) < mp
->m_sb
.sb_agcount
);
1304 ASSERT(XFS_FSB_TO_AGBNO(mp
, bno
) < mp
->m_sb
.sb_agblocks
);
1306 * Go down the tree until leaf level is reached, following the first
1307 * pointer (leftmost) at each level.
1309 while (level
-- > 0) {
1310 error
= xfs_btree_read_bufl(mp
, tp
, bno
, 0, &bp
,
1311 XFS_BMAP_BTREE_REF
, &xfs_bmbt_buf_ops
);
1314 block
= XFS_BUF_TO_BLOCK(bp
);
1315 XFS_WANT_CORRUPTED_GOTO(
1316 xfs_bmap_sanity_check(mp
, bp
, level
),
1320 pp
= XFS_BMBT_PTR_ADDR(mp
, block
, 1, mp
->m_bmap_dmxr
[1]);
1321 bno
= be64_to_cpu(*pp
);
1322 XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp
, bno
), error0
);
1323 xfs_trans_brelse(tp
, bp
);
1326 * Here with bp and block set to the leftmost leaf node in the tree.
1328 room
= ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
);
1331 * Loop over all leaf nodes. Copy information to the extent records.
1334 xfs_bmbt_rec_t
*frp
;
1335 xfs_fsblock_t nextbno
;
1336 xfs_extnum_t num_recs
;
1339 num_recs
= xfs_btree_get_numrecs(block
);
1340 if (unlikely(i
+ num_recs
> room
)) {
1341 ASSERT(i
+ num_recs
<= room
);
1342 xfs_warn(ip
->i_mount
,
1343 "corrupt dinode %Lu, (btree extents).",
1344 (unsigned long long) ip
->i_ino
);
1345 XFS_CORRUPTION_ERROR("xfs_bmap_read_extents(1)",
1346 XFS_ERRLEVEL_LOW
, ip
->i_mount
, block
);
1349 XFS_WANT_CORRUPTED_GOTO(
1350 xfs_bmap_sanity_check(mp
, bp
, 0),
1353 * Read-ahead the next leaf block, if any.
1355 nextbno
= be64_to_cpu(block
->bb_u
.l
.bb_rightsib
);
1356 if (nextbno
!= NULLFSBLOCK
)
1357 xfs_btree_reada_bufl(mp
, nextbno
, 1,
1360 * Copy records into the extent records.
1362 frp
= XFS_BMBT_REC_ADDR(mp
, block
, 1);
1364 for (j
= 0; j
< num_recs
; j
++, i
++, frp
++) {
1365 xfs_bmbt_rec_host_t
*trp
= xfs_iext_get_ext(ifp
, i
);
1366 trp
->l0
= be64_to_cpu(frp
->l0
);
1367 trp
->l1
= be64_to_cpu(frp
->l1
);
1369 if (exntf
== XFS_EXTFMT_NOSTATE
) {
1371 * Check all attribute bmap btree records and
1372 * any "older" data bmap btree records for a
1373 * set bit in the "extent flag" position.
1375 if (unlikely(xfs_check_nostate_extents(ifp
,
1376 start
, num_recs
))) {
1377 XFS_ERROR_REPORT("xfs_bmap_read_extents(2)",
1383 xfs_trans_brelse(tp
, bp
);
1386 * If we've reached the end, stop.
1388 if (bno
== NULLFSBLOCK
)
1390 error
= xfs_btree_read_bufl(mp
, tp
, bno
, 0, &bp
,
1391 XFS_BMAP_BTREE_REF
, &xfs_bmbt_buf_ops
);
1394 block
= XFS_BUF_TO_BLOCK(bp
);
1396 ASSERT(i
== (ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
)));
1397 ASSERT(i
== XFS_IFORK_NEXTENTS(ip
, whichfork
));
1398 XFS_BMAP_TRACE_EXLIST(ip
, i
, whichfork
);
1401 xfs_trans_brelse(tp
, bp
);
1402 return -EFSCORRUPTED
;
1407 * Search the extent records for the entry containing block bno.
1408 * If bno lies in a hole, point to the next entry. If bno lies
1409 * past eof, *eofp will be set, and *prevp will contain the last
1410 * entry (null if none). Else, *lastxp will be set to the index
1411 * of the found entry; *gotp will contain the entry.
1413 STATIC xfs_bmbt_rec_host_t
* /* pointer to found extent entry */
1414 xfs_bmap_search_multi_extents(
1415 xfs_ifork_t
*ifp
, /* inode fork pointer */
1416 xfs_fileoff_t bno
, /* block number searched for */
1417 int *eofp
, /* out: end of file found */
1418 xfs_extnum_t
*lastxp
, /* out: last extent index */
1419 xfs_bmbt_irec_t
*gotp
, /* out: extent entry found */
1420 xfs_bmbt_irec_t
*prevp
) /* out: previous extent entry found */
1422 xfs_bmbt_rec_host_t
*ep
; /* extent record pointer */
1423 xfs_extnum_t lastx
; /* last extent index */
1426 * Initialize the extent entry structure to catch access to
1427 * uninitialized br_startblock field.
1429 gotp
->br_startoff
= 0xffa5a5a5a5a5a5a5LL
;
1430 gotp
->br_blockcount
= 0xa55a5a5a5a5a5a5aLL
;
1431 gotp
->br_state
= XFS_EXT_INVALID
;
1432 gotp
->br_startblock
= 0xffffa5a5a5a5a5a5LL
;
1433 prevp
->br_startoff
= NULLFILEOFF
;
1435 ep
= xfs_iext_bno_to_ext(ifp
, bno
, &lastx
);
1437 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, lastx
- 1), prevp
);
1439 if (lastx
< (ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
))) {
1440 xfs_bmbt_get_all(ep
, gotp
);
1454 * Search the extents list for the inode, for the extent containing bno.
1455 * If bno lies in a hole, point to the next entry. If bno lies past eof,
1456 * *eofp will be set, and *prevp will contain the last entry (null if none).
1457 * Else, *lastxp will be set to the index of the found
1458 * entry; *gotp will contain the entry.
1460 STATIC xfs_bmbt_rec_host_t
* /* pointer to found extent entry */
1461 xfs_bmap_search_extents(
1462 xfs_inode_t
*ip
, /* incore inode pointer */
1463 xfs_fileoff_t bno
, /* block number searched for */
1464 int fork
, /* data or attr fork */
1465 int *eofp
, /* out: end of file found */
1466 xfs_extnum_t
*lastxp
, /* out: last extent index */
1467 xfs_bmbt_irec_t
*gotp
, /* out: extent entry found */
1468 xfs_bmbt_irec_t
*prevp
) /* out: previous extent entry found */
1470 xfs_ifork_t
*ifp
; /* inode fork pointer */
1471 xfs_bmbt_rec_host_t
*ep
; /* extent record pointer */
1473 XFS_STATS_INC(xs_look_exlist
);
1474 ifp
= XFS_IFORK_PTR(ip
, fork
);
1476 ep
= xfs_bmap_search_multi_extents(ifp
, bno
, eofp
, lastxp
, gotp
, prevp
);
1478 if (unlikely(!(gotp
->br_startblock
) && (*lastxp
!= NULLEXTNUM
) &&
1479 !(XFS_IS_REALTIME_INODE(ip
) && fork
== XFS_DATA_FORK
))) {
1480 xfs_alert_tag(ip
->i_mount
, XFS_PTAG_FSBLOCK_ZERO
,
1481 "Access to block zero in inode %llu "
1482 "start_block: %llx start_off: %llx "
1483 "blkcnt: %llx extent-state: %x lastx: %x",
1484 (unsigned long long)ip
->i_ino
,
1485 (unsigned long long)gotp
->br_startblock
,
1486 (unsigned long long)gotp
->br_startoff
,
1487 (unsigned long long)gotp
->br_blockcount
,
1488 gotp
->br_state
, *lastxp
);
1489 *lastxp
= NULLEXTNUM
;
1497 * Returns the file-relative block number of the first unused block(s)
1498 * in the file with at least "len" logically contiguous blocks free.
1499 * This is the lowest-address hole if the file has holes, else the first block
1500 * past the end of file.
1501 * Return 0 if the file is currently local (in-inode).
1504 xfs_bmap_first_unused(
1505 xfs_trans_t
*tp
, /* transaction pointer */
1506 xfs_inode_t
*ip
, /* incore inode */
1507 xfs_extlen_t len
, /* size of hole to find */
1508 xfs_fileoff_t
*first_unused
, /* unused block */
1509 int whichfork
) /* data or attr fork */
1511 int error
; /* error return value */
1512 int idx
; /* extent record index */
1513 xfs_ifork_t
*ifp
; /* inode fork pointer */
1514 xfs_fileoff_t lastaddr
; /* last block number seen */
1515 xfs_fileoff_t lowest
; /* lowest useful block */
1516 xfs_fileoff_t max
; /* starting useful block */
1517 xfs_fileoff_t off
; /* offset for this block */
1518 xfs_extnum_t nextents
; /* number of extent entries */
1520 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_BTREE
||
1521 XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_EXTENTS
||
1522 XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
);
1523 if (XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
) {
1527 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
1528 if (!(ifp
->if_flags
& XFS_IFEXTENTS
) &&
1529 (error
= xfs_iread_extents(tp
, ip
, whichfork
)))
1531 lowest
= *first_unused
;
1532 nextents
= ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
);
1533 for (idx
= 0, lastaddr
= 0, max
= lowest
; idx
< nextents
; idx
++) {
1534 xfs_bmbt_rec_host_t
*ep
= xfs_iext_get_ext(ifp
, idx
);
1535 off
= xfs_bmbt_get_startoff(ep
);
1537 * See if the hole before this extent will work.
1539 if (off
>= lowest
+ len
&& off
- max
>= len
) {
1540 *first_unused
= max
;
1543 lastaddr
= off
+ xfs_bmbt_get_blockcount(ep
);
1544 max
= XFS_FILEOFF_MAX(lastaddr
, lowest
);
1546 *first_unused
= max
;
1551 * Returns the file-relative block number of the last block - 1 before
1552 * last_block (input value) in the file.
1553 * This is not based on i_size, it is based on the extent records.
1554 * Returns 0 for local files, as they do not have extent records.
1557 xfs_bmap_last_before(
1558 xfs_trans_t
*tp
, /* transaction pointer */
1559 xfs_inode_t
*ip
, /* incore inode */
1560 xfs_fileoff_t
*last_block
, /* last block */
1561 int whichfork
) /* data or attr fork */
1563 xfs_fileoff_t bno
; /* input file offset */
1564 int eof
; /* hit end of file */
1565 xfs_bmbt_rec_host_t
*ep
; /* pointer to last extent */
1566 int error
; /* error return value */
1567 xfs_bmbt_irec_t got
; /* current extent value */
1568 xfs_ifork_t
*ifp
; /* inode fork pointer */
1569 xfs_extnum_t lastx
; /* last extent used */
1570 xfs_bmbt_irec_t prev
; /* previous extent value */
1572 if (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
&&
1573 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
&&
1574 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_LOCAL
)
1576 if (XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
) {
1580 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
1581 if (!(ifp
->if_flags
& XFS_IFEXTENTS
) &&
1582 (error
= xfs_iread_extents(tp
, ip
, whichfork
)))
1584 bno
= *last_block
- 1;
1585 ep
= xfs_bmap_search_extents(ip
, bno
, whichfork
, &eof
, &lastx
, &got
,
1587 if (eof
|| xfs_bmbt_get_startoff(ep
) > bno
) {
1588 if (prev
.br_startoff
== NULLFILEOFF
)
1591 *last_block
= prev
.br_startoff
+ prev
.br_blockcount
;
1594 * Otherwise *last_block is already the right answer.
1600 xfs_bmap_last_extent(
1601 struct xfs_trans
*tp
,
1602 struct xfs_inode
*ip
,
1604 struct xfs_bmbt_irec
*rec
,
1607 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(ip
, whichfork
);
1611 if (!(ifp
->if_flags
& XFS_IFEXTENTS
)) {
1612 error
= xfs_iread_extents(tp
, ip
, whichfork
);
1617 nextents
= ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
);
1618 if (nextents
== 0) {
1623 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, nextents
- 1), rec
);
1629 * Check the last inode extent to determine whether this allocation will result
1630 * in blocks being allocated at the end of the file. When we allocate new data
1631 * blocks at the end of the file which do not start at the previous data block,
1632 * we will try to align the new blocks at stripe unit boundaries.
1634 * Returns 1 in bma->aeof if the file (fork) is empty as any new write will be
1635 * at, or past the EOF.
1639 struct xfs_bmalloca
*bma
,
1642 struct xfs_bmbt_irec rec
;
1647 error
= xfs_bmap_last_extent(NULL
, bma
->ip
, whichfork
, &rec
,
1658 * Check if we are allocation or past the last extent, or at least into
1659 * the last delayed allocated extent.
1661 bma
->aeof
= bma
->offset
>= rec
.br_startoff
+ rec
.br_blockcount
||
1662 (bma
->offset
>= rec
.br_startoff
&&
1663 isnullstartblock(rec
.br_startblock
));
1668 * Returns the file-relative block number of the first block past eof in
1669 * the file. This is not based on i_size, it is based on the extent records.
1670 * Returns 0 for local files, as they do not have extent records.
1673 xfs_bmap_last_offset(
1674 struct xfs_inode
*ip
,
1675 xfs_fileoff_t
*last_block
,
1678 struct xfs_bmbt_irec rec
;
1684 if (XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_LOCAL
)
1687 if (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
&&
1688 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
)
1691 error
= xfs_bmap_last_extent(NULL
, ip
, whichfork
, &rec
, &is_empty
);
1692 if (error
|| is_empty
)
1695 *last_block
= rec
.br_startoff
+ rec
.br_blockcount
;
1700 * Returns whether the selected fork of the inode has exactly one
1701 * block or not. For the data fork we check this matches di_size,
1702 * implying the file's range is 0..bsize-1.
1704 int /* 1=>1 block, 0=>otherwise */
1706 xfs_inode_t
*ip
, /* incore inode */
1707 int whichfork
) /* data or attr fork */
1709 xfs_bmbt_rec_host_t
*ep
; /* ptr to fork's extent */
1710 xfs_ifork_t
*ifp
; /* inode fork pointer */
1711 int rval
; /* return value */
1712 xfs_bmbt_irec_t s
; /* internal version of extent */
1715 if (whichfork
== XFS_DATA_FORK
)
1716 return XFS_ISIZE(ip
) == ip
->i_mount
->m_sb
.sb_blocksize
;
1718 if (XFS_IFORK_NEXTENTS(ip
, whichfork
) != 1)
1720 if (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
)
1722 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
1723 ASSERT(ifp
->if_flags
& XFS_IFEXTENTS
);
1724 ep
= xfs_iext_get_ext(ifp
, 0);
1725 xfs_bmbt_get_all(ep
, &s
);
1726 rval
= s
.br_startoff
== 0 && s
.br_blockcount
== 1;
1727 if (rval
&& whichfork
== XFS_DATA_FORK
)
1728 ASSERT(XFS_ISIZE(ip
) == ip
->i_mount
->m_sb
.sb_blocksize
);
1733 * Extent tree manipulation functions used during allocation.
1737 * Convert a delayed allocation to a real allocation.
1739 STATIC
int /* error */
1740 xfs_bmap_add_extent_delay_real(
1741 struct xfs_bmalloca
*bma
)
1743 struct xfs_bmbt_irec
*new = &bma
->got
;
1744 int diff
; /* temp value */
1745 xfs_bmbt_rec_host_t
*ep
; /* extent entry for idx */
1746 int error
; /* error return value */
1747 int i
; /* temp state */
1748 xfs_ifork_t
*ifp
; /* inode fork pointer */
1749 xfs_fileoff_t new_endoff
; /* end offset of new entry */
1750 xfs_bmbt_irec_t r
[3]; /* neighbor extent entries */
1751 /* left is 0, right is 1, prev is 2 */
1752 int rval
=0; /* return value (logging flags) */
1753 int state
= 0;/* state bits, accessed thru macros */
1754 xfs_filblks_t da_new
; /* new count del alloc blocks used */
1755 xfs_filblks_t da_old
; /* old count del alloc blocks used */
1756 xfs_filblks_t temp
=0; /* value for da_new calculations */
1757 xfs_filblks_t temp2
=0;/* value for da_new calculations */
1758 int tmp_rval
; /* partial logging flags */
1760 ifp
= XFS_IFORK_PTR(bma
->ip
, XFS_DATA_FORK
);
1762 ASSERT(bma
->idx
>= 0);
1763 ASSERT(bma
->idx
<= ifp
->if_bytes
/ sizeof(struct xfs_bmbt_rec
));
1764 ASSERT(!isnullstartblock(new->br_startblock
));
1766 (bma
->cur
->bc_private
.b
.flags
& XFS_BTCUR_BPRV_WASDEL
));
1768 XFS_STATS_INC(xs_add_exlist
);
1775 * Set up a bunch of variables to make the tests simpler.
1777 ep
= xfs_iext_get_ext(ifp
, bma
->idx
);
1778 xfs_bmbt_get_all(ep
, &PREV
);
1779 new_endoff
= new->br_startoff
+ new->br_blockcount
;
1780 ASSERT(PREV
.br_startoff
<= new->br_startoff
);
1781 ASSERT(PREV
.br_startoff
+ PREV
.br_blockcount
>= new_endoff
);
1783 da_old
= startblockval(PREV
.br_startblock
);
1787 * Set flags determining what part of the previous delayed allocation
1788 * extent is being replaced by a real allocation.
1790 if (PREV
.br_startoff
== new->br_startoff
)
1791 state
|= BMAP_LEFT_FILLING
;
1792 if (PREV
.br_startoff
+ PREV
.br_blockcount
== new_endoff
)
1793 state
|= BMAP_RIGHT_FILLING
;
1796 * Check and set flags if this segment has a left neighbor.
1797 * Don't set contiguous if the combined extent would be too large.
1800 state
|= BMAP_LEFT_VALID
;
1801 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
- 1), &LEFT
);
1803 if (isnullstartblock(LEFT
.br_startblock
))
1804 state
|= BMAP_LEFT_DELAY
;
1807 if ((state
& BMAP_LEFT_VALID
) && !(state
& BMAP_LEFT_DELAY
) &&
1808 LEFT
.br_startoff
+ LEFT
.br_blockcount
== new->br_startoff
&&
1809 LEFT
.br_startblock
+ LEFT
.br_blockcount
== new->br_startblock
&&
1810 LEFT
.br_state
== new->br_state
&&
1811 LEFT
.br_blockcount
+ new->br_blockcount
<= MAXEXTLEN
)
1812 state
|= BMAP_LEFT_CONTIG
;
1815 * Check and set flags if this segment has a right neighbor.
1816 * Don't set contiguous if the combined extent would be too large.
1817 * Also check for all-three-contiguous being too large.
1819 if (bma
->idx
< bma
->ip
->i_df
.if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
) - 1) {
1820 state
|= BMAP_RIGHT_VALID
;
1821 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
+ 1), &RIGHT
);
1823 if (isnullstartblock(RIGHT
.br_startblock
))
1824 state
|= BMAP_RIGHT_DELAY
;
1827 if ((state
& BMAP_RIGHT_VALID
) && !(state
& BMAP_RIGHT_DELAY
) &&
1828 new_endoff
== RIGHT
.br_startoff
&&
1829 new->br_startblock
+ new->br_blockcount
== RIGHT
.br_startblock
&&
1830 new->br_state
== RIGHT
.br_state
&&
1831 new->br_blockcount
+ RIGHT
.br_blockcount
<= MAXEXTLEN
&&
1832 ((state
& (BMAP_LEFT_CONTIG
| BMAP_LEFT_FILLING
|
1833 BMAP_RIGHT_FILLING
)) !=
1834 (BMAP_LEFT_CONTIG
| BMAP_LEFT_FILLING
|
1835 BMAP_RIGHT_FILLING
) ||
1836 LEFT
.br_blockcount
+ new->br_blockcount
+ RIGHT
.br_blockcount
1838 state
|= BMAP_RIGHT_CONTIG
;
1842 * Switch out based on the FILLING and CONTIG state bits.
1844 switch (state
& (BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
|
1845 BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
)) {
1846 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
|
1847 BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
1849 * Filling in all of a previously delayed allocation extent.
1850 * The left and right neighbors are both contiguous with new.
1853 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1854 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, bma
->idx
),
1855 LEFT
.br_blockcount
+ PREV
.br_blockcount
+
1856 RIGHT
.br_blockcount
);
1857 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1859 xfs_iext_remove(bma
->ip
, bma
->idx
+ 1, 2, state
);
1860 bma
->ip
->i_d
.di_nextents
--;
1861 if (bma
->cur
== NULL
)
1862 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
1864 rval
= XFS_ILOG_CORE
;
1865 error
= xfs_bmbt_lookup_eq(bma
->cur
, RIGHT
.br_startoff
,
1866 RIGHT
.br_startblock
,
1867 RIGHT
.br_blockcount
, &i
);
1870 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1871 error
= xfs_btree_delete(bma
->cur
, &i
);
1874 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1875 error
= xfs_btree_decrement(bma
->cur
, 0, &i
);
1878 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1879 error
= xfs_bmbt_update(bma
->cur
, LEFT
.br_startoff
,
1881 LEFT
.br_blockcount
+
1882 PREV
.br_blockcount
+
1883 RIGHT
.br_blockcount
, LEFT
.br_state
);
1889 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
:
1891 * Filling in all of a previously delayed allocation extent.
1892 * The left neighbor is contiguous, the right is not.
1896 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1897 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, bma
->idx
),
1898 LEFT
.br_blockcount
+ PREV
.br_blockcount
);
1899 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1901 xfs_iext_remove(bma
->ip
, bma
->idx
+ 1, 1, state
);
1902 if (bma
->cur
== NULL
)
1903 rval
= XFS_ILOG_DEXT
;
1906 error
= xfs_bmbt_lookup_eq(bma
->cur
, LEFT
.br_startoff
,
1907 LEFT
.br_startblock
, LEFT
.br_blockcount
,
1911 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1912 error
= xfs_bmbt_update(bma
->cur
, LEFT
.br_startoff
,
1914 LEFT
.br_blockcount
+
1915 PREV
.br_blockcount
, LEFT
.br_state
);
1921 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
1923 * Filling in all of a previously delayed allocation extent.
1924 * The right neighbor is contiguous, the left is not.
1926 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1927 xfs_bmbt_set_startblock(ep
, new->br_startblock
);
1928 xfs_bmbt_set_blockcount(ep
,
1929 PREV
.br_blockcount
+ RIGHT
.br_blockcount
);
1930 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1932 xfs_iext_remove(bma
->ip
, bma
->idx
+ 1, 1, state
);
1933 if (bma
->cur
== NULL
)
1934 rval
= XFS_ILOG_DEXT
;
1937 error
= xfs_bmbt_lookup_eq(bma
->cur
, RIGHT
.br_startoff
,
1938 RIGHT
.br_startblock
,
1939 RIGHT
.br_blockcount
, &i
);
1942 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1943 error
= xfs_bmbt_update(bma
->cur
, PREV
.br_startoff
,
1945 PREV
.br_blockcount
+
1946 RIGHT
.br_blockcount
, PREV
.br_state
);
1952 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
:
1954 * Filling in all of a previously delayed allocation extent.
1955 * Neither the left nor right neighbors are contiguous with
1958 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1959 xfs_bmbt_set_startblock(ep
, new->br_startblock
);
1960 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1962 bma
->ip
->i_d
.di_nextents
++;
1963 if (bma
->cur
== NULL
)
1964 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
1966 rval
= XFS_ILOG_CORE
;
1967 error
= xfs_bmbt_lookup_eq(bma
->cur
, new->br_startoff
,
1968 new->br_startblock
, new->br_blockcount
,
1972 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
1973 bma
->cur
->bc_rec
.b
.br_state
= XFS_EXT_NORM
;
1974 error
= xfs_btree_insert(bma
->cur
, &i
);
1977 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
1981 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
:
1983 * Filling in the first part of a previous delayed allocation.
1984 * The left neighbor is contiguous.
1986 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
- 1, state
, _THIS_IP_
);
1987 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, bma
->idx
- 1),
1988 LEFT
.br_blockcount
+ new->br_blockcount
);
1989 xfs_bmbt_set_startoff(ep
,
1990 PREV
.br_startoff
+ new->br_blockcount
);
1991 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
- 1, state
, _THIS_IP_
);
1993 temp
= PREV
.br_blockcount
- new->br_blockcount
;
1994 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
1995 xfs_bmbt_set_blockcount(ep
, temp
);
1996 if (bma
->cur
== NULL
)
1997 rval
= XFS_ILOG_DEXT
;
2000 error
= xfs_bmbt_lookup_eq(bma
->cur
, LEFT
.br_startoff
,
2001 LEFT
.br_startblock
, LEFT
.br_blockcount
,
2005 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2006 error
= xfs_bmbt_update(bma
->cur
, LEFT
.br_startoff
,
2008 LEFT
.br_blockcount
+
2014 da_new
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(bma
->ip
, temp
),
2015 startblockval(PREV
.br_startblock
));
2016 xfs_bmbt_set_startblock(ep
, nullstartblock(da_new
));
2017 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2022 case BMAP_LEFT_FILLING
:
2024 * Filling in the first part of a previous delayed allocation.
2025 * The left neighbor is not contiguous.
2027 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2028 xfs_bmbt_set_startoff(ep
, new_endoff
);
2029 temp
= PREV
.br_blockcount
- new->br_blockcount
;
2030 xfs_bmbt_set_blockcount(ep
, temp
);
2031 xfs_iext_insert(bma
->ip
, bma
->idx
, 1, new, state
);
2032 bma
->ip
->i_d
.di_nextents
++;
2033 if (bma
->cur
== NULL
)
2034 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2036 rval
= XFS_ILOG_CORE
;
2037 error
= xfs_bmbt_lookup_eq(bma
->cur
, new->br_startoff
,
2038 new->br_startblock
, new->br_blockcount
,
2042 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
2043 bma
->cur
->bc_rec
.b
.br_state
= XFS_EXT_NORM
;
2044 error
= xfs_btree_insert(bma
->cur
, &i
);
2047 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2050 if (xfs_bmap_needs_btree(bma
->ip
, XFS_DATA_FORK
)) {
2051 error
= xfs_bmap_extents_to_btree(bma
->tp
, bma
->ip
,
2052 bma
->firstblock
, bma
->flist
,
2053 &bma
->cur
, 1, &tmp_rval
, XFS_DATA_FORK
);
2058 da_new
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(bma
->ip
, temp
),
2059 startblockval(PREV
.br_startblock
) -
2060 (bma
->cur
? bma
->cur
->bc_private
.b
.allocated
: 0));
2061 ep
= xfs_iext_get_ext(ifp
, bma
->idx
+ 1);
2062 xfs_bmbt_set_startblock(ep
, nullstartblock(da_new
));
2063 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
+ 1, state
, _THIS_IP_
);
2066 case BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
2068 * Filling in the last part of a previous delayed allocation.
2069 * The right neighbor is contiguous with the new allocation.
2071 temp
= PREV
.br_blockcount
- new->br_blockcount
;
2072 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
+ 1, state
, _THIS_IP_
);
2073 xfs_bmbt_set_blockcount(ep
, temp
);
2074 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp
, bma
->idx
+ 1),
2075 new->br_startoff
, new->br_startblock
,
2076 new->br_blockcount
+ RIGHT
.br_blockcount
,
2078 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
+ 1, state
, _THIS_IP_
);
2079 if (bma
->cur
== NULL
)
2080 rval
= XFS_ILOG_DEXT
;
2083 error
= xfs_bmbt_lookup_eq(bma
->cur
, RIGHT
.br_startoff
,
2084 RIGHT
.br_startblock
,
2085 RIGHT
.br_blockcount
, &i
);
2088 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2089 error
= xfs_bmbt_update(bma
->cur
, new->br_startoff
,
2091 new->br_blockcount
+
2092 RIGHT
.br_blockcount
,
2098 da_new
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(bma
->ip
, temp
),
2099 startblockval(PREV
.br_startblock
));
2100 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2101 xfs_bmbt_set_startblock(ep
, nullstartblock(da_new
));
2102 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2107 case BMAP_RIGHT_FILLING
:
2109 * Filling in the last part of a previous delayed allocation.
2110 * The right neighbor is not contiguous.
2112 temp
= PREV
.br_blockcount
- new->br_blockcount
;
2113 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2114 xfs_bmbt_set_blockcount(ep
, temp
);
2115 xfs_iext_insert(bma
->ip
, bma
->idx
+ 1, 1, new, state
);
2116 bma
->ip
->i_d
.di_nextents
++;
2117 if (bma
->cur
== NULL
)
2118 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2120 rval
= XFS_ILOG_CORE
;
2121 error
= xfs_bmbt_lookup_eq(bma
->cur
, new->br_startoff
,
2122 new->br_startblock
, new->br_blockcount
,
2126 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
2127 bma
->cur
->bc_rec
.b
.br_state
= XFS_EXT_NORM
;
2128 error
= xfs_btree_insert(bma
->cur
, &i
);
2131 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2134 if (xfs_bmap_needs_btree(bma
->ip
, XFS_DATA_FORK
)) {
2135 error
= xfs_bmap_extents_to_btree(bma
->tp
, bma
->ip
,
2136 bma
->firstblock
, bma
->flist
, &bma
->cur
, 1,
2137 &tmp_rval
, XFS_DATA_FORK
);
2142 da_new
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(bma
->ip
, temp
),
2143 startblockval(PREV
.br_startblock
) -
2144 (bma
->cur
? bma
->cur
->bc_private
.b
.allocated
: 0));
2145 ep
= xfs_iext_get_ext(ifp
, bma
->idx
);
2146 xfs_bmbt_set_startblock(ep
, nullstartblock(da_new
));
2147 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2154 * Filling in the middle part of a previous delayed allocation.
2155 * Contiguity is impossible here.
2156 * This case is avoided almost all the time.
2158 * We start with a delayed allocation:
2160 * +ddddddddddddddddddddddddddddddddddddddddddddddddddddddd+
2163 * and we are allocating:
2164 * +rrrrrrrrrrrrrrrrr+
2167 * and we set it up for insertion as:
2168 * +ddddddddddddddddddd+rrrrrrrrrrrrrrrrr+ddddddddddddddddd+
2170 * PREV @ idx LEFT RIGHT
2171 * inserted at idx + 1
2173 temp
= new->br_startoff
- PREV
.br_startoff
;
2174 temp2
= PREV
.br_startoff
+ PREV
.br_blockcount
- new_endoff
;
2175 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, 0, _THIS_IP_
);
2176 xfs_bmbt_set_blockcount(ep
, temp
); /* truncate PREV */
2178 RIGHT
.br_state
= PREV
.br_state
;
2179 RIGHT
.br_startblock
= nullstartblock(
2180 (int)xfs_bmap_worst_indlen(bma
->ip
, temp2
));
2181 RIGHT
.br_startoff
= new_endoff
;
2182 RIGHT
.br_blockcount
= temp2
;
2183 /* insert LEFT (r[0]) and RIGHT (r[1]) at the same time */
2184 xfs_iext_insert(bma
->ip
, bma
->idx
+ 1, 2, &LEFT
, state
);
2185 bma
->ip
->i_d
.di_nextents
++;
2186 if (bma
->cur
== NULL
)
2187 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2189 rval
= XFS_ILOG_CORE
;
2190 error
= xfs_bmbt_lookup_eq(bma
->cur
, new->br_startoff
,
2191 new->br_startblock
, new->br_blockcount
,
2195 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
2196 bma
->cur
->bc_rec
.b
.br_state
= XFS_EXT_NORM
;
2197 error
= xfs_btree_insert(bma
->cur
, &i
);
2200 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2203 if (xfs_bmap_needs_btree(bma
->ip
, XFS_DATA_FORK
)) {
2204 error
= xfs_bmap_extents_to_btree(bma
->tp
, bma
->ip
,
2205 bma
->firstblock
, bma
->flist
, &bma
->cur
,
2206 1, &tmp_rval
, XFS_DATA_FORK
);
2211 temp
= xfs_bmap_worst_indlen(bma
->ip
, temp
);
2212 temp2
= xfs_bmap_worst_indlen(bma
->ip
, temp2
);
2213 diff
= (int)(temp
+ temp2
- startblockval(PREV
.br_startblock
) -
2214 (bma
->cur
? bma
->cur
->bc_private
.b
.allocated
: 0));
2216 error
= xfs_icsb_modify_counters(bma
->ip
->i_mount
,
2218 -((int64_t)diff
), 0);
2224 ep
= xfs_iext_get_ext(ifp
, bma
->idx
);
2225 xfs_bmbt_set_startblock(ep
, nullstartblock((int)temp
));
2226 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
2227 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
+ 2, state
, _THIS_IP_
);
2228 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp
, bma
->idx
+ 2),
2229 nullstartblock((int)temp2
));
2230 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
+ 2, state
, _THIS_IP_
);
2233 da_new
= temp
+ temp2
;
2236 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2237 case BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2238 case BMAP_LEFT_FILLING
| BMAP_RIGHT_CONTIG
:
2239 case BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
:
2240 case BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2241 case BMAP_LEFT_CONTIG
:
2242 case BMAP_RIGHT_CONTIG
:
2244 * These cases are all impossible.
2249 /* convert to a btree if necessary */
2250 if (xfs_bmap_needs_btree(bma
->ip
, XFS_DATA_FORK
)) {
2251 int tmp_logflags
; /* partial log flag return val */
2253 ASSERT(bma
->cur
== NULL
);
2254 error
= xfs_bmap_extents_to_btree(bma
->tp
, bma
->ip
,
2255 bma
->firstblock
, bma
->flist
, &bma
->cur
,
2256 da_old
> 0, &tmp_logflags
, XFS_DATA_FORK
);
2257 bma
->logflags
|= tmp_logflags
;
2262 /* adjust for changes in reserved delayed indirect blocks */
2263 if (da_old
|| da_new
) {
2266 temp
+= bma
->cur
->bc_private
.b
.allocated
;
2267 ASSERT(temp
<= da_old
);
2269 xfs_icsb_modify_counters(bma
->ip
->i_mount
,
2271 (int64_t)(da_old
- temp
), 0);
2274 /* clear out the allocated field, done with it now in any case. */
2276 bma
->cur
->bc_private
.b
.allocated
= 0;
2278 xfs_bmap_check_leaf_extents(bma
->cur
, bma
->ip
, XFS_DATA_FORK
);
2280 bma
->logflags
|= rval
;
2288 * Convert an unwritten allocation to a real allocation or vice versa.
2290 STATIC
int /* error */
2291 xfs_bmap_add_extent_unwritten_real(
2292 struct xfs_trans
*tp
,
2293 xfs_inode_t
*ip
, /* incore inode pointer */
2294 xfs_extnum_t
*idx
, /* extent number to update/insert */
2295 xfs_btree_cur_t
**curp
, /* if *curp is null, not a btree */
2296 xfs_bmbt_irec_t
*new, /* new data to add to file extents */
2297 xfs_fsblock_t
*first
, /* pointer to firstblock variable */
2298 xfs_bmap_free_t
*flist
, /* list of extents to be freed */
2299 int *logflagsp
) /* inode logging flags */
2301 xfs_btree_cur_t
*cur
; /* btree cursor */
2302 xfs_bmbt_rec_host_t
*ep
; /* extent entry for idx */
2303 int error
; /* error return value */
2304 int i
; /* temp state */
2305 xfs_ifork_t
*ifp
; /* inode fork pointer */
2306 xfs_fileoff_t new_endoff
; /* end offset of new entry */
2307 xfs_exntst_t newext
; /* new extent state */
2308 xfs_exntst_t oldext
; /* old extent state */
2309 xfs_bmbt_irec_t r
[3]; /* neighbor extent entries */
2310 /* left is 0, right is 1, prev is 2 */
2311 int rval
=0; /* return value (logging flags) */
2312 int state
= 0;/* state bits, accessed thru macros */
2317 ifp
= XFS_IFORK_PTR(ip
, XFS_DATA_FORK
);
2320 ASSERT(*idx
<= ifp
->if_bytes
/ sizeof(struct xfs_bmbt_rec
));
2321 ASSERT(!isnullstartblock(new->br_startblock
));
2323 XFS_STATS_INC(xs_add_exlist
);
2330 * Set up a bunch of variables to make the tests simpler.
2333 ep
= xfs_iext_get_ext(ifp
, *idx
);
2334 xfs_bmbt_get_all(ep
, &PREV
);
2335 newext
= new->br_state
;
2336 oldext
= (newext
== XFS_EXT_UNWRITTEN
) ?
2337 XFS_EXT_NORM
: XFS_EXT_UNWRITTEN
;
2338 ASSERT(PREV
.br_state
== oldext
);
2339 new_endoff
= new->br_startoff
+ new->br_blockcount
;
2340 ASSERT(PREV
.br_startoff
<= new->br_startoff
);
2341 ASSERT(PREV
.br_startoff
+ PREV
.br_blockcount
>= new_endoff
);
2344 * Set flags determining what part of the previous oldext allocation
2345 * extent is being replaced by a newext allocation.
2347 if (PREV
.br_startoff
== new->br_startoff
)
2348 state
|= BMAP_LEFT_FILLING
;
2349 if (PREV
.br_startoff
+ PREV
.br_blockcount
== new_endoff
)
2350 state
|= BMAP_RIGHT_FILLING
;
2353 * Check and set flags if this segment has a left neighbor.
2354 * Don't set contiguous if the combined extent would be too large.
2357 state
|= BMAP_LEFT_VALID
;
2358 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, *idx
- 1), &LEFT
);
2360 if (isnullstartblock(LEFT
.br_startblock
))
2361 state
|= BMAP_LEFT_DELAY
;
2364 if ((state
& BMAP_LEFT_VALID
) && !(state
& BMAP_LEFT_DELAY
) &&
2365 LEFT
.br_startoff
+ LEFT
.br_blockcount
== new->br_startoff
&&
2366 LEFT
.br_startblock
+ LEFT
.br_blockcount
== new->br_startblock
&&
2367 LEFT
.br_state
== newext
&&
2368 LEFT
.br_blockcount
+ new->br_blockcount
<= MAXEXTLEN
)
2369 state
|= BMAP_LEFT_CONTIG
;
2372 * Check and set flags if this segment has a right neighbor.
2373 * Don't set contiguous if the combined extent would be too large.
2374 * Also check for all-three-contiguous being too large.
2376 if (*idx
< ip
->i_df
.if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
) - 1) {
2377 state
|= BMAP_RIGHT_VALID
;
2378 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, *idx
+ 1), &RIGHT
);
2379 if (isnullstartblock(RIGHT
.br_startblock
))
2380 state
|= BMAP_RIGHT_DELAY
;
2383 if ((state
& BMAP_RIGHT_VALID
) && !(state
& BMAP_RIGHT_DELAY
) &&
2384 new_endoff
== RIGHT
.br_startoff
&&
2385 new->br_startblock
+ new->br_blockcount
== RIGHT
.br_startblock
&&
2386 newext
== RIGHT
.br_state
&&
2387 new->br_blockcount
+ RIGHT
.br_blockcount
<= MAXEXTLEN
&&
2388 ((state
& (BMAP_LEFT_CONTIG
| BMAP_LEFT_FILLING
|
2389 BMAP_RIGHT_FILLING
)) !=
2390 (BMAP_LEFT_CONTIG
| BMAP_LEFT_FILLING
|
2391 BMAP_RIGHT_FILLING
) ||
2392 LEFT
.br_blockcount
+ new->br_blockcount
+ RIGHT
.br_blockcount
2394 state
|= BMAP_RIGHT_CONTIG
;
2397 * Switch out based on the FILLING and CONTIG state bits.
2399 switch (state
& (BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
|
2400 BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
)) {
2401 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
|
2402 BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
2404 * Setting all of a previous oldext extent to newext.
2405 * The left and right neighbors are both contiguous with new.
2409 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2410 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, *idx
),
2411 LEFT
.br_blockcount
+ PREV
.br_blockcount
+
2412 RIGHT
.br_blockcount
);
2413 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2415 xfs_iext_remove(ip
, *idx
+ 1, 2, state
);
2416 ip
->i_d
.di_nextents
-= 2;
2418 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2420 rval
= XFS_ILOG_CORE
;
2421 if ((error
= xfs_bmbt_lookup_eq(cur
, RIGHT
.br_startoff
,
2422 RIGHT
.br_startblock
,
2423 RIGHT
.br_blockcount
, &i
)))
2425 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2426 if ((error
= xfs_btree_delete(cur
, &i
)))
2428 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2429 if ((error
= xfs_btree_decrement(cur
, 0, &i
)))
2431 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2432 if ((error
= xfs_btree_delete(cur
, &i
)))
2434 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2435 if ((error
= xfs_btree_decrement(cur
, 0, &i
)))
2437 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2438 if ((error
= xfs_bmbt_update(cur
, LEFT
.br_startoff
,
2440 LEFT
.br_blockcount
+ PREV
.br_blockcount
+
2441 RIGHT
.br_blockcount
, LEFT
.br_state
)))
2446 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
:
2448 * Setting all of a previous oldext extent to newext.
2449 * The left neighbor is contiguous, the right is not.
2453 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2454 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, *idx
),
2455 LEFT
.br_blockcount
+ PREV
.br_blockcount
);
2456 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2458 xfs_iext_remove(ip
, *idx
+ 1, 1, state
);
2459 ip
->i_d
.di_nextents
--;
2461 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2463 rval
= XFS_ILOG_CORE
;
2464 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2465 PREV
.br_startblock
, PREV
.br_blockcount
,
2468 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2469 if ((error
= xfs_btree_delete(cur
, &i
)))
2471 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2472 if ((error
= xfs_btree_decrement(cur
, 0, &i
)))
2474 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2475 if ((error
= xfs_bmbt_update(cur
, LEFT
.br_startoff
,
2477 LEFT
.br_blockcount
+ PREV
.br_blockcount
,
2483 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
2485 * Setting all of a previous oldext extent to newext.
2486 * The right neighbor is contiguous, the left is not.
2488 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2489 xfs_bmbt_set_blockcount(ep
,
2490 PREV
.br_blockcount
+ RIGHT
.br_blockcount
);
2491 xfs_bmbt_set_state(ep
, newext
);
2492 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2493 xfs_iext_remove(ip
, *idx
+ 1, 1, state
);
2494 ip
->i_d
.di_nextents
--;
2496 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2498 rval
= XFS_ILOG_CORE
;
2499 if ((error
= xfs_bmbt_lookup_eq(cur
, RIGHT
.br_startoff
,
2500 RIGHT
.br_startblock
,
2501 RIGHT
.br_blockcount
, &i
)))
2503 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2504 if ((error
= xfs_btree_delete(cur
, &i
)))
2506 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2507 if ((error
= xfs_btree_decrement(cur
, 0, &i
)))
2509 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2510 if ((error
= xfs_bmbt_update(cur
, new->br_startoff
,
2512 new->br_blockcount
+ RIGHT
.br_blockcount
,
2518 case BMAP_LEFT_FILLING
| BMAP_RIGHT_FILLING
:
2520 * Setting all of a previous oldext extent to newext.
2521 * Neither the left nor right neighbors are contiguous with
2524 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2525 xfs_bmbt_set_state(ep
, newext
);
2526 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2529 rval
= XFS_ILOG_DEXT
;
2532 if ((error
= xfs_bmbt_lookup_eq(cur
, new->br_startoff
,
2533 new->br_startblock
, new->br_blockcount
,
2536 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2537 if ((error
= xfs_bmbt_update(cur
, new->br_startoff
,
2538 new->br_startblock
, new->br_blockcount
,
2544 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
:
2546 * Setting the first part of a previous oldext extent to newext.
2547 * The left neighbor is contiguous.
2549 trace_xfs_bmap_pre_update(ip
, *idx
- 1, state
, _THIS_IP_
);
2550 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, *idx
- 1),
2551 LEFT
.br_blockcount
+ new->br_blockcount
);
2552 xfs_bmbt_set_startoff(ep
,
2553 PREV
.br_startoff
+ new->br_blockcount
);
2554 trace_xfs_bmap_post_update(ip
, *idx
- 1, state
, _THIS_IP_
);
2556 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2557 xfs_bmbt_set_startblock(ep
,
2558 new->br_startblock
+ new->br_blockcount
);
2559 xfs_bmbt_set_blockcount(ep
,
2560 PREV
.br_blockcount
- new->br_blockcount
);
2561 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2566 rval
= XFS_ILOG_DEXT
;
2569 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2570 PREV
.br_startblock
, PREV
.br_blockcount
,
2573 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2574 if ((error
= xfs_bmbt_update(cur
,
2575 PREV
.br_startoff
+ new->br_blockcount
,
2576 PREV
.br_startblock
+ new->br_blockcount
,
2577 PREV
.br_blockcount
- new->br_blockcount
,
2580 if ((error
= xfs_btree_decrement(cur
, 0, &i
)))
2582 error
= xfs_bmbt_update(cur
, LEFT
.br_startoff
,
2584 LEFT
.br_blockcount
+ new->br_blockcount
,
2591 case BMAP_LEFT_FILLING
:
2593 * Setting the first part of a previous oldext extent to newext.
2594 * The left neighbor is not contiguous.
2596 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2597 ASSERT(ep
&& xfs_bmbt_get_state(ep
) == oldext
);
2598 xfs_bmbt_set_startoff(ep
, new_endoff
);
2599 xfs_bmbt_set_blockcount(ep
,
2600 PREV
.br_blockcount
- new->br_blockcount
);
2601 xfs_bmbt_set_startblock(ep
,
2602 new->br_startblock
+ new->br_blockcount
);
2603 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2605 xfs_iext_insert(ip
, *idx
, 1, new, state
);
2606 ip
->i_d
.di_nextents
++;
2608 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2610 rval
= XFS_ILOG_CORE
;
2611 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2612 PREV
.br_startblock
, PREV
.br_blockcount
,
2615 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2616 if ((error
= xfs_bmbt_update(cur
,
2617 PREV
.br_startoff
+ new->br_blockcount
,
2618 PREV
.br_startblock
+ new->br_blockcount
,
2619 PREV
.br_blockcount
- new->br_blockcount
,
2622 cur
->bc_rec
.b
= *new;
2623 if ((error
= xfs_btree_insert(cur
, &i
)))
2625 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2629 case BMAP_RIGHT_FILLING
| BMAP_RIGHT_CONTIG
:
2631 * Setting the last part of a previous oldext extent to newext.
2632 * The right neighbor is contiguous with the new allocation.
2634 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2635 xfs_bmbt_set_blockcount(ep
,
2636 PREV
.br_blockcount
- new->br_blockcount
);
2637 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2641 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2642 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp
, *idx
),
2643 new->br_startoff
, new->br_startblock
,
2644 new->br_blockcount
+ RIGHT
.br_blockcount
, newext
);
2645 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2648 rval
= XFS_ILOG_DEXT
;
2651 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2653 PREV
.br_blockcount
, &i
)))
2655 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2656 if ((error
= xfs_bmbt_update(cur
, PREV
.br_startoff
,
2658 PREV
.br_blockcount
- new->br_blockcount
,
2661 if ((error
= xfs_btree_increment(cur
, 0, &i
)))
2663 if ((error
= xfs_bmbt_update(cur
, new->br_startoff
,
2665 new->br_blockcount
+ RIGHT
.br_blockcount
,
2671 case BMAP_RIGHT_FILLING
:
2673 * Setting the last part of a previous oldext extent to newext.
2674 * The right neighbor is not contiguous.
2676 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2677 xfs_bmbt_set_blockcount(ep
,
2678 PREV
.br_blockcount
- new->br_blockcount
);
2679 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2682 xfs_iext_insert(ip
, *idx
, 1, new, state
);
2684 ip
->i_d
.di_nextents
++;
2686 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2688 rval
= XFS_ILOG_CORE
;
2689 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2690 PREV
.br_startblock
, PREV
.br_blockcount
,
2693 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2694 if ((error
= xfs_bmbt_update(cur
, PREV
.br_startoff
,
2696 PREV
.br_blockcount
- new->br_blockcount
,
2699 if ((error
= xfs_bmbt_lookup_eq(cur
, new->br_startoff
,
2700 new->br_startblock
, new->br_blockcount
,
2703 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
2704 cur
->bc_rec
.b
.br_state
= XFS_EXT_NORM
;
2705 if ((error
= xfs_btree_insert(cur
, &i
)))
2707 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2713 * Setting the middle part of a previous oldext extent to
2714 * newext. Contiguity is impossible here.
2715 * One extent becomes three extents.
2717 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2718 xfs_bmbt_set_blockcount(ep
,
2719 new->br_startoff
- PREV
.br_startoff
);
2720 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2723 r
[1].br_startoff
= new_endoff
;
2724 r
[1].br_blockcount
=
2725 PREV
.br_startoff
+ PREV
.br_blockcount
- new_endoff
;
2726 r
[1].br_startblock
= new->br_startblock
+ new->br_blockcount
;
2727 r
[1].br_state
= oldext
;
2730 xfs_iext_insert(ip
, *idx
, 2, &r
[0], state
);
2732 ip
->i_d
.di_nextents
+= 2;
2734 rval
= XFS_ILOG_CORE
| XFS_ILOG_DEXT
;
2736 rval
= XFS_ILOG_CORE
;
2737 if ((error
= xfs_bmbt_lookup_eq(cur
, PREV
.br_startoff
,
2738 PREV
.br_startblock
, PREV
.br_blockcount
,
2741 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2742 /* new right extent - oldext */
2743 if ((error
= xfs_bmbt_update(cur
, r
[1].br_startoff
,
2744 r
[1].br_startblock
, r
[1].br_blockcount
,
2747 /* new left extent - oldext */
2748 cur
->bc_rec
.b
= PREV
;
2749 cur
->bc_rec
.b
.br_blockcount
=
2750 new->br_startoff
- PREV
.br_startoff
;
2751 if ((error
= xfs_btree_insert(cur
, &i
)))
2753 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2755 * Reset the cursor to the position of the new extent
2756 * we are about to insert as we can't trust it after
2757 * the previous insert.
2759 if ((error
= xfs_bmbt_lookup_eq(cur
, new->br_startoff
,
2760 new->br_startblock
, new->br_blockcount
,
2763 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
2764 /* new middle extent - newext */
2765 cur
->bc_rec
.b
.br_state
= new->br_state
;
2766 if ((error
= xfs_btree_insert(cur
, &i
)))
2768 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
2772 case BMAP_LEFT_FILLING
| BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2773 case BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2774 case BMAP_LEFT_FILLING
| BMAP_RIGHT_CONTIG
:
2775 case BMAP_RIGHT_FILLING
| BMAP_LEFT_CONTIG
:
2776 case BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2777 case BMAP_LEFT_CONTIG
:
2778 case BMAP_RIGHT_CONTIG
:
2780 * These cases are all impossible.
2785 /* convert to a btree if necessary */
2786 if (xfs_bmap_needs_btree(ip
, XFS_DATA_FORK
)) {
2787 int tmp_logflags
; /* partial log flag return val */
2789 ASSERT(cur
== NULL
);
2790 error
= xfs_bmap_extents_to_btree(tp
, ip
, first
, flist
, &cur
,
2791 0, &tmp_logflags
, XFS_DATA_FORK
);
2792 *logflagsp
|= tmp_logflags
;
2797 /* clear out the allocated field, done with it now in any case. */
2799 cur
->bc_private
.b
.allocated
= 0;
2803 xfs_bmap_check_leaf_extents(*curp
, ip
, XFS_DATA_FORK
);
2813 * Convert a hole to a delayed allocation.
2816 xfs_bmap_add_extent_hole_delay(
2817 xfs_inode_t
*ip
, /* incore inode pointer */
2818 xfs_extnum_t
*idx
, /* extent number to update/insert */
2819 xfs_bmbt_irec_t
*new) /* new data to add to file extents */
2821 xfs_ifork_t
*ifp
; /* inode fork pointer */
2822 xfs_bmbt_irec_t left
; /* left neighbor extent entry */
2823 xfs_filblks_t newlen
=0; /* new indirect size */
2824 xfs_filblks_t oldlen
=0; /* old indirect size */
2825 xfs_bmbt_irec_t right
; /* right neighbor extent entry */
2826 int state
; /* state bits, accessed thru macros */
2827 xfs_filblks_t temp
=0; /* temp for indirect calculations */
2829 ifp
= XFS_IFORK_PTR(ip
, XFS_DATA_FORK
);
2831 ASSERT(isnullstartblock(new->br_startblock
));
2834 * Check and set flags if this segment has a left neighbor
2837 state
|= BMAP_LEFT_VALID
;
2838 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, *idx
- 1), &left
);
2840 if (isnullstartblock(left
.br_startblock
))
2841 state
|= BMAP_LEFT_DELAY
;
2845 * Check and set flags if the current (right) segment exists.
2846 * If it doesn't exist, we're converting the hole at end-of-file.
2848 if (*idx
< ip
->i_df
.if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
)) {
2849 state
|= BMAP_RIGHT_VALID
;
2850 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, *idx
), &right
);
2852 if (isnullstartblock(right
.br_startblock
))
2853 state
|= BMAP_RIGHT_DELAY
;
2857 * Set contiguity flags on the left and right neighbors.
2858 * Don't let extents get too large, even if the pieces are contiguous.
2860 if ((state
& BMAP_LEFT_VALID
) && (state
& BMAP_LEFT_DELAY
) &&
2861 left
.br_startoff
+ left
.br_blockcount
== new->br_startoff
&&
2862 left
.br_blockcount
+ new->br_blockcount
<= MAXEXTLEN
)
2863 state
|= BMAP_LEFT_CONTIG
;
2865 if ((state
& BMAP_RIGHT_VALID
) && (state
& BMAP_RIGHT_DELAY
) &&
2866 new->br_startoff
+ new->br_blockcount
== right
.br_startoff
&&
2867 new->br_blockcount
+ right
.br_blockcount
<= MAXEXTLEN
&&
2868 (!(state
& BMAP_LEFT_CONTIG
) ||
2869 (left
.br_blockcount
+ new->br_blockcount
+
2870 right
.br_blockcount
<= MAXEXTLEN
)))
2871 state
|= BMAP_RIGHT_CONTIG
;
2874 * Switch out based on the contiguity flags.
2876 switch (state
& (BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
)) {
2877 case BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
2879 * New allocation is contiguous with delayed allocations
2880 * on the left and on the right.
2881 * Merge all three into a single extent record.
2884 temp
= left
.br_blockcount
+ new->br_blockcount
+
2885 right
.br_blockcount
;
2887 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2888 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, *idx
), temp
);
2889 oldlen
= startblockval(left
.br_startblock
) +
2890 startblockval(new->br_startblock
) +
2891 startblockval(right
.br_startblock
);
2892 newlen
= xfs_bmap_worst_indlen(ip
, temp
);
2893 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp
, *idx
),
2894 nullstartblock((int)newlen
));
2895 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2897 xfs_iext_remove(ip
, *idx
+ 1, 1, state
);
2900 case BMAP_LEFT_CONTIG
:
2902 * New allocation is contiguous with a delayed allocation
2904 * Merge the new allocation with the left neighbor.
2907 temp
= left
.br_blockcount
+ new->br_blockcount
;
2909 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2910 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, *idx
), temp
);
2911 oldlen
= startblockval(left
.br_startblock
) +
2912 startblockval(new->br_startblock
);
2913 newlen
= xfs_bmap_worst_indlen(ip
, temp
);
2914 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp
, *idx
),
2915 nullstartblock((int)newlen
));
2916 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2919 case BMAP_RIGHT_CONTIG
:
2921 * New allocation is contiguous with a delayed allocation
2923 * Merge the new allocation with the right neighbor.
2925 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
2926 temp
= new->br_blockcount
+ right
.br_blockcount
;
2927 oldlen
= startblockval(new->br_startblock
) +
2928 startblockval(right
.br_startblock
);
2929 newlen
= xfs_bmap_worst_indlen(ip
, temp
);
2930 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp
, *idx
),
2932 nullstartblock((int)newlen
), temp
, right
.br_state
);
2933 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
2938 * New allocation is not contiguous with another
2939 * delayed allocation.
2940 * Insert a new entry.
2942 oldlen
= newlen
= 0;
2943 xfs_iext_insert(ip
, *idx
, 1, new, state
);
2946 if (oldlen
!= newlen
) {
2947 ASSERT(oldlen
> newlen
);
2948 xfs_icsb_modify_counters(ip
->i_mount
, XFS_SBS_FDBLOCKS
,
2949 (int64_t)(oldlen
- newlen
), 0);
2951 * Nothing to do for disk quota accounting here.
2957 * Convert a hole to a real allocation.
2959 STATIC
int /* error */
2960 xfs_bmap_add_extent_hole_real(
2961 struct xfs_bmalloca
*bma
,
2964 struct xfs_bmbt_irec
*new = &bma
->got
;
2965 int error
; /* error return value */
2966 int i
; /* temp state */
2967 xfs_ifork_t
*ifp
; /* inode fork pointer */
2968 xfs_bmbt_irec_t left
; /* left neighbor extent entry */
2969 xfs_bmbt_irec_t right
; /* right neighbor extent entry */
2970 int rval
=0; /* return value (logging flags) */
2971 int state
; /* state bits, accessed thru macros */
2973 ifp
= XFS_IFORK_PTR(bma
->ip
, whichfork
);
2975 ASSERT(bma
->idx
>= 0);
2976 ASSERT(bma
->idx
<= ifp
->if_bytes
/ sizeof(struct xfs_bmbt_rec
));
2977 ASSERT(!isnullstartblock(new->br_startblock
));
2979 !(bma
->cur
->bc_private
.b
.flags
& XFS_BTCUR_BPRV_WASDEL
));
2981 XFS_STATS_INC(xs_add_exlist
);
2984 if (whichfork
== XFS_ATTR_FORK
)
2985 state
|= BMAP_ATTRFORK
;
2988 * Check and set flags if this segment has a left neighbor.
2991 state
|= BMAP_LEFT_VALID
;
2992 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
- 1), &left
);
2993 if (isnullstartblock(left
.br_startblock
))
2994 state
|= BMAP_LEFT_DELAY
;
2998 * Check and set flags if this segment has a current value.
2999 * Not true if we're inserting into the "hole" at eof.
3001 if (bma
->idx
< ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
)) {
3002 state
|= BMAP_RIGHT_VALID
;
3003 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
), &right
);
3004 if (isnullstartblock(right
.br_startblock
))
3005 state
|= BMAP_RIGHT_DELAY
;
3009 * We're inserting a real allocation between "left" and "right".
3010 * Set the contiguity flags. Don't let extents get too large.
3012 if ((state
& BMAP_LEFT_VALID
) && !(state
& BMAP_LEFT_DELAY
) &&
3013 left
.br_startoff
+ left
.br_blockcount
== new->br_startoff
&&
3014 left
.br_startblock
+ left
.br_blockcount
== new->br_startblock
&&
3015 left
.br_state
== new->br_state
&&
3016 left
.br_blockcount
+ new->br_blockcount
<= MAXEXTLEN
)
3017 state
|= BMAP_LEFT_CONTIG
;
3019 if ((state
& BMAP_RIGHT_VALID
) && !(state
& BMAP_RIGHT_DELAY
) &&
3020 new->br_startoff
+ new->br_blockcount
== right
.br_startoff
&&
3021 new->br_startblock
+ new->br_blockcount
== right
.br_startblock
&&
3022 new->br_state
== right
.br_state
&&
3023 new->br_blockcount
+ right
.br_blockcount
<= MAXEXTLEN
&&
3024 (!(state
& BMAP_LEFT_CONTIG
) ||
3025 left
.br_blockcount
+ new->br_blockcount
+
3026 right
.br_blockcount
<= MAXEXTLEN
))
3027 state
|= BMAP_RIGHT_CONTIG
;
3031 * Select which case we're in here, and implement it.
3033 switch (state
& (BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
)) {
3034 case BMAP_LEFT_CONTIG
| BMAP_RIGHT_CONTIG
:
3036 * New allocation is contiguous with real allocations on the
3037 * left and on the right.
3038 * Merge all three into a single extent record.
3041 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3042 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, bma
->idx
),
3043 left
.br_blockcount
+ new->br_blockcount
+
3044 right
.br_blockcount
);
3045 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3047 xfs_iext_remove(bma
->ip
, bma
->idx
+ 1, 1, state
);
3049 XFS_IFORK_NEXT_SET(bma
->ip
, whichfork
,
3050 XFS_IFORK_NEXTENTS(bma
->ip
, whichfork
) - 1);
3051 if (bma
->cur
== NULL
) {
3052 rval
= XFS_ILOG_CORE
| xfs_ilog_fext(whichfork
);
3054 rval
= XFS_ILOG_CORE
;
3055 error
= xfs_bmbt_lookup_eq(bma
->cur
, right
.br_startoff
,
3056 right
.br_startblock
, right
.br_blockcount
,
3060 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3061 error
= xfs_btree_delete(bma
->cur
, &i
);
3064 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3065 error
= xfs_btree_decrement(bma
->cur
, 0, &i
);
3068 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3069 error
= xfs_bmbt_update(bma
->cur
, left
.br_startoff
,
3071 left
.br_blockcount
+
3072 new->br_blockcount
+
3073 right
.br_blockcount
,
3080 case BMAP_LEFT_CONTIG
:
3082 * New allocation is contiguous with a real allocation
3084 * Merge the new allocation with the left neighbor.
3087 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3088 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp
, bma
->idx
),
3089 left
.br_blockcount
+ new->br_blockcount
);
3090 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3092 if (bma
->cur
== NULL
) {
3093 rval
= xfs_ilog_fext(whichfork
);
3096 error
= xfs_bmbt_lookup_eq(bma
->cur
, left
.br_startoff
,
3097 left
.br_startblock
, left
.br_blockcount
,
3101 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3102 error
= xfs_bmbt_update(bma
->cur
, left
.br_startoff
,
3104 left
.br_blockcount
+
3112 case BMAP_RIGHT_CONTIG
:
3114 * New allocation is contiguous with a real allocation
3116 * Merge the new allocation with the right neighbor.
3118 trace_xfs_bmap_pre_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3119 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp
, bma
->idx
),
3120 new->br_startoff
, new->br_startblock
,
3121 new->br_blockcount
+ right
.br_blockcount
,
3123 trace_xfs_bmap_post_update(bma
->ip
, bma
->idx
, state
, _THIS_IP_
);
3125 if (bma
->cur
== NULL
) {
3126 rval
= xfs_ilog_fext(whichfork
);
3129 error
= xfs_bmbt_lookup_eq(bma
->cur
,
3131 right
.br_startblock
,
3132 right
.br_blockcount
, &i
);
3135 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3136 error
= xfs_bmbt_update(bma
->cur
, new->br_startoff
,
3138 new->br_blockcount
+
3139 right
.br_blockcount
,
3148 * New allocation is not contiguous with another
3150 * Insert a new entry.
3152 xfs_iext_insert(bma
->ip
, bma
->idx
, 1, new, state
);
3153 XFS_IFORK_NEXT_SET(bma
->ip
, whichfork
,
3154 XFS_IFORK_NEXTENTS(bma
->ip
, whichfork
) + 1);
3155 if (bma
->cur
== NULL
) {
3156 rval
= XFS_ILOG_CORE
| xfs_ilog_fext(whichfork
);
3158 rval
= XFS_ILOG_CORE
;
3159 error
= xfs_bmbt_lookup_eq(bma
->cur
,
3162 new->br_blockcount
, &i
);
3165 XFS_WANT_CORRUPTED_GOTO(i
== 0, done
);
3166 bma
->cur
->bc_rec
.b
.br_state
= new->br_state
;
3167 error
= xfs_btree_insert(bma
->cur
, &i
);
3170 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
3175 /* convert to a btree if necessary */
3176 if (xfs_bmap_needs_btree(bma
->ip
, whichfork
)) {
3177 int tmp_logflags
; /* partial log flag return val */
3179 ASSERT(bma
->cur
== NULL
);
3180 error
= xfs_bmap_extents_to_btree(bma
->tp
, bma
->ip
,
3181 bma
->firstblock
, bma
->flist
, &bma
->cur
,
3182 0, &tmp_logflags
, whichfork
);
3183 bma
->logflags
|= tmp_logflags
;
3188 /* clear out the allocated field, done with it now in any case. */
3190 bma
->cur
->bc_private
.b
.allocated
= 0;
3192 xfs_bmap_check_leaf_extents(bma
->cur
, bma
->ip
, whichfork
);
3194 bma
->logflags
|= rval
;
3199 * Functions used in the extent read, allocate and remove paths
3203 * Adjust the size of the new extent based on di_extsize and rt extsize.
3206 xfs_bmap_extsize_align(
3208 xfs_bmbt_irec_t
*gotp
, /* next extent pointer */
3209 xfs_bmbt_irec_t
*prevp
, /* previous extent pointer */
3210 xfs_extlen_t extsz
, /* align to this extent size */
3211 int rt
, /* is this a realtime inode? */
3212 int eof
, /* is extent at end-of-file? */
3213 int delay
, /* creating delalloc extent? */
3214 int convert
, /* overwriting unwritten extent? */
3215 xfs_fileoff_t
*offp
, /* in/out: aligned offset */
3216 xfs_extlen_t
*lenp
) /* in/out: aligned length */
3218 xfs_fileoff_t orig_off
; /* original offset */
3219 xfs_extlen_t orig_alen
; /* original length */
3220 xfs_fileoff_t orig_end
; /* original off+len */
3221 xfs_fileoff_t nexto
; /* next file offset */
3222 xfs_fileoff_t prevo
; /* previous file offset */
3223 xfs_fileoff_t align_off
; /* temp for offset */
3224 xfs_extlen_t align_alen
; /* temp for length */
3225 xfs_extlen_t temp
; /* temp for calculations */
3230 orig_off
= align_off
= *offp
;
3231 orig_alen
= align_alen
= *lenp
;
3232 orig_end
= orig_off
+ orig_alen
;
3235 * If this request overlaps an existing extent, then don't
3236 * attempt to perform any additional alignment.
3238 if (!delay
&& !eof
&&
3239 (orig_off
>= gotp
->br_startoff
) &&
3240 (orig_end
<= gotp
->br_startoff
+ gotp
->br_blockcount
)) {
3245 * If the file offset is unaligned vs. the extent size
3246 * we need to align it. This will be possible unless
3247 * the file was previously written with a kernel that didn't
3248 * perform this alignment, or if a truncate shot us in the
3251 temp
= do_mod(orig_off
, extsz
);
3257 * Same adjustment for the end of the requested area.
3259 if ((temp
= (align_alen
% extsz
))) {
3260 align_alen
+= extsz
- temp
;
3263 * If the previous block overlaps with this proposed allocation
3264 * then move the start forward without adjusting the length.
3266 if (prevp
->br_startoff
!= NULLFILEOFF
) {
3267 if (prevp
->br_startblock
== HOLESTARTBLOCK
)
3268 prevo
= prevp
->br_startoff
;
3270 prevo
= prevp
->br_startoff
+ prevp
->br_blockcount
;
3273 if (align_off
!= orig_off
&& align_off
< prevo
)
3276 * If the next block overlaps with this proposed allocation
3277 * then move the start back without adjusting the length,
3278 * but not before offset 0.
3279 * This may of course make the start overlap previous block,
3280 * and if we hit the offset 0 limit then the next block
3281 * can still overlap too.
3283 if (!eof
&& gotp
->br_startoff
!= NULLFILEOFF
) {
3284 if ((delay
&& gotp
->br_startblock
== HOLESTARTBLOCK
) ||
3285 (!delay
&& gotp
->br_startblock
== DELAYSTARTBLOCK
))
3286 nexto
= gotp
->br_startoff
+ gotp
->br_blockcount
;
3288 nexto
= gotp
->br_startoff
;
3290 nexto
= NULLFILEOFF
;
3292 align_off
+ align_alen
!= orig_end
&&
3293 align_off
+ align_alen
> nexto
)
3294 align_off
= nexto
> align_alen
? nexto
- align_alen
: 0;
3296 * If we're now overlapping the next or previous extent that
3297 * means we can't fit an extsz piece in this hole. Just move
3298 * the start forward to the first valid spot and set
3299 * the length so we hit the end.
3301 if (align_off
!= orig_off
&& align_off
< prevo
)
3303 if (align_off
+ align_alen
!= orig_end
&&
3304 align_off
+ align_alen
> nexto
&&
3305 nexto
!= NULLFILEOFF
) {
3306 ASSERT(nexto
> prevo
);
3307 align_alen
= nexto
- align_off
;
3311 * If realtime, and the result isn't a multiple of the realtime
3312 * extent size we need to remove blocks until it is.
3314 if (rt
&& (temp
= (align_alen
% mp
->m_sb
.sb_rextsize
))) {
3316 * We're not covering the original request, or
3317 * we won't be able to once we fix the length.
3319 if (orig_off
< align_off
||
3320 orig_end
> align_off
+ align_alen
||
3321 align_alen
- temp
< orig_alen
)
3324 * Try to fix it by moving the start up.
3326 if (align_off
+ temp
<= orig_off
) {
3331 * Try to fix it by moving the end in.
3333 else if (align_off
+ align_alen
- temp
>= orig_end
)
3336 * Set the start to the minimum then trim the length.
3339 align_alen
-= orig_off
- align_off
;
3340 align_off
= orig_off
;
3341 align_alen
-= align_alen
% mp
->m_sb
.sb_rextsize
;
3344 * Result doesn't cover the request, fail it.
3346 if (orig_off
< align_off
|| orig_end
> align_off
+ align_alen
)
3349 ASSERT(orig_off
>= align_off
);
3350 ASSERT(orig_end
<= align_off
+ align_alen
);
3354 if (!eof
&& gotp
->br_startoff
!= NULLFILEOFF
)
3355 ASSERT(align_off
+ align_alen
<= gotp
->br_startoff
);
3356 if (prevp
->br_startoff
!= NULLFILEOFF
)
3357 ASSERT(align_off
>= prevp
->br_startoff
+ prevp
->br_blockcount
);
3365 #define XFS_ALLOC_GAP_UNITS 4
3369 struct xfs_bmalloca
*ap
) /* bmap alloc argument struct */
3371 xfs_fsblock_t adjust
; /* adjustment to block numbers */
3372 xfs_agnumber_t fb_agno
; /* ag number of ap->firstblock */
3373 xfs_mount_t
*mp
; /* mount point structure */
3374 int nullfb
; /* true if ap->firstblock isn't set */
3375 int rt
; /* true if inode is realtime */
3377 #define ISVALID(x,y) \
3379 (x) < mp->m_sb.sb_rblocks : \
3380 XFS_FSB_TO_AGNO(mp, x) == XFS_FSB_TO_AGNO(mp, y) && \
3381 XFS_FSB_TO_AGNO(mp, x) < mp->m_sb.sb_agcount && \
3382 XFS_FSB_TO_AGBNO(mp, x) < mp->m_sb.sb_agblocks)
3384 mp
= ap
->ip
->i_mount
;
3385 nullfb
= *ap
->firstblock
== NULLFSBLOCK
;
3386 rt
= XFS_IS_REALTIME_INODE(ap
->ip
) && ap
->userdata
;
3387 fb_agno
= nullfb
? NULLAGNUMBER
: XFS_FSB_TO_AGNO(mp
, *ap
->firstblock
);
3389 * If allocating at eof, and there's a previous real block,
3390 * try to use its last block as our starting point.
3392 if (ap
->eof
&& ap
->prev
.br_startoff
!= NULLFILEOFF
&&
3393 !isnullstartblock(ap
->prev
.br_startblock
) &&
3394 ISVALID(ap
->prev
.br_startblock
+ ap
->prev
.br_blockcount
,
3395 ap
->prev
.br_startblock
)) {
3396 ap
->blkno
= ap
->prev
.br_startblock
+ ap
->prev
.br_blockcount
;
3398 * Adjust for the gap between prevp and us.
3400 adjust
= ap
->offset
-
3401 (ap
->prev
.br_startoff
+ ap
->prev
.br_blockcount
);
3403 ISVALID(ap
->blkno
+ adjust
, ap
->prev
.br_startblock
))
3404 ap
->blkno
+= adjust
;
3407 * If not at eof, then compare the two neighbor blocks.
3408 * Figure out whether either one gives us a good starting point,
3409 * and pick the better one.
3411 else if (!ap
->eof
) {
3412 xfs_fsblock_t gotbno
; /* right side block number */
3413 xfs_fsblock_t gotdiff
=0; /* right side difference */
3414 xfs_fsblock_t prevbno
; /* left side block number */
3415 xfs_fsblock_t prevdiff
=0; /* left side difference */
3418 * If there's a previous (left) block, select a requested
3419 * start block based on it.
3421 if (ap
->prev
.br_startoff
!= NULLFILEOFF
&&
3422 !isnullstartblock(ap
->prev
.br_startblock
) &&
3423 (prevbno
= ap
->prev
.br_startblock
+
3424 ap
->prev
.br_blockcount
) &&
3425 ISVALID(prevbno
, ap
->prev
.br_startblock
)) {
3427 * Calculate gap to end of previous block.
3429 adjust
= prevdiff
= ap
->offset
-
3430 (ap
->prev
.br_startoff
+
3431 ap
->prev
.br_blockcount
);
3433 * Figure the startblock based on the previous block's
3434 * end and the gap size.
3436 * If the gap is large relative to the piece we're
3437 * allocating, or using it gives us an invalid block
3438 * number, then just use the end of the previous block.
3440 if (prevdiff
<= XFS_ALLOC_GAP_UNITS
* ap
->length
&&
3441 ISVALID(prevbno
+ prevdiff
,
3442 ap
->prev
.br_startblock
))
3447 * If the firstblock forbids it, can't use it,
3450 if (!rt
&& !nullfb
&&
3451 XFS_FSB_TO_AGNO(mp
, prevbno
) != fb_agno
)
3452 prevbno
= NULLFSBLOCK
;
3455 * No previous block or can't follow it, just default.
3458 prevbno
= NULLFSBLOCK
;
3460 * If there's a following (right) block, select a requested
3461 * start block based on it.
3463 if (!isnullstartblock(ap
->got
.br_startblock
)) {
3465 * Calculate gap to start of next block.
3467 adjust
= gotdiff
= ap
->got
.br_startoff
- ap
->offset
;
3469 * Figure the startblock based on the next block's
3470 * start and the gap size.
3472 gotbno
= ap
->got
.br_startblock
;
3475 * If the gap is large relative to the piece we're
3476 * allocating, or using it gives us an invalid block
3477 * number, then just use the start of the next block
3478 * offset by our length.
3480 if (gotdiff
<= XFS_ALLOC_GAP_UNITS
* ap
->length
&&
3481 ISVALID(gotbno
- gotdiff
, gotbno
))
3483 else if (ISVALID(gotbno
- ap
->length
, gotbno
)) {
3484 gotbno
-= ap
->length
;
3485 gotdiff
+= adjust
- ap
->length
;
3489 * If the firstblock forbids it, can't use it,
3492 if (!rt
&& !nullfb
&&
3493 XFS_FSB_TO_AGNO(mp
, gotbno
) != fb_agno
)
3494 gotbno
= NULLFSBLOCK
;
3497 * No next block, just default.
3500 gotbno
= NULLFSBLOCK
;
3502 * If both valid, pick the better one, else the only good
3503 * one, else ap->blkno is already set (to 0 or the inode block).
3505 if (prevbno
!= NULLFSBLOCK
&& gotbno
!= NULLFSBLOCK
)
3506 ap
->blkno
= prevdiff
<= gotdiff
? prevbno
: gotbno
;
3507 else if (prevbno
!= NULLFSBLOCK
)
3508 ap
->blkno
= prevbno
;
3509 else if (gotbno
!= NULLFSBLOCK
)
3516 xfs_bmap_longest_free_extent(
3517 struct xfs_trans
*tp
,
3522 struct xfs_mount
*mp
= tp
->t_mountp
;
3523 struct xfs_perag
*pag
;
3524 xfs_extlen_t longest
;
3527 pag
= xfs_perag_get(mp
, ag
);
3528 if (!pag
->pagf_init
) {
3529 error
= xfs_alloc_pagf_init(mp
, tp
, ag
, XFS_ALLOC_FLAG_TRYLOCK
);
3533 if (!pag
->pagf_init
) {
3539 longest
= xfs_alloc_longest_free_extent(mp
, pag
);
3540 if (*blen
< longest
)
3549 xfs_bmap_select_minlen(
3550 struct xfs_bmalloca
*ap
,
3551 struct xfs_alloc_arg
*args
,
3555 if (notinit
|| *blen
< ap
->minlen
) {
3557 * Since we did a BUF_TRYLOCK above, it is possible that
3558 * there is space for this request.
3560 args
->minlen
= ap
->minlen
;
3561 } else if (*blen
< args
->maxlen
) {
3563 * If the best seen length is less than the request length,
3564 * use the best as the minimum.
3566 args
->minlen
= *blen
;
3569 * Otherwise we've seen an extent as big as maxlen, use that
3572 args
->minlen
= args
->maxlen
;
3577 xfs_bmap_btalloc_nullfb(
3578 struct xfs_bmalloca
*ap
,
3579 struct xfs_alloc_arg
*args
,
3582 struct xfs_mount
*mp
= ap
->ip
->i_mount
;
3583 xfs_agnumber_t ag
, startag
;
3587 args
->type
= XFS_ALLOCTYPE_START_BNO
;
3588 args
->total
= ap
->total
;
3590 startag
= ag
= XFS_FSB_TO_AGNO(mp
, args
->fsbno
);
3591 if (startag
== NULLAGNUMBER
)
3594 while (*blen
< args
->maxlen
) {
3595 error
= xfs_bmap_longest_free_extent(args
->tp
, ag
, blen
,
3600 if (++ag
== mp
->m_sb
.sb_agcount
)
3606 xfs_bmap_select_minlen(ap
, args
, blen
, notinit
);
3611 xfs_bmap_btalloc_filestreams(
3612 struct xfs_bmalloca
*ap
,
3613 struct xfs_alloc_arg
*args
,
3616 struct xfs_mount
*mp
= ap
->ip
->i_mount
;
3621 args
->type
= XFS_ALLOCTYPE_NEAR_BNO
;
3622 args
->total
= ap
->total
;
3624 ag
= XFS_FSB_TO_AGNO(mp
, args
->fsbno
);
3625 if (ag
== NULLAGNUMBER
)
3628 error
= xfs_bmap_longest_free_extent(args
->tp
, ag
, blen
, ¬init
);
3632 if (*blen
< args
->maxlen
) {
3633 error
= xfs_filestream_new_ag(ap
, &ag
);
3637 error
= xfs_bmap_longest_free_extent(args
->tp
, ag
, blen
,
3644 xfs_bmap_select_minlen(ap
, args
, blen
, notinit
);
3647 * Set the failure fallback case to look in the selected AG as stream
3650 ap
->blkno
= args
->fsbno
= XFS_AGB_TO_FSB(mp
, ag
, 0);
3656 struct xfs_bmalloca
*ap
) /* bmap alloc argument struct */
3658 xfs_mount_t
*mp
; /* mount point structure */
3659 xfs_alloctype_t atype
= 0; /* type for allocation routines */
3660 xfs_extlen_t align
; /* minimum allocation alignment */
3661 xfs_agnumber_t fb_agno
; /* ag number of ap->firstblock */
3663 xfs_alloc_arg_t args
;
3665 xfs_extlen_t nextminlen
= 0;
3666 int nullfb
; /* true if ap->firstblock isn't set */
3674 mp
= ap
->ip
->i_mount
;
3676 /* stripe alignment for allocation is determined by mount parameters */
3678 if (mp
->m_swidth
&& (mp
->m_flags
& XFS_MOUNT_SWALLOC
))
3679 stripe_align
= mp
->m_swidth
;
3680 else if (mp
->m_dalign
)
3681 stripe_align
= mp
->m_dalign
;
3683 align
= ap
->userdata
? xfs_get_extsz_hint(ap
->ip
) : 0;
3684 if (unlikely(align
)) {
3685 error
= xfs_bmap_extsize_align(mp
, &ap
->got
, &ap
->prev
,
3686 align
, 0, ap
->eof
, 0, ap
->conv
,
3687 &ap
->offset
, &ap
->length
);
3693 nullfb
= *ap
->firstblock
== NULLFSBLOCK
;
3694 fb_agno
= nullfb
? NULLAGNUMBER
: XFS_FSB_TO_AGNO(mp
, *ap
->firstblock
);
3696 if (ap
->userdata
&& xfs_inode_is_filestream(ap
->ip
)) {
3697 ag
= xfs_filestream_lookup_ag(ap
->ip
);
3698 ag
= (ag
!= NULLAGNUMBER
) ? ag
: 0;
3699 ap
->blkno
= XFS_AGB_TO_FSB(mp
, ag
, 0);
3701 ap
->blkno
= XFS_INO_TO_FSB(mp
, ap
->ip
->i_ino
);
3704 ap
->blkno
= *ap
->firstblock
;
3706 xfs_bmap_adjacent(ap
);
3709 * If allowed, use ap->blkno; otherwise must use firstblock since
3710 * it's in the right allocation group.
3712 if (nullfb
|| XFS_FSB_TO_AGNO(mp
, ap
->blkno
) == fb_agno
)
3715 ap
->blkno
= *ap
->firstblock
;
3717 * Normal allocation, done through xfs_alloc_vextent.
3719 tryagain
= isaligned
= 0;
3720 memset(&args
, 0, sizeof(args
));
3723 args
.fsbno
= ap
->blkno
;
3725 /* Trim the allocation back to the maximum an AG can fit. */
3726 args
.maxlen
= MIN(ap
->length
, XFS_ALLOC_AG_MAX_USABLE(mp
));
3727 args
.firstblock
= *ap
->firstblock
;
3731 * Search for an allocation group with a single extent large
3732 * enough for the request. If one isn't found, then adjust
3733 * the minimum allocation size to the largest space found.
3735 if (ap
->userdata
&& xfs_inode_is_filestream(ap
->ip
))
3736 error
= xfs_bmap_btalloc_filestreams(ap
, &args
, &blen
);
3738 error
= xfs_bmap_btalloc_nullfb(ap
, &args
, &blen
);
3741 } else if (ap
->flist
->xbf_low
) {
3742 if (xfs_inode_is_filestream(ap
->ip
))
3743 args
.type
= XFS_ALLOCTYPE_FIRST_AG
;
3745 args
.type
= XFS_ALLOCTYPE_START_BNO
;
3746 args
.total
= args
.minlen
= ap
->minlen
;
3748 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
3749 args
.total
= ap
->total
;
3750 args
.minlen
= ap
->minlen
;
3752 /* apply extent size hints if obtained earlier */
3753 if (unlikely(align
)) {
3755 if ((args
.mod
= (xfs_extlen_t
)do_mod(ap
->offset
, args
.prod
)))
3756 args
.mod
= (xfs_extlen_t
)(args
.prod
- args
.mod
);
3757 } else if (mp
->m_sb
.sb_blocksize
>= PAGE_CACHE_SIZE
) {
3761 args
.prod
= PAGE_CACHE_SIZE
>> mp
->m_sb
.sb_blocklog
;
3762 if ((args
.mod
= (xfs_extlen_t
)(do_mod(ap
->offset
, args
.prod
))))
3763 args
.mod
= (xfs_extlen_t
)(args
.prod
- args
.mod
);
3766 * If we are not low on available data blocks, and the
3767 * underlying logical volume manager is a stripe, and
3768 * the file offset is zero then try to allocate data
3769 * blocks on stripe unit boundary.
3770 * NOTE: ap->aeof is only set if the allocation length
3771 * is >= the stripe unit and the allocation offset is
3772 * at the end of file.
3774 if (!ap
->flist
->xbf_low
&& ap
->aeof
) {
3776 args
.alignment
= stripe_align
;
3780 * Adjust for alignment
3782 if (blen
> args
.alignment
&& blen
<= args
.maxlen
)
3783 args
.minlen
= blen
- args
.alignment
;
3784 args
.minalignslop
= 0;
3787 * First try an exact bno allocation.
3788 * If it fails then do a near or start bno
3789 * allocation with alignment turned on.
3793 args
.type
= XFS_ALLOCTYPE_THIS_BNO
;
3796 * Compute the minlen+alignment for the
3797 * next case. Set slop so that the value
3798 * of minlen+alignment+slop doesn't go up
3799 * between the calls.
3801 if (blen
> stripe_align
&& blen
<= args
.maxlen
)
3802 nextminlen
= blen
- stripe_align
;
3804 nextminlen
= args
.minlen
;
3805 if (nextminlen
+ stripe_align
> args
.minlen
+ 1)
3807 nextminlen
+ stripe_align
-
3810 args
.minalignslop
= 0;
3814 args
.minalignslop
= 0;
3816 args
.minleft
= ap
->minleft
;
3817 args
.wasdel
= ap
->wasdel
;
3819 args
.userdata
= ap
->userdata
;
3820 if ((error
= xfs_alloc_vextent(&args
)))
3822 if (tryagain
&& args
.fsbno
== NULLFSBLOCK
) {
3824 * Exact allocation failed. Now try with alignment
3828 args
.fsbno
= ap
->blkno
;
3829 args
.alignment
= stripe_align
;
3830 args
.minlen
= nextminlen
;
3831 args
.minalignslop
= 0;
3833 if ((error
= xfs_alloc_vextent(&args
)))
3836 if (isaligned
&& args
.fsbno
== NULLFSBLOCK
) {
3838 * allocation failed, so turn off alignment and
3842 args
.fsbno
= ap
->blkno
;
3844 if ((error
= xfs_alloc_vextent(&args
)))
3847 if (args
.fsbno
== NULLFSBLOCK
&& nullfb
&&
3848 args
.minlen
> ap
->minlen
) {
3849 args
.minlen
= ap
->minlen
;
3850 args
.type
= XFS_ALLOCTYPE_START_BNO
;
3851 args
.fsbno
= ap
->blkno
;
3852 if ((error
= xfs_alloc_vextent(&args
)))
3855 if (args
.fsbno
== NULLFSBLOCK
&& nullfb
) {
3857 args
.type
= XFS_ALLOCTYPE_FIRST_AG
;
3858 args
.total
= ap
->minlen
;
3860 if ((error
= xfs_alloc_vextent(&args
)))
3862 ap
->flist
->xbf_low
= 1;
3864 if (args
.fsbno
!= NULLFSBLOCK
) {
3866 * check the allocation happened at the same or higher AG than
3867 * the first block that was allocated.
3869 ASSERT(*ap
->firstblock
== NULLFSBLOCK
||
3870 XFS_FSB_TO_AGNO(mp
, *ap
->firstblock
) ==
3871 XFS_FSB_TO_AGNO(mp
, args
.fsbno
) ||
3872 (ap
->flist
->xbf_low
&&
3873 XFS_FSB_TO_AGNO(mp
, *ap
->firstblock
) <
3874 XFS_FSB_TO_AGNO(mp
, args
.fsbno
)));
3876 ap
->blkno
= args
.fsbno
;
3877 if (*ap
->firstblock
== NULLFSBLOCK
)
3878 *ap
->firstblock
= args
.fsbno
;
3879 ASSERT(nullfb
|| fb_agno
== args
.agno
||
3880 (ap
->flist
->xbf_low
&& fb_agno
< args
.agno
));
3881 ap
->length
= args
.len
;
3882 ap
->ip
->i_d
.di_nblocks
+= args
.len
;
3883 xfs_trans_log_inode(ap
->tp
, ap
->ip
, XFS_ILOG_CORE
);
3885 ap
->ip
->i_delayed_blks
-= args
.len
;
3887 * Adjust the disk quota also. This was reserved
3890 xfs_trans_mod_dquot_byino(ap
->tp
, ap
->ip
,
3891 ap
->wasdel
? XFS_TRANS_DQ_DELBCOUNT
:
3892 XFS_TRANS_DQ_BCOUNT
,
3895 ap
->blkno
= NULLFSBLOCK
;
3902 * xfs_bmap_alloc is called by xfs_bmapi to allocate an extent for a file.
3903 * It figures out where to ask the underlying allocator to put the new extent.
3907 struct xfs_bmalloca
*ap
) /* bmap alloc argument struct */
3909 if (XFS_IS_REALTIME_INODE(ap
->ip
) && ap
->userdata
)
3910 return xfs_bmap_rtalloc(ap
);
3911 return xfs_bmap_btalloc(ap
);
3915 * Trim the returned map to the required bounds
3919 struct xfs_bmbt_irec
*mval
,
3920 struct xfs_bmbt_irec
*got
,
3928 if ((flags
& XFS_BMAPI_ENTIRE
) ||
3929 got
->br_startoff
+ got
->br_blockcount
<= obno
) {
3931 if (isnullstartblock(got
->br_startblock
))
3932 mval
->br_startblock
= DELAYSTARTBLOCK
;
3938 ASSERT((*bno
>= obno
) || (n
== 0));
3940 mval
->br_startoff
= *bno
;
3941 if (isnullstartblock(got
->br_startblock
))
3942 mval
->br_startblock
= DELAYSTARTBLOCK
;
3944 mval
->br_startblock
= got
->br_startblock
+
3945 (*bno
- got
->br_startoff
);
3947 * Return the minimum of what we got and what we asked for for
3948 * the length. We can use the len variable here because it is
3949 * modified below and we could have been there before coming
3950 * here if the first part of the allocation didn't overlap what
3953 mval
->br_blockcount
= XFS_FILBLKS_MIN(end
- *bno
,
3954 got
->br_blockcount
- (*bno
- got
->br_startoff
));
3955 mval
->br_state
= got
->br_state
;
3956 ASSERT(mval
->br_blockcount
<= len
);
3961 * Update and validate the extent map to return
3964 xfs_bmapi_update_map(
3965 struct xfs_bmbt_irec
**map
,
3973 xfs_bmbt_irec_t
*mval
= *map
;
3975 ASSERT((flags
& XFS_BMAPI_ENTIRE
) ||
3976 ((mval
->br_startoff
+ mval
->br_blockcount
) <= end
));
3977 ASSERT((flags
& XFS_BMAPI_ENTIRE
) || (mval
->br_blockcount
<= *len
) ||
3978 (mval
->br_startoff
< obno
));
3980 *bno
= mval
->br_startoff
+ mval
->br_blockcount
;
3982 if (*n
> 0 && mval
->br_startoff
== mval
[-1].br_startoff
) {
3983 /* update previous map with new information */
3984 ASSERT(mval
->br_startblock
== mval
[-1].br_startblock
);
3985 ASSERT(mval
->br_blockcount
> mval
[-1].br_blockcount
);
3986 ASSERT(mval
->br_state
== mval
[-1].br_state
);
3987 mval
[-1].br_blockcount
= mval
->br_blockcount
;
3988 mval
[-1].br_state
= mval
->br_state
;
3989 } else if (*n
> 0 && mval
->br_startblock
!= DELAYSTARTBLOCK
&&
3990 mval
[-1].br_startblock
!= DELAYSTARTBLOCK
&&
3991 mval
[-1].br_startblock
!= HOLESTARTBLOCK
&&
3992 mval
->br_startblock
== mval
[-1].br_startblock
+
3993 mval
[-1].br_blockcount
&&
3994 ((flags
& XFS_BMAPI_IGSTATE
) ||
3995 mval
[-1].br_state
== mval
->br_state
)) {
3996 ASSERT(mval
->br_startoff
==
3997 mval
[-1].br_startoff
+ mval
[-1].br_blockcount
);
3998 mval
[-1].br_blockcount
+= mval
->br_blockcount
;
3999 } else if (*n
> 0 &&
4000 mval
->br_startblock
== DELAYSTARTBLOCK
&&
4001 mval
[-1].br_startblock
== DELAYSTARTBLOCK
&&
4002 mval
->br_startoff
==
4003 mval
[-1].br_startoff
+ mval
[-1].br_blockcount
) {
4004 mval
[-1].br_blockcount
+= mval
->br_blockcount
;
4005 mval
[-1].br_state
= mval
->br_state
;
4006 } else if (!((*n
== 0) &&
4007 ((mval
->br_startoff
+ mval
->br_blockcount
) <=
4016 * Map file blocks to filesystem blocks without allocation.
4020 struct xfs_inode
*ip
,
4023 struct xfs_bmbt_irec
*mval
,
4027 struct xfs_mount
*mp
= ip
->i_mount
;
4028 struct xfs_ifork
*ifp
;
4029 struct xfs_bmbt_irec got
;
4030 struct xfs_bmbt_irec prev
;
4037 int whichfork
= (flags
& XFS_BMAPI_ATTRFORK
) ?
4038 XFS_ATTR_FORK
: XFS_DATA_FORK
;
4041 ASSERT(!(flags
& ~(XFS_BMAPI_ATTRFORK
|XFS_BMAPI_ENTIRE
|
4042 XFS_BMAPI_IGSTATE
)));
4043 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_SHARED
|XFS_ILOCK_EXCL
));
4045 if (unlikely(XFS_TEST_ERROR(
4046 (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
&&
4047 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
),
4048 mp
, XFS_ERRTAG_BMAPIFORMAT
, XFS_RANDOM_BMAPIFORMAT
))) {
4049 XFS_ERROR_REPORT("xfs_bmapi_read", XFS_ERRLEVEL_LOW
, mp
);
4050 return -EFSCORRUPTED
;
4053 if (XFS_FORCED_SHUTDOWN(mp
))
4056 XFS_STATS_INC(xs_blk_mapr
);
4058 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
4060 if (!(ifp
->if_flags
& XFS_IFEXTENTS
)) {
4061 error
= xfs_iread_extents(NULL
, ip
, whichfork
);
4066 xfs_bmap_search_extents(ip
, bno
, whichfork
, &eof
, &lastx
, &got
, &prev
);
4070 while (bno
< end
&& n
< *nmap
) {
4071 /* Reading past eof, act as though there's a hole up to end. */
4073 got
.br_startoff
= end
;
4074 if (got
.br_startoff
> bno
) {
4075 /* Reading in a hole. */
4076 mval
->br_startoff
= bno
;
4077 mval
->br_startblock
= HOLESTARTBLOCK
;
4078 mval
->br_blockcount
=
4079 XFS_FILBLKS_MIN(len
, got
.br_startoff
- bno
);
4080 mval
->br_state
= XFS_EXT_NORM
;
4081 bno
+= mval
->br_blockcount
;
4082 len
-= mval
->br_blockcount
;
4088 /* set up the extent map to return. */
4089 xfs_bmapi_trim_map(mval
, &got
, &bno
, len
, obno
, end
, n
, flags
);
4090 xfs_bmapi_update_map(&mval
, &bno
, &len
, obno
, end
, &n
, flags
);
4092 /* If we're done, stop now. */
4093 if (bno
>= end
|| n
>= *nmap
)
4096 /* Else go on to the next record. */
4097 if (++lastx
< ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
))
4098 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, lastx
), &got
);
4107 xfs_bmapi_reserve_delalloc(
4108 struct xfs_inode
*ip
,
4111 struct xfs_bmbt_irec
*got
,
4112 struct xfs_bmbt_irec
*prev
,
4113 xfs_extnum_t
*lastx
,
4116 struct xfs_mount
*mp
= ip
->i_mount
;
4117 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(ip
, XFS_DATA_FORK
);
4119 xfs_extlen_t indlen
;
4120 char rt
= XFS_IS_REALTIME_INODE(ip
);
4124 alen
= XFS_FILBLKS_MIN(len
, MAXEXTLEN
);
4126 alen
= XFS_FILBLKS_MIN(alen
, got
->br_startoff
- aoff
);
4128 /* Figure out the extent size, adjust alen */
4129 extsz
= xfs_get_extsz_hint(ip
);
4132 * Make sure we don't exceed a single extent length when we
4133 * align the extent by reducing length we are going to
4134 * allocate by the maximum amount extent size aligment may
4137 alen
= XFS_FILBLKS_MIN(len
, MAXEXTLEN
- (2 * extsz
- 1));
4138 error
= xfs_bmap_extsize_align(mp
, got
, prev
, extsz
, rt
, eof
,
4139 1, 0, &aoff
, &alen
);
4144 extsz
= alen
/ mp
->m_sb
.sb_rextsize
;
4147 * Make a transaction-less quota reservation for delayed allocation
4148 * blocks. This number gets adjusted later. We return if we haven't
4149 * allocated blocks already inside this loop.
4151 error
= xfs_trans_reserve_quota_nblks(NULL
, ip
, (long)alen
, 0,
4152 rt
? XFS_QMOPT_RES_RTBLKS
: XFS_QMOPT_RES_REGBLKS
);
4157 * Split changing sb for alen and indlen since they could be coming
4158 * from different places.
4160 indlen
= (xfs_extlen_t
)xfs_bmap_worst_indlen(ip
, alen
);
4164 error
= xfs_mod_incore_sb(mp
, XFS_SBS_FREXTENTS
,
4165 -((int64_t)extsz
), 0);
4167 error
= xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
,
4168 -((int64_t)alen
), 0);
4172 goto out_unreserve_quota
;
4174 error
= xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
,
4175 -((int64_t)indlen
), 0);
4177 goto out_unreserve_blocks
;
4180 ip
->i_delayed_blks
+= alen
;
4182 got
->br_startoff
= aoff
;
4183 got
->br_startblock
= nullstartblock(indlen
);
4184 got
->br_blockcount
= alen
;
4185 got
->br_state
= XFS_EXT_NORM
;
4186 xfs_bmap_add_extent_hole_delay(ip
, lastx
, got
);
4189 * Update our extent pointer, given that xfs_bmap_add_extent_hole_delay
4190 * might have merged it into one of the neighbouring ones.
4192 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, *lastx
), got
);
4194 ASSERT(got
->br_startoff
<= aoff
);
4195 ASSERT(got
->br_startoff
+ got
->br_blockcount
>= aoff
+ alen
);
4196 ASSERT(isnullstartblock(got
->br_startblock
));
4197 ASSERT(got
->br_state
== XFS_EXT_NORM
);
4200 out_unreserve_blocks
:
4202 xfs_mod_incore_sb(mp
, XFS_SBS_FREXTENTS
, extsz
, 0);
4204 xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
, alen
, 0);
4205 out_unreserve_quota
:
4206 if (XFS_IS_QUOTA_ON(mp
))
4207 xfs_trans_unreserve_quota_nblks(NULL
, ip
, (long)alen
, 0, rt
?
4208 XFS_QMOPT_RES_RTBLKS
: XFS_QMOPT_RES_REGBLKS
);
4213 * Map file blocks to filesystem blocks, adding delayed allocations as needed.
4217 struct xfs_inode
*ip
, /* incore inode */
4218 xfs_fileoff_t bno
, /* starting file offs. mapped */
4219 xfs_filblks_t len
, /* length to map in file */
4220 struct xfs_bmbt_irec
*mval
, /* output: map values */
4221 int *nmap
, /* i/o: mval size/count */
4222 int flags
) /* XFS_BMAPI_... */
4224 struct xfs_mount
*mp
= ip
->i_mount
;
4225 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(ip
, XFS_DATA_FORK
);
4226 struct xfs_bmbt_irec got
; /* current file extent record */
4227 struct xfs_bmbt_irec prev
; /* previous file extent record */
4228 xfs_fileoff_t obno
; /* old block number (offset) */
4229 xfs_fileoff_t end
; /* end of mapped file region */
4230 xfs_extnum_t lastx
; /* last useful extent number */
4231 int eof
; /* we've hit the end of extents */
4232 int n
= 0; /* current extent index */
4236 ASSERT(*nmap
<= XFS_BMAP_MAX_NMAP
);
4237 ASSERT(!(flags
& ~XFS_BMAPI_ENTIRE
));
4238 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_EXCL
));
4240 if (unlikely(XFS_TEST_ERROR(
4241 (XFS_IFORK_FORMAT(ip
, XFS_DATA_FORK
) != XFS_DINODE_FMT_EXTENTS
&&
4242 XFS_IFORK_FORMAT(ip
, XFS_DATA_FORK
) != XFS_DINODE_FMT_BTREE
),
4243 mp
, XFS_ERRTAG_BMAPIFORMAT
, XFS_RANDOM_BMAPIFORMAT
))) {
4244 XFS_ERROR_REPORT("xfs_bmapi_delay", XFS_ERRLEVEL_LOW
, mp
);
4245 return -EFSCORRUPTED
;
4248 if (XFS_FORCED_SHUTDOWN(mp
))
4251 XFS_STATS_INC(xs_blk_mapw
);
4253 if (!(ifp
->if_flags
& XFS_IFEXTENTS
)) {
4254 error
= xfs_iread_extents(NULL
, ip
, XFS_DATA_FORK
);
4259 xfs_bmap_search_extents(ip
, bno
, XFS_DATA_FORK
, &eof
, &lastx
, &got
, &prev
);
4263 while (bno
< end
&& n
< *nmap
) {
4264 if (eof
|| got
.br_startoff
> bno
) {
4265 error
= xfs_bmapi_reserve_delalloc(ip
, bno
, len
, &got
,
4266 &prev
, &lastx
, eof
);
4276 /* set up the extent map to return. */
4277 xfs_bmapi_trim_map(mval
, &got
, &bno
, len
, obno
, end
, n
, flags
);
4278 xfs_bmapi_update_map(&mval
, &bno
, &len
, obno
, end
, &n
, flags
);
4280 /* If we're done, stop now. */
4281 if (bno
>= end
|| n
>= *nmap
)
4284 /* Else go on to the next record. */
4286 if (++lastx
< ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
))
4287 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, lastx
), &got
);
4299 struct xfs_bmalloca
*bma
)
4301 struct xfs_mount
*mp
= bma
->ip
->i_mount
;
4302 int whichfork
= (bma
->flags
& XFS_BMAPI_ATTRFORK
) ?
4303 XFS_ATTR_FORK
: XFS_DATA_FORK
;
4304 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(bma
->ip
, whichfork
);
4305 int tmp_logflags
= 0;
4308 ASSERT(bma
->length
> 0);
4311 * For the wasdelay case, we could also just allocate the stuff asked
4312 * for in this bmap call but that wouldn't be as good.
4315 bma
->length
= (xfs_extlen_t
)bma
->got
.br_blockcount
;
4316 bma
->offset
= bma
->got
.br_startoff
;
4317 if (bma
->idx
!= NULLEXTNUM
&& bma
->idx
) {
4318 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
- 1),
4322 bma
->length
= XFS_FILBLKS_MIN(bma
->length
, MAXEXTLEN
);
4324 bma
->length
= XFS_FILBLKS_MIN(bma
->length
,
4325 bma
->got
.br_startoff
- bma
->offset
);
4329 * Indicate if this is the first user data in the file, or just any
4332 if (!(bma
->flags
& XFS_BMAPI_METADATA
)) {
4333 bma
->userdata
= (bma
->offset
== 0) ?
4334 XFS_ALLOC_INITIAL_USER_DATA
: XFS_ALLOC_USERDATA
;
4337 bma
->minlen
= (bma
->flags
& XFS_BMAPI_CONTIG
) ? bma
->length
: 1;
4340 * Only want to do the alignment at the eof if it is userdata and
4341 * allocation length is larger than a stripe unit.
4343 if (mp
->m_dalign
&& bma
->length
>= mp
->m_dalign
&&
4344 !(bma
->flags
& XFS_BMAPI_METADATA
) && whichfork
== XFS_DATA_FORK
) {
4345 error
= xfs_bmap_isaeof(bma
, whichfork
);
4350 error
= xfs_bmap_alloc(bma
);
4354 if (bma
->flist
->xbf_low
)
4357 bma
->cur
->bc_private
.b
.firstblock
= *bma
->firstblock
;
4358 if (bma
->blkno
== NULLFSBLOCK
)
4360 if ((ifp
->if_flags
& XFS_IFBROOT
) && !bma
->cur
) {
4361 bma
->cur
= xfs_bmbt_init_cursor(mp
, bma
->tp
, bma
->ip
, whichfork
);
4362 bma
->cur
->bc_private
.b
.firstblock
= *bma
->firstblock
;
4363 bma
->cur
->bc_private
.b
.flist
= bma
->flist
;
4366 * Bump the number of extents we've allocated
4372 bma
->cur
->bc_private
.b
.flags
=
4373 bma
->wasdel
? XFS_BTCUR_BPRV_WASDEL
: 0;
4375 bma
->got
.br_startoff
= bma
->offset
;
4376 bma
->got
.br_startblock
= bma
->blkno
;
4377 bma
->got
.br_blockcount
= bma
->length
;
4378 bma
->got
.br_state
= XFS_EXT_NORM
;
4381 * A wasdelay extent has been initialized, so shouldn't be flagged
4384 if (!bma
->wasdel
&& (bma
->flags
& XFS_BMAPI_PREALLOC
) &&
4385 xfs_sb_version_hasextflgbit(&mp
->m_sb
))
4386 bma
->got
.br_state
= XFS_EXT_UNWRITTEN
;
4389 error
= xfs_bmap_add_extent_delay_real(bma
);
4391 error
= xfs_bmap_add_extent_hole_real(bma
, whichfork
);
4393 bma
->logflags
|= tmp_logflags
;
4398 * Update our extent pointer, given that xfs_bmap_add_extent_delay_real
4399 * or xfs_bmap_add_extent_hole_real might have merged it into one of
4400 * the neighbouring ones.
4402 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
), &bma
->got
);
4404 ASSERT(bma
->got
.br_startoff
<= bma
->offset
);
4405 ASSERT(bma
->got
.br_startoff
+ bma
->got
.br_blockcount
>=
4406 bma
->offset
+ bma
->length
);
4407 ASSERT(bma
->got
.br_state
== XFS_EXT_NORM
||
4408 bma
->got
.br_state
== XFS_EXT_UNWRITTEN
);
4413 xfs_bmapi_convert_unwritten(
4414 struct xfs_bmalloca
*bma
,
4415 struct xfs_bmbt_irec
*mval
,
4419 int whichfork
= (flags
& XFS_BMAPI_ATTRFORK
) ?
4420 XFS_ATTR_FORK
: XFS_DATA_FORK
;
4421 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(bma
->ip
, whichfork
);
4422 int tmp_logflags
= 0;
4425 /* check if we need to do unwritten->real conversion */
4426 if (mval
->br_state
== XFS_EXT_UNWRITTEN
&&
4427 (flags
& XFS_BMAPI_PREALLOC
))
4430 /* check if we need to do real->unwritten conversion */
4431 if (mval
->br_state
== XFS_EXT_NORM
&&
4432 (flags
& (XFS_BMAPI_PREALLOC
| XFS_BMAPI_CONVERT
)) !=
4433 (XFS_BMAPI_PREALLOC
| XFS_BMAPI_CONVERT
))
4437 * Modify (by adding) the state flag, if writing.
4439 ASSERT(mval
->br_blockcount
<= len
);
4440 if ((ifp
->if_flags
& XFS_IFBROOT
) && !bma
->cur
) {
4441 bma
->cur
= xfs_bmbt_init_cursor(bma
->ip
->i_mount
, bma
->tp
,
4442 bma
->ip
, whichfork
);
4443 bma
->cur
->bc_private
.b
.firstblock
= *bma
->firstblock
;
4444 bma
->cur
->bc_private
.b
.flist
= bma
->flist
;
4446 mval
->br_state
= (mval
->br_state
== XFS_EXT_UNWRITTEN
)
4447 ? XFS_EXT_NORM
: XFS_EXT_UNWRITTEN
;
4449 error
= xfs_bmap_add_extent_unwritten_real(bma
->tp
, bma
->ip
, &bma
->idx
,
4450 &bma
->cur
, mval
, bma
->firstblock
, bma
->flist
,
4452 bma
->logflags
|= tmp_logflags
;
4457 * Update our extent pointer, given that
4458 * xfs_bmap_add_extent_unwritten_real might have merged it into one
4459 * of the neighbouring ones.
4461 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
->idx
), &bma
->got
);
4464 * We may have combined previously unwritten space with written space,
4465 * so generate another request.
4467 if (mval
->br_blockcount
< len
)
4473 * Map file blocks to filesystem blocks, and allocate blocks or convert the
4474 * extent state if necessary. Details behaviour is controlled by the flags
4475 * parameter. Only allocates blocks from a single allocation group, to avoid
4478 * The returned value in "firstblock" from the first call in a transaction
4479 * must be remembered and presented to subsequent calls in "firstblock".
4480 * An upper bound for the number of blocks to be allocated is supplied to
4481 * the first call in "total"; if no allocation group has that many free
4482 * blocks then the call will fail (return NULLFSBLOCK in "firstblock").
4486 struct xfs_trans
*tp
, /* transaction pointer */
4487 struct xfs_inode
*ip
, /* incore inode */
4488 xfs_fileoff_t bno
, /* starting file offs. mapped */
4489 xfs_filblks_t len
, /* length to map in file */
4490 int flags
, /* XFS_BMAPI_... */
4491 xfs_fsblock_t
*firstblock
, /* first allocated block
4492 controls a.g. for allocs */
4493 xfs_extlen_t total
, /* total blocks needed */
4494 struct xfs_bmbt_irec
*mval
, /* output: map values */
4495 int *nmap
, /* i/o: mval size/count */
4496 struct xfs_bmap_free
*flist
) /* i/o: list extents to free */
4498 struct xfs_mount
*mp
= ip
->i_mount
;
4499 struct xfs_ifork
*ifp
;
4500 struct xfs_bmalloca bma
= { NULL
}; /* args for xfs_bmap_alloc */
4501 xfs_fileoff_t end
; /* end of mapped file region */
4502 int eof
; /* after the end of extents */
4503 int error
; /* error return */
4504 int n
; /* current extent index */
4505 xfs_fileoff_t obno
; /* old block number (offset) */
4506 int whichfork
; /* data or attr fork */
4507 char inhole
; /* current location is hole in file */
4508 char wasdelay
; /* old extent was delayed */
4511 xfs_fileoff_t orig_bno
; /* original block number value */
4512 int orig_flags
; /* original flags arg value */
4513 xfs_filblks_t orig_len
; /* original value of len arg */
4514 struct xfs_bmbt_irec
*orig_mval
; /* original value of mval */
4515 int orig_nmap
; /* original value of *nmap */
4523 whichfork
= (flags
& XFS_BMAPI_ATTRFORK
) ?
4524 XFS_ATTR_FORK
: XFS_DATA_FORK
;
4527 ASSERT(*nmap
<= XFS_BMAP_MAX_NMAP
);
4528 ASSERT(!(flags
& XFS_BMAPI_IGSTATE
));
4531 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_LOCAL
);
4532 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_EXCL
));
4534 if (unlikely(XFS_TEST_ERROR(
4535 (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
&&
4536 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
),
4537 mp
, XFS_ERRTAG_BMAPIFORMAT
, XFS_RANDOM_BMAPIFORMAT
))) {
4538 XFS_ERROR_REPORT("xfs_bmapi_write", XFS_ERRLEVEL_LOW
, mp
);
4539 return -EFSCORRUPTED
;
4542 if (XFS_FORCED_SHUTDOWN(mp
))
4545 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
4547 XFS_STATS_INC(xs_blk_mapw
);
4549 if (*firstblock
== NULLFSBLOCK
) {
4550 if (XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_BTREE
)
4551 bma
.minleft
= be16_to_cpu(ifp
->if_broot
->bb_level
) + 1;
4558 if (!(ifp
->if_flags
& XFS_IFEXTENTS
)) {
4559 error
= xfs_iread_extents(tp
, ip
, whichfork
);
4564 xfs_bmap_search_extents(ip
, bno
, whichfork
, &eof
, &bma
.idx
, &bma
.got
,
4575 bma
.firstblock
= firstblock
;
4577 while (bno
< end
&& n
< *nmap
) {
4578 inhole
= eof
|| bma
.got
.br_startoff
> bno
;
4579 wasdelay
= !inhole
&& isnullstartblock(bma
.got
.br_startblock
);
4582 * First, deal with the hole before the allocated space
4583 * that we found, if any.
4585 if (inhole
|| wasdelay
) {
4587 bma
.conv
= !!(flags
& XFS_BMAPI_CONVERT
);
4588 bma
.wasdel
= wasdelay
;
4593 * There's a 32/64 bit type mismatch between the
4594 * allocation length request (which can be 64 bits in
4595 * length) and the bma length request, which is
4596 * xfs_extlen_t and therefore 32 bits. Hence we have to
4597 * check for 32-bit overflows and handle them here.
4599 if (len
> (xfs_filblks_t
)MAXEXTLEN
)
4600 bma
.length
= MAXEXTLEN
;
4605 ASSERT(bma
.length
> 0);
4606 error
= xfs_bmapi_allocate(&bma
);
4609 if (bma
.blkno
== NULLFSBLOCK
)
4613 /* Deal with the allocated space we found. */
4614 xfs_bmapi_trim_map(mval
, &bma
.got
, &bno
, len
, obno
,
4617 /* Execute unwritten extent conversion if necessary */
4618 error
= xfs_bmapi_convert_unwritten(&bma
, mval
, len
, flags
);
4619 if (error
== -EAGAIN
)
4624 /* update the extent map to return */
4625 xfs_bmapi_update_map(&mval
, &bno
, &len
, obno
, end
, &n
, flags
);
4628 * If we're done, stop now. Stop when we've allocated
4629 * XFS_BMAP_MAX_NMAP extents no matter what. Otherwise
4630 * the transaction may get too big.
4632 if (bno
>= end
|| n
>= *nmap
|| bma
.nallocs
>= *nmap
)
4635 /* Else go on to the next record. */
4637 if (++bma
.idx
< ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
)) {
4638 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
, bma
.idx
),
4646 * Transform from btree to extents, give it cur.
4648 if (xfs_bmap_wants_extents(ip
, whichfork
)) {
4649 int tmp_logflags
= 0;
4652 error
= xfs_bmap_btree_to_extents(tp
, ip
, bma
.cur
,
4653 &tmp_logflags
, whichfork
);
4654 bma
.logflags
|= tmp_logflags
;
4659 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
||
4660 XFS_IFORK_NEXTENTS(ip
, whichfork
) >
4661 XFS_IFORK_MAXEXT(ip
, whichfork
));
4665 * Log everything. Do this after conversion, there's no point in
4666 * logging the extent records if we've converted to btree format.
4668 if ((bma
.logflags
& xfs_ilog_fext(whichfork
)) &&
4669 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
)
4670 bma
.logflags
&= ~xfs_ilog_fext(whichfork
);
4671 else if ((bma
.logflags
& xfs_ilog_fbroot(whichfork
)) &&
4672 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
)
4673 bma
.logflags
&= ~xfs_ilog_fbroot(whichfork
);
4675 * Log whatever the flags say, even if error. Otherwise we might miss
4676 * detecting a case where the data is changed, there's an error,
4677 * and it's not logged so we don't shutdown when we should.
4680 xfs_trans_log_inode(tp
, ip
, bma
.logflags
);
4684 ASSERT(*firstblock
== NULLFSBLOCK
||
4685 XFS_FSB_TO_AGNO(mp
, *firstblock
) ==
4687 bma
.cur
->bc_private
.b
.firstblock
) ||
4689 XFS_FSB_TO_AGNO(mp
, *firstblock
) <
4691 bma
.cur
->bc_private
.b
.firstblock
)));
4692 *firstblock
= bma
.cur
->bc_private
.b
.firstblock
;
4694 xfs_btree_del_cursor(bma
.cur
,
4695 error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);
4698 xfs_bmap_validate_ret(orig_bno
, orig_len
, orig_flags
, orig_mval
,
4704 * Called by xfs_bmapi to update file extent records and the btree
4705 * after removing space (or undoing a delayed allocation).
4707 STATIC
int /* error */
4708 xfs_bmap_del_extent(
4709 xfs_inode_t
*ip
, /* incore inode pointer */
4710 xfs_trans_t
*tp
, /* current transaction pointer */
4711 xfs_extnum_t
*idx
, /* extent number to update/delete */
4712 xfs_bmap_free_t
*flist
, /* list of extents to be freed */
4713 xfs_btree_cur_t
*cur
, /* if null, not a btree */
4714 xfs_bmbt_irec_t
*del
, /* data to remove from extents */
4715 int *logflagsp
, /* inode logging flags */
4716 int whichfork
) /* data or attr fork */
4718 xfs_filblks_t da_new
; /* new delay-alloc indirect blocks */
4719 xfs_filblks_t da_old
; /* old delay-alloc indirect blocks */
4720 xfs_fsblock_t del_endblock
=0; /* first block past del */
4721 xfs_fileoff_t del_endoff
; /* first offset past del */
4722 int delay
; /* current block is delayed allocated */
4723 int do_fx
; /* free extent at end of routine */
4724 xfs_bmbt_rec_host_t
*ep
; /* current extent entry pointer */
4725 int error
; /* error return value */
4726 int flags
; /* inode logging flags */
4727 xfs_bmbt_irec_t got
; /* current extent entry */
4728 xfs_fileoff_t got_endoff
; /* first offset past got */
4729 int i
; /* temp state */
4730 xfs_ifork_t
*ifp
; /* inode fork pointer */
4731 xfs_mount_t
*mp
; /* mount structure */
4732 xfs_filblks_t nblks
; /* quota/sb block count */
4733 xfs_bmbt_irec_t
new; /* new record to be inserted */
4735 uint qfield
; /* quota field to update */
4736 xfs_filblks_t temp
; /* for indirect length calculations */
4737 xfs_filblks_t temp2
; /* for indirect length calculations */
4740 XFS_STATS_INC(xs_del_exlist
);
4742 if (whichfork
== XFS_ATTR_FORK
)
4743 state
|= BMAP_ATTRFORK
;
4746 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
4747 ASSERT((*idx
>= 0) && (*idx
< ifp
->if_bytes
/
4748 (uint
)sizeof(xfs_bmbt_rec_t
)));
4749 ASSERT(del
->br_blockcount
> 0);
4750 ep
= xfs_iext_get_ext(ifp
, *idx
);
4751 xfs_bmbt_get_all(ep
, &got
);
4752 ASSERT(got
.br_startoff
<= del
->br_startoff
);
4753 del_endoff
= del
->br_startoff
+ del
->br_blockcount
;
4754 got_endoff
= got
.br_startoff
+ got
.br_blockcount
;
4755 ASSERT(got_endoff
>= del_endoff
);
4756 delay
= isnullstartblock(got
.br_startblock
);
4757 ASSERT(isnullstartblock(del
->br_startblock
) == delay
);
4762 * If deleting a real allocation, must free up the disk space.
4765 flags
= XFS_ILOG_CORE
;
4767 * Realtime allocation. Free it and record di_nblocks update.
4769 if (whichfork
== XFS_DATA_FORK
&& XFS_IS_REALTIME_INODE(ip
)) {
4773 ASSERT(do_mod(del
->br_blockcount
,
4774 mp
->m_sb
.sb_rextsize
) == 0);
4775 ASSERT(do_mod(del
->br_startblock
,
4776 mp
->m_sb
.sb_rextsize
) == 0);
4777 bno
= del
->br_startblock
;
4778 len
= del
->br_blockcount
;
4779 do_div(bno
, mp
->m_sb
.sb_rextsize
);
4780 do_div(len
, mp
->m_sb
.sb_rextsize
);
4781 error
= xfs_rtfree_extent(tp
, bno
, (xfs_extlen_t
)len
);
4785 nblks
= len
* mp
->m_sb
.sb_rextsize
;
4786 qfield
= XFS_TRANS_DQ_RTBCOUNT
;
4789 * Ordinary allocation.
4793 nblks
= del
->br_blockcount
;
4794 qfield
= XFS_TRANS_DQ_BCOUNT
;
4797 * Set up del_endblock and cur for later.
4799 del_endblock
= del
->br_startblock
+ del
->br_blockcount
;
4801 if ((error
= xfs_bmbt_lookup_eq(cur
, got
.br_startoff
,
4802 got
.br_startblock
, got
.br_blockcount
,
4805 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
4807 da_old
= da_new
= 0;
4809 da_old
= startblockval(got
.br_startblock
);
4815 * Set flag value to use in switch statement.
4816 * Left-contig is 2, right-contig is 1.
4818 switch (((got
.br_startoff
== del
->br_startoff
) << 1) |
4819 (got_endoff
== del_endoff
)) {
4822 * Matches the whole extent. Delete the entry.
4824 xfs_iext_remove(ip
, *idx
, 1,
4825 whichfork
== XFS_ATTR_FORK
? BMAP_ATTRFORK
: 0);
4830 XFS_IFORK_NEXT_SET(ip
, whichfork
,
4831 XFS_IFORK_NEXTENTS(ip
, whichfork
) - 1);
4832 flags
|= XFS_ILOG_CORE
;
4834 flags
|= xfs_ilog_fext(whichfork
);
4837 if ((error
= xfs_btree_delete(cur
, &i
)))
4839 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
4844 * Deleting the first part of the extent.
4846 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
4847 xfs_bmbt_set_startoff(ep
, del_endoff
);
4848 temp
= got
.br_blockcount
- del
->br_blockcount
;
4849 xfs_bmbt_set_blockcount(ep
, temp
);
4851 temp
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip
, temp
),
4853 xfs_bmbt_set_startblock(ep
, nullstartblock((int)temp
));
4854 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
4858 xfs_bmbt_set_startblock(ep
, del_endblock
);
4859 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
4861 flags
|= xfs_ilog_fext(whichfork
);
4864 if ((error
= xfs_bmbt_update(cur
, del_endoff
, del_endblock
,
4865 got
.br_blockcount
- del
->br_blockcount
,
4872 * Deleting the last part of the extent.
4874 temp
= got
.br_blockcount
- del
->br_blockcount
;
4875 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
4876 xfs_bmbt_set_blockcount(ep
, temp
);
4878 temp
= XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip
, temp
),
4880 xfs_bmbt_set_startblock(ep
, nullstartblock((int)temp
));
4881 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
4885 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
4887 flags
|= xfs_ilog_fext(whichfork
);
4890 if ((error
= xfs_bmbt_update(cur
, got
.br_startoff
,
4892 got
.br_blockcount
- del
->br_blockcount
,
4899 * Deleting the middle of the extent.
4901 temp
= del
->br_startoff
- got
.br_startoff
;
4902 trace_xfs_bmap_pre_update(ip
, *idx
, state
, _THIS_IP_
);
4903 xfs_bmbt_set_blockcount(ep
, temp
);
4904 new.br_startoff
= del_endoff
;
4905 temp2
= got_endoff
- del_endoff
;
4906 new.br_blockcount
= temp2
;
4907 new.br_state
= got
.br_state
;
4909 new.br_startblock
= del_endblock
;
4910 flags
|= XFS_ILOG_CORE
;
4912 if ((error
= xfs_bmbt_update(cur
,
4914 got
.br_startblock
, temp
,
4917 if ((error
= xfs_btree_increment(cur
, 0, &i
)))
4919 cur
->bc_rec
.b
= new;
4920 error
= xfs_btree_insert(cur
, &i
);
4921 if (error
&& error
!= -ENOSPC
)
4924 * If get no-space back from btree insert,
4925 * it tried a split, and we have a zero
4926 * block reservation.
4927 * Fix up our state and return the error.
4929 if (error
== -ENOSPC
) {
4931 * Reset the cursor, don't trust
4932 * it after any insert operation.
4934 if ((error
= xfs_bmbt_lookup_eq(cur
,
4939 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
4941 * Update the btree record back
4942 * to the original value.
4944 if ((error
= xfs_bmbt_update(cur
,
4951 * Reset the extent record back
4952 * to the original value.
4954 xfs_bmbt_set_blockcount(ep
,
4960 XFS_WANT_CORRUPTED_GOTO(i
== 1, done
);
4962 flags
|= xfs_ilog_fext(whichfork
);
4963 XFS_IFORK_NEXT_SET(ip
, whichfork
,
4964 XFS_IFORK_NEXTENTS(ip
, whichfork
) + 1);
4966 ASSERT(whichfork
== XFS_DATA_FORK
);
4967 temp
= xfs_bmap_worst_indlen(ip
, temp
);
4968 xfs_bmbt_set_startblock(ep
, nullstartblock((int)temp
));
4969 temp2
= xfs_bmap_worst_indlen(ip
, temp2
);
4970 new.br_startblock
= nullstartblock((int)temp2
);
4971 da_new
= temp
+ temp2
;
4972 while (da_new
> da_old
) {
4976 xfs_bmbt_set_startblock(ep
,
4977 nullstartblock((int)temp
));
4979 if (da_new
== da_old
)
4985 nullstartblock((int)temp2
);
4989 trace_xfs_bmap_post_update(ip
, *idx
, state
, _THIS_IP_
);
4990 xfs_iext_insert(ip
, *idx
+ 1, 1, &new, state
);
4995 * If we need to, add to list of extents to delete.
4998 xfs_bmap_add_free(del
->br_startblock
, del
->br_blockcount
, flist
,
5001 * Adjust inode # blocks in the file.
5004 ip
->i_d
.di_nblocks
-= nblks
;
5006 * Adjust quota data.
5009 xfs_trans_mod_dquot_byino(tp
, ip
, qfield
, (long)-nblks
);
5012 * Account for change in delayed indirect blocks.
5013 * Nothing to do for disk quota accounting here.
5015 ASSERT(da_old
>= da_new
);
5016 if (da_old
> da_new
) {
5017 xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
,
5018 (int64_t)(da_old
- da_new
), 0);
5026 * Unmap (remove) blocks from a file.
5027 * If nexts is nonzero then the number of extents to remove is limited to
5028 * that value. If not all extents in the block range can be removed then
5033 xfs_trans_t
*tp
, /* transaction pointer */
5034 struct xfs_inode
*ip
, /* incore inode */
5035 xfs_fileoff_t bno
, /* starting offset to unmap */
5036 xfs_filblks_t len
, /* length to unmap in file */
5037 int flags
, /* misc flags */
5038 xfs_extnum_t nexts
, /* number of extents max */
5039 xfs_fsblock_t
*firstblock
, /* first allocated block
5040 controls a.g. for allocs */
5041 xfs_bmap_free_t
*flist
, /* i/o: list extents to free */
5042 int *done
) /* set if not done yet */
5044 xfs_btree_cur_t
*cur
; /* bmap btree cursor */
5045 xfs_bmbt_irec_t del
; /* extent being deleted */
5046 int eof
; /* is deleting at eof */
5047 xfs_bmbt_rec_host_t
*ep
; /* extent record pointer */
5048 int error
; /* error return value */
5049 xfs_extnum_t extno
; /* extent number in list */
5050 xfs_bmbt_irec_t got
; /* current extent record */
5051 xfs_ifork_t
*ifp
; /* inode fork pointer */
5052 int isrt
; /* freeing in rt area */
5053 xfs_extnum_t lastx
; /* last extent index used */
5054 int logflags
; /* transaction logging flags */
5055 xfs_extlen_t mod
; /* rt extent offset */
5056 xfs_mount_t
*mp
; /* mount structure */
5057 xfs_extnum_t nextents
; /* number of file extents */
5058 xfs_bmbt_irec_t prev
; /* previous extent record */
5059 xfs_fileoff_t start
; /* first file offset deleted */
5060 int tmp_logflags
; /* partial logging flags */
5061 int wasdel
; /* was a delayed alloc extent */
5062 int whichfork
; /* data or attribute fork */
5065 trace_xfs_bunmap(ip
, bno
, len
, flags
, _RET_IP_
);
5067 whichfork
= (flags
& XFS_BMAPI_ATTRFORK
) ?
5068 XFS_ATTR_FORK
: XFS_DATA_FORK
;
5069 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
5071 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
&&
5072 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
)) {
5073 XFS_ERROR_REPORT("xfs_bunmapi", XFS_ERRLEVEL_LOW
,
5075 return -EFSCORRUPTED
;
5078 if (XFS_FORCED_SHUTDOWN(mp
))
5081 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_EXCL
));
5085 if (!(ifp
->if_flags
& XFS_IFEXTENTS
) &&
5086 (error
= xfs_iread_extents(tp
, ip
, whichfork
)))
5088 nextents
= ifp
->if_bytes
/ (uint
)sizeof(xfs_bmbt_rec_t
);
5089 if (nextents
== 0) {
5093 XFS_STATS_INC(xs_blk_unmap
);
5094 isrt
= (whichfork
== XFS_DATA_FORK
) && XFS_IS_REALTIME_INODE(ip
);
5096 bno
= start
+ len
- 1;
5097 ep
= xfs_bmap_search_extents(ip
, bno
, whichfork
, &eof
, &lastx
, &got
,
5101 * Check to see if the given block number is past the end of the
5102 * file, back up to the last block if so...
5105 ep
= xfs_iext_get_ext(ifp
, --lastx
);
5106 xfs_bmbt_get_all(ep
, &got
);
5107 bno
= got
.br_startoff
+ got
.br_blockcount
- 1;
5110 if (ifp
->if_flags
& XFS_IFBROOT
) {
5111 ASSERT(XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_BTREE
);
5112 cur
= xfs_bmbt_init_cursor(mp
, tp
, ip
, whichfork
);
5113 cur
->bc_private
.b
.firstblock
= *firstblock
;
5114 cur
->bc_private
.b
.flist
= flist
;
5115 cur
->bc_private
.b
.flags
= 0;
5121 * Synchronize by locking the bitmap inode.
5123 xfs_ilock(mp
->m_rbmip
, XFS_ILOCK_EXCL
);
5124 xfs_trans_ijoin(tp
, mp
->m_rbmip
, XFS_ILOCK_EXCL
);
5128 while (bno
!= (xfs_fileoff_t
)-1 && bno
>= start
&& lastx
>= 0 &&
5129 (nexts
== 0 || extno
< nexts
)) {
5131 * Is the found extent after a hole in which bno lives?
5132 * Just back up to the previous extent, if so.
5134 if (got
.br_startoff
> bno
) {
5137 ep
= xfs_iext_get_ext(ifp
, lastx
);
5138 xfs_bmbt_get_all(ep
, &got
);
5141 * Is the last block of this extent before the range
5142 * we're supposed to delete? If so, we're done.
5144 bno
= XFS_FILEOFF_MIN(bno
,
5145 got
.br_startoff
+ got
.br_blockcount
- 1);
5149 * Then deal with the (possibly delayed) allocated space
5154 wasdel
= isnullstartblock(del
.br_startblock
);
5155 if (got
.br_startoff
< start
) {
5156 del
.br_startoff
= start
;
5157 del
.br_blockcount
-= start
- got
.br_startoff
;
5159 del
.br_startblock
+= start
- got
.br_startoff
;
5161 if (del
.br_startoff
+ del
.br_blockcount
> bno
+ 1)
5162 del
.br_blockcount
= bno
+ 1 - del
.br_startoff
;
5163 sum
= del
.br_startblock
+ del
.br_blockcount
;
5165 (mod
= do_mod(sum
, mp
->m_sb
.sb_rextsize
))) {
5167 * Realtime extent not lined up at the end.
5168 * The extent could have been split into written
5169 * and unwritten pieces, or we could just be
5170 * unmapping part of it. But we can't really
5171 * get rid of part of a realtime extent.
5173 if (del
.br_state
== XFS_EXT_UNWRITTEN
||
5174 !xfs_sb_version_hasextflgbit(&mp
->m_sb
)) {
5176 * This piece is unwritten, or we're not
5177 * using unwritten extents. Skip over it.
5180 bno
-= mod
> del
.br_blockcount
?
5181 del
.br_blockcount
: mod
;
5182 if (bno
< got
.br_startoff
) {
5184 xfs_bmbt_get_all(xfs_iext_get_ext(
5190 * It's written, turn it unwritten.
5191 * This is better than zeroing it.
5193 ASSERT(del
.br_state
== XFS_EXT_NORM
);
5194 ASSERT(xfs_trans_get_block_res(tp
) > 0);
5196 * If this spans a realtime extent boundary,
5197 * chop it back to the start of the one we end at.
5199 if (del
.br_blockcount
> mod
) {
5200 del
.br_startoff
+= del
.br_blockcount
- mod
;
5201 del
.br_startblock
+= del
.br_blockcount
- mod
;
5202 del
.br_blockcount
= mod
;
5204 del
.br_state
= XFS_EXT_UNWRITTEN
;
5205 error
= xfs_bmap_add_extent_unwritten_real(tp
, ip
,
5206 &lastx
, &cur
, &del
, firstblock
, flist
,
5212 if (isrt
&& (mod
= do_mod(del
.br_startblock
, mp
->m_sb
.sb_rextsize
))) {
5214 * Realtime extent is lined up at the end but not
5215 * at the front. We'll get rid of full extents if
5218 mod
= mp
->m_sb
.sb_rextsize
- mod
;
5219 if (del
.br_blockcount
> mod
) {
5220 del
.br_blockcount
-= mod
;
5221 del
.br_startoff
+= mod
;
5222 del
.br_startblock
+= mod
;
5223 } else if ((del
.br_startoff
== start
&&
5224 (del
.br_state
== XFS_EXT_UNWRITTEN
||
5225 xfs_trans_get_block_res(tp
) == 0)) ||
5226 !xfs_sb_version_hasextflgbit(&mp
->m_sb
)) {
5228 * Can't make it unwritten. There isn't
5229 * a full extent here so just skip it.
5231 ASSERT(bno
>= del
.br_blockcount
);
5232 bno
-= del
.br_blockcount
;
5233 if (got
.br_startoff
> bno
) {
5235 ep
= xfs_iext_get_ext(ifp
,
5237 xfs_bmbt_get_all(ep
, &got
);
5241 } else if (del
.br_state
== XFS_EXT_UNWRITTEN
) {
5243 * This one is already unwritten.
5244 * It must have a written left neighbor.
5245 * Unwrite the killed part of that one and
5249 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
,
5251 ASSERT(prev
.br_state
== XFS_EXT_NORM
);
5252 ASSERT(!isnullstartblock(prev
.br_startblock
));
5253 ASSERT(del
.br_startblock
==
5254 prev
.br_startblock
+ prev
.br_blockcount
);
5255 if (prev
.br_startoff
< start
) {
5256 mod
= start
- prev
.br_startoff
;
5257 prev
.br_blockcount
-= mod
;
5258 prev
.br_startblock
+= mod
;
5259 prev
.br_startoff
= start
;
5261 prev
.br_state
= XFS_EXT_UNWRITTEN
;
5263 error
= xfs_bmap_add_extent_unwritten_real(tp
,
5264 ip
, &lastx
, &cur
, &prev
,
5265 firstblock
, flist
, &logflags
);
5270 ASSERT(del
.br_state
== XFS_EXT_NORM
);
5271 del
.br_state
= XFS_EXT_UNWRITTEN
;
5272 error
= xfs_bmap_add_extent_unwritten_real(tp
,
5273 ip
, &lastx
, &cur
, &del
,
5274 firstblock
, flist
, &logflags
);
5281 ASSERT(startblockval(del
.br_startblock
) > 0);
5282 /* Update realtime/data freespace, unreserve quota */
5284 xfs_filblks_t rtexts
;
5286 rtexts
= XFS_FSB_TO_B(mp
, del
.br_blockcount
);
5287 do_div(rtexts
, mp
->m_sb
.sb_rextsize
);
5288 xfs_mod_incore_sb(mp
, XFS_SBS_FREXTENTS
,
5289 (int64_t)rtexts
, 0);
5290 (void)xfs_trans_reserve_quota_nblks(NULL
,
5291 ip
, -((long)del
.br_blockcount
), 0,
5292 XFS_QMOPT_RES_RTBLKS
);
5294 xfs_icsb_modify_counters(mp
, XFS_SBS_FDBLOCKS
,
5295 (int64_t)del
.br_blockcount
, 0);
5296 (void)xfs_trans_reserve_quota_nblks(NULL
,
5297 ip
, -((long)del
.br_blockcount
), 0,
5298 XFS_QMOPT_RES_REGBLKS
);
5300 ip
->i_delayed_blks
-= del
.br_blockcount
;
5302 cur
->bc_private
.b
.flags
|=
5303 XFS_BTCUR_BPRV_WASDEL
;
5305 cur
->bc_private
.b
.flags
&= ~XFS_BTCUR_BPRV_WASDEL
;
5307 * If it's the case where the directory code is running
5308 * with no block reservation, and the deleted block is in
5309 * the middle of its extent, and the resulting insert
5310 * of an extent would cause transformation to btree format,
5311 * then reject it. The calling code will then swap
5312 * blocks around instead.
5313 * We have to do this now, rather than waiting for the
5314 * conversion to btree format, since the transaction
5317 if (!wasdel
&& xfs_trans_get_block_res(tp
) == 0 &&
5318 XFS_IFORK_FORMAT(ip
, whichfork
) == XFS_DINODE_FMT_EXTENTS
&&
5319 XFS_IFORK_NEXTENTS(ip
, whichfork
) >= /* Note the >= */
5320 XFS_IFORK_MAXEXT(ip
, whichfork
) &&
5321 del
.br_startoff
> got
.br_startoff
&&
5322 del
.br_startoff
+ del
.br_blockcount
<
5323 got
.br_startoff
+ got
.br_blockcount
) {
5327 error
= xfs_bmap_del_extent(ip
, tp
, &lastx
, flist
, cur
, &del
,
5328 &tmp_logflags
, whichfork
);
5329 logflags
|= tmp_logflags
;
5332 bno
= del
.br_startoff
- 1;
5335 * If not done go on to the next (previous) record.
5337 if (bno
!= (xfs_fileoff_t
)-1 && bno
>= start
) {
5339 ep
= xfs_iext_get_ext(ifp
, lastx
);
5340 if (xfs_bmbt_get_startoff(ep
) > bno
) {
5342 ep
= xfs_iext_get_ext(ifp
,
5345 xfs_bmbt_get_all(ep
, &got
);
5350 *done
= bno
== (xfs_fileoff_t
)-1 || bno
< start
|| lastx
< 0;
5353 * Convert to a btree if necessary.
5355 if (xfs_bmap_needs_btree(ip
, whichfork
)) {
5356 ASSERT(cur
== NULL
);
5357 error
= xfs_bmap_extents_to_btree(tp
, ip
, firstblock
, flist
,
5358 &cur
, 0, &tmp_logflags
, whichfork
);
5359 logflags
|= tmp_logflags
;
5364 * transform from btree to extents, give it cur
5366 else if (xfs_bmap_wants_extents(ip
, whichfork
)) {
5367 ASSERT(cur
!= NULL
);
5368 error
= xfs_bmap_btree_to_extents(tp
, ip
, cur
, &tmp_logflags
,
5370 logflags
|= tmp_logflags
;
5375 * transform from extents to local?
5380 * Log everything. Do this after conversion, there's no point in
5381 * logging the extent records if we've converted to btree format.
5383 if ((logflags
& xfs_ilog_fext(whichfork
)) &&
5384 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
)
5385 logflags
&= ~xfs_ilog_fext(whichfork
);
5386 else if ((logflags
& xfs_ilog_fbroot(whichfork
)) &&
5387 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
)
5388 logflags
&= ~xfs_ilog_fbroot(whichfork
);
5390 * Log inode even in the error case, if the transaction
5391 * is dirty we'll need to shut down the filesystem.
5394 xfs_trans_log_inode(tp
, ip
, logflags
);
5397 *firstblock
= cur
->bc_private
.b
.firstblock
;
5398 cur
->bc_private
.b
.allocated
= 0;
5400 xfs_btree_del_cursor(cur
,
5401 error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);
5407 * Shift extent records to the left to cover a hole.
5409 * The maximum number of extents to be shifted in a single operation
5410 * is @num_exts, and @current_ext keeps track of the current extent
5411 * index we have shifted. @offset_shift_fsb is the length by which each
5412 * extent is shifted. If there is no hole to shift the extents
5413 * into, this will be considered invalid operation and we abort immediately.
5416 xfs_bmap_shift_extents(
5417 struct xfs_trans
*tp
,
5418 struct xfs_inode
*ip
,
5420 xfs_fileoff_t start_fsb
,
5421 xfs_fileoff_t offset_shift_fsb
,
5422 xfs_extnum_t
*current_ext
,
5423 xfs_fsblock_t
*firstblock
,
5424 struct xfs_bmap_free
*flist
,
5427 struct xfs_btree_cur
*cur
;
5428 struct xfs_bmbt_rec_host
*gotp
;
5429 struct xfs_bmbt_irec got
;
5430 struct xfs_bmbt_irec left
;
5431 struct xfs_mount
*mp
= ip
->i_mount
;
5432 struct xfs_ifork
*ifp
;
5433 xfs_extnum_t nexts
= 0;
5434 xfs_fileoff_t startoff
;
5437 int whichfork
= XFS_DATA_FORK
;
5439 xfs_filblks_t blockcount
= 0;
5442 if (unlikely(XFS_TEST_ERROR(
5443 (XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_EXTENTS
&&
5444 XFS_IFORK_FORMAT(ip
, whichfork
) != XFS_DINODE_FMT_BTREE
),
5445 mp
, XFS_ERRTAG_BMAPIFORMAT
, XFS_RANDOM_BMAPIFORMAT
))) {
5446 XFS_ERROR_REPORT("xfs_bmap_shift_extents",
5447 XFS_ERRLEVEL_LOW
, mp
);
5448 return -EFSCORRUPTED
;
5451 if (XFS_FORCED_SHUTDOWN(mp
))
5454 ASSERT(current_ext
!= NULL
);
5456 ifp
= XFS_IFORK_PTR(ip
, whichfork
);
5457 if (!(ifp
->if_flags
& XFS_IFEXTENTS
)) {
5458 /* Read in all the extents */
5459 error
= xfs_iread_extents(tp
, ip
, whichfork
);
5465 * If *current_ext is 0, we would need to lookup the extent
5466 * from where we would start shifting and store it in gotp.
5468 if (!*current_ext
) {
5469 gotp
= xfs_iext_bno_to_ext(ifp
, start_fsb
, current_ext
);
5471 * gotp can be null in 2 cases: 1) if there are no extents
5472 * or 2) start_fsb lies in a hole beyond which there are
5473 * no extents. Either way, we are done.
5481 /* We are going to change core inode */
5482 logflags
= XFS_ILOG_CORE
;
5483 if (ifp
->if_flags
& XFS_IFBROOT
) {
5484 cur
= xfs_bmbt_init_cursor(mp
, tp
, ip
, whichfork
);
5485 cur
->bc_private
.b
.firstblock
= *firstblock
;
5486 cur
->bc_private
.b
.flist
= flist
;
5487 cur
->bc_private
.b
.flags
= 0;
5490 logflags
|= XFS_ILOG_DEXT
;
5494 * There may be delalloc extents in the data fork before the range we
5495 * are collapsing out, so we cannot
5496 * use the count of real extents here. Instead we have to calculate it
5497 * from the incore fork.
5499 total_extents
= ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
);
5500 while (nexts
++ < num_exts
&& *current_ext
< total_extents
) {
5502 gotp
= xfs_iext_get_ext(ifp
, *current_ext
);
5503 xfs_bmbt_get_all(gotp
, &got
);
5504 startoff
= got
.br_startoff
- offset_shift_fsb
;
5507 * Before shifting extent into hole, make sure that the hole
5508 * is large enough to accomodate the shift.
5511 xfs_bmbt_get_all(xfs_iext_get_ext(ifp
,
5512 *current_ext
- 1), &left
);
5514 if (startoff
< left
.br_startoff
+ left
.br_blockcount
)
5516 } else if (offset_shift_fsb
> got
.br_startoff
) {
5518 * When first extent is shifted, offset_shift_fsb
5519 * should be less than the stating offset of
5529 error
= xfs_bmbt_lookup_eq(cur
, got
.br_startoff
,
5535 XFS_WANT_CORRUPTED_GOTO(i
== 1, del_cursor
);
5538 /* Check if we can merge 2 adjacent extents */
5540 left
.br_startoff
+ left
.br_blockcount
== startoff
&&
5541 left
.br_startblock
+ left
.br_blockcount
==
5542 got
.br_startblock
&&
5543 left
.br_state
== got
.br_state
&&
5544 left
.br_blockcount
+ got
.br_blockcount
<= MAXEXTLEN
) {
5545 blockcount
= left
.br_blockcount
+
5547 xfs_iext_remove(ip
, *current_ext
, 1, 0);
5549 error
= xfs_btree_delete(cur
, &i
);
5552 XFS_WANT_CORRUPTED_GOTO(i
== 1, del_cursor
);
5554 XFS_IFORK_NEXT_SET(ip
, whichfork
,
5555 XFS_IFORK_NEXTENTS(ip
, whichfork
) - 1);
5556 gotp
= xfs_iext_get_ext(ifp
, --*current_ext
);
5557 xfs_bmbt_get_all(gotp
, &got
);
5559 /* Make cursor point to the extent we will update */
5561 error
= xfs_bmbt_lookup_eq(cur
, got
.br_startoff
,
5567 XFS_WANT_CORRUPTED_GOTO(i
== 1, del_cursor
);
5570 xfs_bmbt_set_blockcount(gotp
, blockcount
);
5571 got
.br_blockcount
= blockcount
;
5573 /* We have to update the startoff */
5574 xfs_bmbt_set_startoff(gotp
, startoff
);
5575 got
.br_startoff
= startoff
;
5579 error
= xfs_bmbt_update(cur
, got
.br_startoff
,
5588 total_extents
= ifp
->if_bytes
/ sizeof(xfs_bmbt_rec_t
);
5591 /* Check if we are done */
5592 if (*current_ext
== total_extents
)
5597 xfs_btree_del_cursor(cur
,
5598 error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);
5600 xfs_trans_log_inode(tp
, ip
, logflags
);