2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
25 #include "xfs_mount.h"
26 #include "xfs_defer.h"
27 #include "xfs_inode.h"
28 #include "xfs_trans.h"
29 #include "xfs_inode_item.h"
30 #include "xfs_alloc.h"
31 #include "xfs_btree.h"
32 #include "xfs_bmap_btree.h"
34 #include "xfs_error.h"
35 #include "xfs_quota.h"
36 #include "xfs_trace.h"
37 #include "xfs_cksum.h"
41 * Determine the extent state.
50 ASSERT(blks
!= 0); /* saved for DMIG */
51 return XFS_EXT_UNWRITTEN
;
57 * Convert on-disk form of btree root to in-memory form.
62 xfs_bmdr_block_t
*dblock
,
64 struct xfs_btree_block
*rblock
,
67 struct xfs_mount
*mp
= ip
->i_mount
;
74 xfs_btree_init_block_int(mp
, rblock
, XFS_BUF_DADDR_NULL
,
75 XFS_BTNUM_BMAP
, 0, 0, ip
->i_ino
,
77 rblock
->bb_level
= dblock
->bb_level
;
78 ASSERT(be16_to_cpu(rblock
->bb_level
) > 0);
79 rblock
->bb_numrecs
= dblock
->bb_numrecs
;
80 dmxr
= xfs_bmdr_maxrecs(dblocklen
, 0);
81 fkp
= XFS_BMDR_KEY_ADDR(dblock
, 1);
82 tkp
= XFS_BMBT_KEY_ADDR(mp
, rblock
, 1);
83 fpp
= XFS_BMDR_PTR_ADDR(dblock
, 1, dmxr
);
84 tpp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, rblock
, 1, rblocklen
);
85 dmxr
= be16_to_cpu(dblock
->bb_numrecs
);
86 memcpy(tkp
, fkp
, sizeof(*fkp
) * dmxr
);
87 memcpy(tpp
, fpp
, sizeof(*fpp
) * dmxr
);
91 * Convert a compressed bmap extent record to an uncompressed form.
92 * This code must be in sync with the routines xfs_bmbt_get_startoff,
93 * xfs_bmbt_get_startblock, xfs_bmbt_get_blockcount and xfs_bmbt_get_state.
104 ext_flag
= (int)(l0
>> (64 - BMBT_EXNTFLAG_BITLEN
));
105 s
->br_startoff
= ((xfs_fileoff_t
)l0
&
106 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN
)) >> 9;
107 s
->br_startblock
= (((xfs_fsblock_t
)l0
& xfs_mask64lo(9)) << 43) |
108 (((xfs_fsblock_t
)l1
) >> 21);
109 s
->br_blockcount
= (xfs_filblks_t
)(l1
& xfs_mask64lo(21));
110 /* This is xfs_extent_state() in-line */
112 ASSERT(s
->br_blockcount
!= 0); /* saved for DMIG */
113 st
= XFS_EXT_UNWRITTEN
;
121 xfs_bmbt_rec_host_t
*r
,
124 __xfs_bmbt_get_all(r
->l0
, r
->l1
, s
);
128 * Extract the blockcount field from an in memory bmap extent record.
131 xfs_bmbt_get_blockcount(
132 xfs_bmbt_rec_host_t
*r
)
134 return (xfs_filblks_t
)(r
->l1
& xfs_mask64lo(21));
138 * Extract the startblock field from an in memory bmap extent record.
141 xfs_bmbt_get_startblock(
142 xfs_bmbt_rec_host_t
*r
)
144 return (((xfs_fsblock_t
)r
->l0
& xfs_mask64lo(9)) << 43) |
145 (((xfs_fsblock_t
)r
->l1
) >> 21);
149 * Extract the startoff field from an in memory bmap extent record.
152 xfs_bmbt_get_startoff(
153 xfs_bmbt_rec_host_t
*r
)
155 return ((xfs_fileoff_t
)r
->l0
&
156 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN
)) >> 9;
161 xfs_bmbt_rec_host_t
*r
)
165 ext_flag
= (int)((r
->l0
) >> (64 - BMBT_EXNTFLAG_BITLEN
));
166 return xfs_extent_state(xfs_bmbt_get_blockcount(r
),
171 * Extract the blockcount field from an on disk bmap extent record.
174 xfs_bmbt_disk_get_blockcount(
177 return (xfs_filblks_t
)(be64_to_cpu(r
->l1
) & xfs_mask64lo(21));
181 * Extract the startoff field from a disk format bmap extent record.
184 xfs_bmbt_disk_get_startoff(
187 return ((xfs_fileoff_t
)be64_to_cpu(r
->l0
) &
188 xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN
)) >> 9;
193 * Set all the fields in a bmap extent record from the arguments.
197 xfs_bmbt_rec_host_t
*r
,
198 xfs_fileoff_t startoff
,
199 xfs_fsblock_t startblock
,
200 xfs_filblks_t blockcount
,
203 int extent_flag
= (state
== XFS_EXT_NORM
) ? 0 : 1;
205 ASSERT(state
== XFS_EXT_NORM
|| state
== XFS_EXT_UNWRITTEN
);
206 ASSERT((startoff
& xfs_mask64hi(64-BMBT_STARTOFF_BITLEN
)) == 0);
207 ASSERT((blockcount
& xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN
)) == 0);
209 ASSERT((startblock
& xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN
)) == 0);
211 r
->l0
= ((xfs_bmbt_rec_base_t
)extent_flag
<< 63) |
212 ((xfs_bmbt_rec_base_t
)startoff
<< 9) |
213 ((xfs_bmbt_rec_base_t
)startblock
>> 43);
214 r
->l1
= ((xfs_bmbt_rec_base_t
)startblock
<< 21) |
215 ((xfs_bmbt_rec_base_t
)blockcount
&
216 (xfs_bmbt_rec_base_t
)xfs_mask64lo(21));
220 * Set all the fields in a bmap extent record from the uncompressed form.
224 xfs_bmbt_rec_host_t
*r
,
227 xfs_bmbt_set_allf(r
, s
->br_startoff
, s
->br_startblock
,
228 s
->br_blockcount
, s
->br_state
);
233 * Set all the fields in a disk format bmap extent record from the arguments.
236 xfs_bmbt_disk_set_allf(
238 xfs_fileoff_t startoff
,
239 xfs_fsblock_t startblock
,
240 xfs_filblks_t blockcount
,
243 int extent_flag
= (state
== XFS_EXT_NORM
) ? 0 : 1;
245 ASSERT(state
== XFS_EXT_NORM
|| state
== XFS_EXT_UNWRITTEN
);
246 ASSERT((startoff
& xfs_mask64hi(64-BMBT_STARTOFF_BITLEN
)) == 0);
247 ASSERT((blockcount
& xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN
)) == 0);
248 ASSERT((startblock
& xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN
)) == 0);
251 ((xfs_bmbt_rec_base_t
)extent_flag
<< 63) |
252 ((xfs_bmbt_rec_base_t
)startoff
<< 9) |
253 ((xfs_bmbt_rec_base_t
)startblock
>> 43));
255 ((xfs_bmbt_rec_base_t
)startblock
<< 21) |
256 ((xfs_bmbt_rec_base_t
)blockcount
&
257 (xfs_bmbt_rec_base_t
)xfs_mask64lo(21)));
261 * Set all the fields in a bmap extent record from the uncompressed form.
264 xfs_bmbt_disk_set_all(
268 xfs_bmbt_disk_set_allf(r
, s
->br_startoff
, s
->br_startblock
,
269 s
->br_blockcount
, s
->br_state
);
273 * Set the blockcount field in a bmap extent record.
276 xfs_bmbt_set_blockcount(
277 xfs_bmbt_rec_host_t
*r
,
280 ASSERT((v
& xfs_mask64hi(43)) == 0);
281 r
->l1
= (r
->l1
& (xfs_bmbt_rec_base_t
)xfs_mask64hi(43)) |
282 (xfs_bmbt_rec_base_t
)(v
& xfs_mask64lo(21));
286 * Set the startblock field in a bmap extent record.
289 xfs_bmbt_set_startblock(
290 xfs_bmbt_rec_host_t
*r
,
293 ASSERT((v
& xfs_mask64hi(12)) == 0);
294 r
->l0
= (r
->l0
& (xfs_bmbt_rec_base_t
)xfs_mask64hi(55)) |
295 (xfs_bmbt_rec_base_t
)(v
>> 43);
296 r
->l1
= (r
->l1
& (xfs_bmbt_rec_base_t
)xfs_mask64lo(21)) |
297 (xfs_bmbt_rec_base_t
)(v
<< 21);
301 * Set the startoff field in a bmap extent record.
304 xfs_bmbt_set_startoff(
305 xfs_bmbt_rec_host_t
*r
,
308 ASSERT((v
& xfs_mask64hi(9)) == 0);
309 r
->l0
= (r
->l0
& (xfs_bmbt_rec_base_t
) xfs_mask64hi(1)) |
310 ((xfs_bmbt_rec_base_t
)v
<< 9) |
311 (r
->l0
& (xfs_bmbt_rec_base_t
)xfs_mask64lo(9));
315 * Set the extent state field in a bmap extent record.
319 xfs_bmbt_rec_host_t
*r
,
322 ASSERT(v
== XFS_EXT_NORM
|| v
== XFS_EXT_UNWRITTEN
);
323 if (v
== XFS_EXT_NORM
)
324 r
->l0
&= xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN
);
326 r
->l0
|= xfs_mask64hi(BMBT_EXNTFLAG_BITLEN
);
330 * Convert in-memory form of btree root to on-disk form.
334 struct xfs_mount
*mp
,
335 struct xfs_btree_block
*rblock
,
337 xfs_bmdr_block_t
*dblock
,
346 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
347 ASSERT(rblock
->bb_magic
== cpu_to_be32(XFS_BMAP_CRC_MAGIC
));
348 ASSERT(uuid_equal(&rblock
->bb_u
.l
.bb_uuid
,
349 &mp
->m_sb
.sb_meta_uuid
));
350 ASSERT(rblock
->bb_u
.l
.bb_blkno
==
351 cpu_to_be64(XFS_BUF_DADDR_NULL
));
353 ASSERT(rblock
->bb_magic
== cpu_to_be32(XFS_BMAP_MAGIC
));
354 ASSERT(rblock
->bb_u
.l
.bb_leftsib
== cpu_to_be64(NULLFSBLOCK
));
355 ASSERT(rblock
->bb_u
.l
.bb_rightsib
== cpu_to_be64(NULLFSBLOCK
));
356 ASSERT(rblock
->bb_level
!= 0);
357 dblock
->bb_level
= rblock
->bb_level
;
358 dblock
->bb_numrecs
= rblock
->bb_numrecs
;
359 dmxr
= xfs_bmdr_maxrecs(dblocklen
, 0);
360 fkp
= XFS_BMBT_KEY_ADDR(mp
, rblock
, 1);
361 tkp
= XFS_BMDR_KEY_ADDR(dblock
, 1);
362 fpp
= XFS_BMAP_BROOT_PTR_ADDR(mp
, rblock
, 1, rblocklen
);
363 tpp
= XFS_BMDR_PTR_ADDR(dblock
, 1, dmxr
);
364 dmxr
= be16_to_cpu(dblock
->bb_numrecs
);
365 memcpy(tkp
, fkp
, sizeof(*fkp
) * dmxr
);
366 memcpy(tpp
, fpp
, sizeof(*fpp
) * dmxr
);
370 * Check extent records, which have just been read, for
371 * any bit in the extent flag field. ASSERT on debug
372 * kernels, as this condition should not occur.
373 * Return an error condition (1) if any flags found,
374 * otherwise return 0.
378 xfs_check_nostate_extents(
383 for (; num
> 0; num
--, idx
++) {
384 xfs_bmbt_rec_host_t
*ep
= xfs_iext_get_ext(ifp
, idx
);
386 (64 - BMBT_EXNTFLAG_BITLEN
)) != 0) {
395 STATIC
struct xfs_btree_cur
*
397 struct xfs_btree_cur
*cur
)
399 struct xfs_btree_cur
*new;
401 new = xfs_bmbt_init_cursor(cur
->bc_mp
, cur
->bc_tp
,
402 cur
->bc_private
.b
.ip
, cur
->bc_private
.b
.whichfork
);
405 * Copy the firstblock, dfops, and flags values,
406 * since init cursor doesn't get them.
408 new->bc_private
.b
.firstblock
= cur
->bc_private
.b
.firstblock
;
409 new->bc_private
.b
.dfops
= cur
->bc_private
.b
.dfops
;
410 new->bc_private
.b
.flags
= cur
->bc_private
.b
.flags
;
416 xfs_bmbt_update_cursor(
417 struct xfs_btree_cur
*src
,
418 struct xfs_btree_cur
*dst
)
420 ASSERT((dst
->bc_private
.b
.firstblock
!= NULLFSBLOCK
) ||
421 (dst
->bc_private
.b
.ip
->i_d
.di_flags
& XFS_DIFLAG_REALTIME
));
422 ASSERT(dst
->bc_private
.b
.dfops
== src
->bc_private
.b
.dfops
);
424 dst
->bc_private
.b
.allocated
+= src
->bc_private
.b
.allocated
;
425 dst
->bc_private
.b
.firstblock
= src
->bc_private
.b
.firstblock
;
427 src
->bc_private
.b
.allocated
= 0;
431 xfs_bmbt_alloc_block(
432 struct xfs_btree_cur
*cur
,
433 union xfs_btree_ptr
*start
,
434 union xfs_btree_ptr
*new,
437 xfs_alloc_arg_t args
; /* block allocation args */
438 int error
; /* error return value */
440 memset(&args
, 0, sizeof(args
));
441 args
.tp
= cur
->bc_tp
;
442 args
.mp
= cur
->bc_mp
;
443 args
.fsbno
= cur
->bc_private
.b
.firstblock
;
444 args
.firstblock
= args
.fsbno
;
445 xfs_rmap_ino_bmbt_owner(&args
.oinfo
, cur
->bc_private
.b
.ip
->i_ino
,
446 cur
->bc_private
.b
.whichfork
);
448 if (args
.fsbno
== NULLFSBLOCK
) {
449 args
.fsbno
= be64_to_cpu(start
->l
);
451 args
.type
= XFS_ALLOCTYPE_START_BNO
;
453 * Make sure there is sufficient room left in the AG to
454 * complete a full tree split for an extent insert. If
455 * we are converting the middle part of an extent then
456 * we may need space for two tree splits.
458 * We are relying on the caller to make the correct block
459 * reservation for this operation to succeed. If the
460 * reservation amount is insufficient then we may fail a
461 * block allocation here and corrupt the filesystem.
463 args
.minleft
= args
.tp
->t_blk_res
;
464 } else if (cur
->bc_private
.b
.dfops
->dop_low
) {
465 args
.type
= XFS_ALLOCTYPE_START_BNO
;
467 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
470 args
.minlen
= args
.maxlen
= args
.prod
= 1;
471 args
.wasdel
= cur
->bc_private
.b
.flags
& XFS_BTCUR_BPRV_WASDEL
;
472 if (!args
.wasdel
&& args
.tp
->t_blk_res
== 0) {
476 error
= xfs_alloc_vextent(&args
);
481 * During a CoW operation, the allocation and bmbt updates occur in
482 * different transactions. The mapping code tries to put new bmbt
483 * blocks near extents being mapped, but the only way to guarantee this
484 * is if the alloc and the mapping happen in a single transaction that
485 * has a block reservation. That isn't the case here, so if we run out
486 * of space we'll try again with another AG.
488 if (xfs_sb_version_hasreflink(&cur
->bc_mp
->m_sb
) &&
489 args
.fsbno
== NULLFSBLOCK
&&
490 args
.type
== XFS_ALLOCTYPE_NEAR_BNO
) {
491 cur
->bc_private
.b
.dfops
->dop_low
= true;
492 args
.fsbno
= cur
->bc_private
.b
.firstblock
;
496 if (args
.fsbno
== NULLFSBLOCK
&& args
.minleft
) {
498 * Could not find an AG with enough free space to satisfy
499 * a full btree split. Try again and if
500 * successful activate the lowspace algorithm.
503 args
.type
= XFS_ALLOCTYPE_FIRST_AG
;
504 error
= xfs_alloc_vextent(&args
);
507 cur
->bc_private
.b
.dfops
->dop_low
= true;
509 if (args
.fsbno
== NULLFSBLOCK
) {
510 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
514 ASSERT(args
.len
== 1);
515 cur
->bc_private
.b
.firstblock
= args
.fsbno
;
516 cur
->bc_private
.b
.allocated
++;
517 cur
->bc_private
.b
.ip
->i_d
.di_nblocks
++;
518 xfs_trans_log_inode(args
.tp
, cur
->bc_private
.b
.ip
, XFS_ILOG_CORE
);
519 xfs_trans_mod_dquot_byino(args
.tp
, cur
->bc_private
.b
.ip
,
520 XFS_TRANS_DQ_BCOUNT
, 1L);
522 new->l
= cpu_to_be64(args
.fsbno
);
524 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
529 XFS_BTREE_TRACE_CURSOR(cur
, XBT_ERROR
);
535 struct xfs_btree_cur
*cur
,
538 struct xfs_mount
*mp
= cur
->bc_mp
;
539 struct xfs_inode
*ip
= cur
->bc_private
.b
.ip
;
540 struct xfs_trans
*tp
= cur
->bc_tp
;
541 xfs_fsblock_t fsbno
= XFS_DADDR_TO_FSB(mp
, XFS_BUF_ADDR(bp
));
542 struct xfs_owner_info oinfo
;
544 xfs_rmap_ino_bmbt_owner(&oinfo
, ip
->i_ino
, cur
->bc_private
.b
.whichfork
);
545 xfs_bmap_add_free(mp
, cur
->bc_private
.b
.dfops
, fsbno
, 1, &oinfo
);
546 ip
->i_d
.di_nblocks
--;
548 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
549 xfs_trans_mod_dquot_byino(tp
, ip
, XFS_TRANS_DQ_BCOUNT
, -1L);
554 xfs_bmbt_get_minrecs(
555 struct xfs_btree_cur
*cur
,
558 if (level
== cur
->bc_nlevels
- 1) {
559 struct xfs_ifork
*ifp
;
561 ifp
= XFS_IFORK_PTR(cur
->bc_private
.b
.ip
,
562 cur
->bc_private
.b
.whichfork
);
564 return xfs_bmbt_maxrecs(cur
->bc_mp
,
565 ifp
->if_broot_bytes
, level
== 0) / 2;
568 return cur
->bc_mp
->m_bmap_dmnr
[level
!= 0];
572 xfs_bmbt_get_maxrecs(
573 struct xfs_btree_cur
*cur
,
576 if (level
== cur
->bc_nlevels
- 1) {
577 struct xfs_ifork
*ifp
;
579 ifp
= XFS_IFORK_PTR(cur
->bc_private
.b
.ip
,
580 cur
->bc_private
.b
.whichfork
);
582 return xfs_bmbt_maxrecs(cur
->bc_mp
,
583 ifp
->if_broot_bytes
, level
== 0);
586 return cur
->bc_mp
->m_bmap_dmxr
[level
!= 0];
591 * Get the maximum records we could store in the on-disk format.
593 * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
594 * for the root node this checks the available space in the dinode fork
595 * so that we can resize the in-memory buffer to match it. After a
596 * resize to the maximum size this function returns the same value
597 * as xfs_bmbt_get_maxrecs for the root node, too.
600 xfs_bmbt_get_dmaxrecs(
601 struct xfs_btree_cur
*cur
,
604 if (level
!= cur
->bc_nlevels
- 1)
605 return cur
->bc_mp
->m_bmap_dmxr
[level
!= 0];
606 return xfs_bmdr_maxrecs(cur
->bc_private
.b
.forksize
, level
== 0);
610 xfs_bmbt_init_key_from_rec(
611 union xfs_btree_key
*key
,
612 union xfs_btree_rec
*rec
)
614 key
->bmbt
.br_startoff
=
615 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec
->bmbt
));
619 xfs_bmbt_init_rec_from_cur(
620 struct xfs_btree_cur
*cur
,
621 union xfs_btree_rec
*rec
)
623 xfs_bmbt_disk_set_all(&rec
->bmbt
, &cur
->bc_rec
.b
);
627 xfs_bmbt_init_ptr_from_cur(
628 struct xfs_btree_cur
*cur
,
629 union xfs_btree_ptr
*ptr
)
636 struct xfs_btree_cur
*cur
,
637 union xfs_btree_key
*key
)
639 return (__int64_t
)be64_to_cpu(key
->bmbt
.br_startoff
) -
640 cur
->bc_rec
.b
.br_startoff
;
647 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
648 struct xfs_btree_block
*block
= XFS_BUF_TO_BLOCK(bp
);
651 switch (block
->bb_magic
) {
652 case cpu_to_be32(XFS_BMAP_CRC_MAGIC
):
653 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
655 if (!uuid_equal(&block
->bb_u
.l
.bb_uuid
, &mp
->m_sb
.sb_meta_uuid
))
657 if (be64_to_cpu(block
->bb_u
.l
.bb_blkno
) != bp
->b_bn
)
660 * XXX: need a better way of verifying the owner here. Right now
661 * just make sure there has been one set.
663 if (be64_to_cpu(block
->bb_u
.l
.bb_owner
) == 0)
666 case cpu_to_be32(XFS_BMAP_MAGIC
):
673 * numrecs and level verification.
675 * We don't know what fork we belong to, so just verify that the level
676 * is less than the maximum of the two. Later checks will be more
679 level
= be16_to_cpu(block
->bb_level
);
680 if (level
> max(mp
->m_bm_maxlevels
[0], mp
->m_bm_maxlevels
[1]))
682 if (be16_to_cpu(block
->bb_numrecs
) > mp
->m_bmap_dmxr
[level
!= 0])
685 /* sibling pointer verification */
686 if (!block
->bb_u
.l
.bb_leftsib
||
687 (block
->bb_u
.l
.bb_leftsib
!= cpu_to_be64(NULLFSBLOCK
) &&
688 !XFS_FSB_SANITY_CHECK(mp
, be64_to_cpu(block
->bb_u
.l
.bb_leftsib
))))
690 if (!block
->bb_u
.l
.bb_rightsib
||
691 (block
->bb_u
.l
.bb_rightsib
!= cpu_to_be64(NULLFSBLOCK
) &&
692 !XFS_FSB_SANITY_CHECK(mp
, be64_to_cpu(block
->bb_u
.l
.bb_rightsib
))))
699 xfs_bmbt_read_verify(
702 if (!xfs_btree_lblock_verify_crc(bp
))
703 xfs_buf_ioerror(bp
, -EFSBADCRC
);
704 else if (!xfs_bmbt_verify(bp
))
705 xfs_buf_ioerror(bp
, -EFSCORRUPTED
);
708 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
709 xfs_verifier_error(bp
);
714 xfs_bmbt_write_verify(
717 if (!xfs_bmbt_verify(bp
)) {
718 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
719 xfs_buf_ioerror(bp
, -EFSCORRUPTED
);
720 xfs_verifier_error(bp
);
723 xfs_btree_lblock_calc_crc(bp
);
726 const struct xfs_buf_ops xfs_bmbt_buf_ops
= {
728 .verify_read
= xfs_bmbt_read_verify
,
729 .verify_write
= xfs_bmbt_write_verify
,
733 #if defined(DEBUG) || defined(XFS_WARN)
735 xfs_bmbt_keys_inorder(
736 struct xfs_btree_cur
*cur
,
737 union xfs_btree_key
*k1
,
738 union xfs_btree_key
*k2
)
740 return be64_to_cpu(k1
->bmbt
.br_startoff
) <
741 be64_to_cpu(k2
->bmbt
.br_startoff
);
745 xfs_bmbt_recs_inorder(
746 struct xfs_btree_cur
*cur
,
747 union xfs_btree_rec
*r1
,
748 union xfs_btree_rec
*r2
)
750 return xfs_bmbt_disk_get_startoff(&r1
->bmbt
) +
751 xfs_bmbt_disk_get_blockcount(&r1
->bmbt
) <=
752 xfs_bmbt_disk_get_startoff(&r2
->bmbt
);
756 static const struct xfs_btree_ops xfs_bmbt_ops
= {
757 .rec_len
= sizeof(xfs_bmbt_rec_t
),
758 .key_len
= sizeof(xfs_bmbt_key_t
),
760 .dup_cursor
= xfs_bmbt_dup_cursor
,
761 .update_cursor
= xfs_bmbt_update_cursor
,
762 .alloc_block
= xfs_bmbt_alloc_block
,
763 .free_block
= xfs_bmbt_free_block
,
764 .get_maxrecs
= xfs_bmbt_get_maxrecs
,
765 .get_minrecs
= xfs_bmbt_get_minrecs
,
766 .get_dmaxrecs
= xfs_bmbt_get_dmaxrecs
,
767 .init_key_from_rec
= xfs_bmbt_init_key_from_rec
,
768 .init_rec_from_cur
= xfs_bmbt_init_rec_from_cur
,
769 .init_ptr_from_cur
= xfs_bmbt_init_ptr_from_cur
,
770 .key_diff
= xfs_bmbt_key_diff
,
771 .buf_ops
= &xfs_bmbt_buf_ops
,
772 #if defined(DEBUG) || defined(XFS_WARN)
773 .keys_inorder
= xfs_bmbt_keys_inorder
,
774 .recs_inorder
= xfs_bmbt_recs_inorder
,
779 * Allocate a new bmap btree cursor.
781 struct xfs_btree_cur
* /* new bmap btree cursor */
782 xfs_bmbt_init_cursor(
783 struct xfs_mount
*mp
, /* file system mount point */
784 struct xfs_trans
*tp
, /* transaction pointer */
785 struct xfs_inode
*ip
, /* inode owning the btree */
786 int whichfork
) /* data or attr fork */
788 struct xfs_ifork
*ifp
= XFS_IFORK_PTR(ip
, whichfork
);
789 struct xfs_btree_cur
*cur
;
790 ASSERT(whichfork
!= XFS_COW_FORK
);
792 cur
= kmem_zone_zalloc(xfs_btree_cur_zone
, KM_NOFS
);
796 cur
->bc_nlevels
= be16_to_cpu(ifp
->if_broot
->bb_level
) + 1;
797 cur
->bc_btnum
= XFS_BTNUM_BMAP
;
798 cur
->bc_blocklog
= mp
->m_sb
.sb_blocklog
;
799 cur
->bc_statoff
= XFS_STATS_CALC_INDEX(xs_bmbt_2
);
801 cur
->bc_ops
= &xfs_bmbt_ops
;
802 cur
->bc_flags
= XFS_BTREE_LONG_PTRS
| XFS_BTREE_ROOT_IN_INODE
;
803 if (xfs_sb_version_hascrc(&mp
->m_sb
))
804 cur
->bc_flags
|= XFS_BTREE_CRC_BLOCKS
;
806 cur
->bc_private
.b
.forksize
= XFS_IFORK_SIZE(ip
, whichfork
);
807 cur
->bc_private
.b
.ip
= ip
;
808 cur
->bc_private
.b
.firstblock
= NULLFSBLOCK
;
809 cur
->bc_private
.b
.dfops
= NULL
;
810 cur
->bc_private
.b
.allocated
= 0;
811 cur
->bc_private
.b
.flags
= 0;
812 cur
->bc_private
.b
.whichfork
= whichfork
;
818 * Calculate number of records in a bmap btree block.
822 struct xfs_mount
*mp
,
826 blocklen
-= XFS_BMBT_BLOCK_LEN(mp
);
829 return blocklen
/ sizeof(xfs_bmbt_rec_t
);
830 return blocklen
/ (sizeof(xfs_bmbt_key_t
) + sizeof(xfs_bmbt_ptr_t
));
834 * Calculate number of records in a bmap btree inode root.
841 blocklen
-= sizeof(xfs_bmdr_block_t
);
844 return blocklen
/ sizeof(xfs_bmdr_rec_t
);
845 return blocklen
/ (sizeof(xfs_bmdr_key_t
) + sizeof(xfs_bmdr_ptr_t
));
849 * Change the owner of a btree format fork fo the inode passed in. Change it to
850 * the owner of that is passed in so that we can change owners before or after
851 * we switch forks between inodes. The operation that the caller is doing will
852 * determine whether is needs to change owner before or after the switch.
854 * For demand paged transactional modification, the fork switch should be done
855 * after reading in all the blocks, modifying them and pinning them in the
856 * transaction. For modification when the buffers are already pinned in memory,
857 * the fork switch can be done before changing the owner as we won't need to
858 * validate the owner until the btree buffers are unpinned and writes can occur
861 * For recovery based ownership change, there is no transactional context and
862 * so a buffer list must be supplied so that we can record the buffers that we
863 * modified for the caller to issue IO on.
866 xfs_bmbt_change_owner(
867 struct xfs_trans
*tp
,
868 struct xfs_inode
*ip
,
871 struct list_head
*buffer_list
)
873 struct xfs_btree_cur
*cur
;
876 ASSERT(tp
|| buffer_list
);
877 ASSERT(!(tp
&& buffer_list
));
878 if (whichfork
== XFS_DATA_FORK
)
879 ASSERT(ip
->i_d
.di_format
== XFS_DINODE_FMT_BTREE
);
881 ASSERT(ip
->i_d
.di_aformat
== XFS_DINODE_FMT_BTREE
);
883 cur
= xfs_bmbt_init_cursor(ip
->i_mount
, tp
, ip
, whichfork
);
887 error
= xfs_btree_change_owner(cur
, new_owner
, buffer_list
);
888 xfs_btree_del_cursor(cur
, error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);