2 * Copyright (c) 2000-2001,2005 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_inode.h"
27 #include "xfs_btree.h"
28 #include "xfs_ialloc.h"
29 #include "xfs_ialloc_btree.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_trace.h"
33 #include "xfs_cksum.h"
34 #include "xfs_trans.h"
39 xfs_inobt_get_minrecs(
40 struct xfs_btree_cur
*cur
,
43 return cur
->bc_mp
->m_inobt_mnr
[level
!= 0];
46 STATIC
struct xfs_btree_cur
*
48 struct xfs_btree_cur
*cur
)
50 return xfs_inobt_init_cursor(cur
->bc_mp
, cur
->bc_tp
,
51 cur
->bc_private
.a
.agbp
, cur
->bc_private
.a
.agno
,
57 struct xfs_btree_cur
*cur
,
58 union xfs_btree_ptr
*nptr
,
59 int inc
) /* level change */
61 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
62 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
64 agi
->agi_root
= nptr
->s
;
65 be32_add_cpu(&agi
->agi_level
, inc
);
66 xfs_ialloc_log_agi(cur
->bc_tp
, agbp
, XFS_AGI_ROOT
| XFS_AGI_LEVEL
);
71 struct xfs_btree_cur
*cur
,
72 union xfs_btree_ptr
*nptr
,
73 int inc
) /* level change */
75 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
76 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
78 agi
->agi_free_root
= nptr
->s
;
79 be32_add_cpu(&agi
->agi_free_level
, inc
);
80 xfs_ialloc_log_agi(cur
->bc_tp
, agbp
,
81 XFS_AGI_FREE_ROOT
| XFS_AGI_FREE_LEVEL
);
85 __xfs_inobt_alloc_block(
86 struct xfs_btree_cur
*cur
,
87 union xfs_btree_ptr
*start
,
88 union xfs_btree_ptr
*new,
90 enum xfs_ag_resv_type resv
)
92 xfs_alloc_arg_t args
; /* block allocation args */
93 int error
; /* error return value */
94 xfs_agblock_t sbno
= be32_to_cpu(start
->s
);
96 XFS_BTREE_TRACE_CURSOR(cur
, XBT_ENTRY
);
98 memset(&args
, 0, sizeof(args
));
100 args
.mp
= cur
->bc_mp
;
101 xfs_rmap_ag_owner(&args
.oinfo
, XFS_RMAP_OWN_INOBT
);
102 args
.fsbno
= XFS_AGB_TO_FSB(args
.mp
, cur
->bc_private
.a
.agno
, sbno
);
106 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
109 error
= xfs_alloc_vextent(&args
);
111 XFS_BTREE_TRACE_CURSOR(cur
, XBT_ERROR
);
114 if (args
.fsbno
== NULLFSBLOCK
) {
115 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
119 ASSERT(args
.len
== 1);
120 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
122 new->s
= cpu_to_be32(XFS_FSB_TO_AGBNO(args
.mp
, args
.fsbno
));
128 xfs_inobt_alloc_block(
129 struct xfs_btree_cur
*cur
,
130 union xfs_btree_ptr
*start
,
131 union xfs_btree_ptr
*new,
134 return __xfs_inobt_alloc_block(cur
, start
, new, stat
, XFS_AG_RESV_NONE
);
138 xfs_finobt_alloc_block(
139 struct xfs_btree_cur
*cur
,
140 union xfs_btree_ptr
*start
,
141 union xfs_btree_ptr
*new,
144 return __xfs_inobt_alloc_block(cur
, start
, new, stat
,
145 XFS_AG_RESV_METADATA
);
149 xfs_inobt_free_block(
150 struct xfs_btree_cur
*cur
,
153 struct xfs_owner_info oinfo
;
155 xfs_rmap_ag_owner(&oinfo
, XFS_RMAP_OWN_INOBT
);
156 return xfs_free_extent(cur
->bc_tp
,
157 XFS_DADDR_TO_FSB(cur
->bc_mp
, XFS_BUF_ADDR(bp
)), 1,
158 &oinfo
, XFS_AG_RESV_NONE
);
162 xfs_inobt_get_maxrecs(
163 struct xfs_btree_cur
*cur
,
166 return cur
->bc_mp
->m_inobt_mxr
[level
!= 0];
170 xfs_inobt_init_key_from_rec(
171 union xfs_btree_key
*key
,
172 union xfs_btree_rec
*rec
)
174 key
->inobt
.ir_startino
= rec
->inobt
.ir_startino
;
178 xfs_inobt_init_high_key_from_rec(
179 union xfs_btree_key
*key
,
180 union xfs_btree_rec
*rec
)
184 x
= be32_to_cpu(rec
->inobt
.ir_startino
);
185 x
+= XFS_INODES_PER_CHUNK
- 1;
186 key
->inobt
.ir_startino
= cpu_to_be32(x
);
190 xfs_inobt_init_rec_from_cur(
191 struct xfs_btree_cur
*cur
,
192 union xfs_btree_rec
*rec
)
194 rec
->inobt
.ir_startino
= cpu_to_be32(cur
->bc_rec
.i
.ir_startino
);
195 if (xfs_sb_version_hassparseinodes(&cur
->bc_mp
->m_sb
)) {
196 rec
->inobt
.ir_u
.sp
.ir_holemask
=
197 cpu_to_be16(cur
->bc_rec
.i
.ir_holemask
);
198 rec
->inobt
.ir_u
.sp
.ir_count
= cur
->bc_rec
.i
.ir_count
;
199 rec
->inobt
.ir_u
.sp
.ir_freecount
= cur
->bc_rec
.i
.ir_freecount
;
201 /* ir_holemask/ir_count not supported on-disk */
202 rec
->inobt
.ir_u
.f
.ir_freecount
=
203 cpu_to_be32(cur
->bc_rec
.i
.ir_freecount
);
205 rec
->inobt
.ir_free
= cpu_to_be64(cur
->bc_rec
.i
.ir_free
);
209 * initial value of ptr for lookup
212 xfs_inobt_init_ptr_from_cur(
213 struct xfs_btree_cur
*cur
,
214 union xfs_btree_ptr
*ptr
)
216 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(cur
->bc_private
.a
.agbp
);
218 ASSERT(cur
->bc_private
.a
.agno
== be32_to_cpu(agi
->agi_seqno
));
220 ptr
->s
= agi
->agi_root
;
224 xfs_finobt_init_ptr_from_cur(
225 struct xfs_btree_cur
*cur
,
226 union xfs_btree_ptr
*ptr
)
228 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(cur
->bc_private
.a
.agbp
);
230 ASSERT(cur
->bc_private
.a
.agno
== be32_to_cpu(agi
->agi_seqno
));
231 ptr
->s
= agi
->agi_free_root
;
236 struct xfs_btree_cur
*cur
,
237 union xfs_btree_key
*key
)
239 return (int64_t)be32_to_cpu(key
->inobt
.ir_startino
) -
240 cur
->bc_rec
.i
.ir_startino
;
244 xfs_inobt_diff_two_keys(
245 struct xfs_btree_cur
*cur
,
246 union xfs_btree_key
*k1
,
247 union xfs_btree_key
*k2
)
249 return (int64_t)be32_to_cpu(k1
->inobt
.ir_startino
) -
250 be32_to_cpu(k2
->inobt
.ir_startino
);
257 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
258 struct xfs_btree_block
*block
= XFS_BUF_TO_BLOCK(bp
);
262 * During growfs operations, we can't verify the exact owner as the
263 * perag is not fully initialised and hence not attached to the buffer.
265 * Similarly, during log recovery we will have a perag structure
266 * attached, but the agi information will not yet have been initialised
267 * from the on disk AGI. We don't currently use any of this information,
268 * but beware of the landmine (i.e. need to check pag->pagi_init) if we
271 switch (block
->bb_magic
) {
272 case cpu_to_be32(XFS_IBT_CRC_MAGIC
):
273 case cpu_to_be32(XFS_FIBT_CRC_MAGIC
):
274 if (!xfs_btree_sblock_v5hdr_verify(bp
))
277 case cpu_to_be32(XFS_IBT_MAGIC
):
278 case cpu_to_be32(XFS_FIBT_MAGIC
):
284 /* level verification */
285 level
= be16_to_cpu(block
->bb_level
);
286 if (level
>= mp
->m_in_maxlevels
)
289 return xfs_btree_sblock_verify(bp
, mp
->m_inobt_mxr
[level
!= 0]);
293 xfs_inobt_read_verify(
296 if (!xfs_btree_sblock_verify_crc(bp
))
297 xfs_buf_ioerror(bp
, -EFSBADCRC
);
298 else if (!xfs_inobt_verify(bp
))
299 xfs_buf_ioerror(bp
, -EFSCORRUPTED
);
302 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
303 xfs_verifier_error(bp
);
308 xfs_inobt_write_verify(
311 if (!xfs_inobt_verify(bp
)) {
312 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
313 xfs_buf_ioerror(bp
, -EFSCORRUPTED
);
314 xfs_verifier_error(bp
);
317 xfs_btree_sblock_calc_crc(bp
);
321 const struct xfs_buf_ops xfs_inobt_buf_ops
= {
323 .verify_read
= xfs_inobt_read_verify
,
324 .verify_write
= xfs_inobt_write_verify
,
328 xfs_inobt_keys_inorder(
329 struct xfs_btree_cur
*cur
,
330 union xfs_btree_key
*k1
,
331 union xfs_btree_key
*k2
)
333 return be32_to_cpu(k1
->inobt
.ir_startino
) <
334 be32_to_cpu(k2
->inobt
.ir_startino
);
338 xfs_inobt_recs_inorder(
339 struct xfs_btree_cur
*cur
,
340 union xfs_btree_rec
*r1
,
341 union xfs_btree_rec
*r2
)
343 return be32_to_cpu(r1
->inobt
.ir_startino
) + XFS_INODES_PER_CHUNK
<=
344 be32_to_cpu(r2
->inobt
.ir_startino
);
347 static const struct xfs_btree_ops xfs_inobt_ops
= {
348 .rec_len
= sizeof(xfs_inobt_rec_t
),
349 .key_len
= sizeof(xfs_inobt_key_t
),
351 .dup_cursor
= xfs_inobt_dup_cursor
,
352 .set_root
= xfs_inobt_set_root
,
353 .alloc_block
= xfs_inobt_alloc_block
,
354 .free_block
= xfs_inobt_free_block
,
355 .get_minrecs
= xfs_inobt_get_minrecs
,
356 .get_maxrecs
= xfs_inobt_get_maxrecs
,
357 .init_key_from_rec
= xfs_inobt_init_key_from_rec
,
358 .init_high_key_from_rec
= xfs_inobt_init_high_key_from_rec
,
359 .init_rec_from_cur
= xfs_inobt_init_rec_from_cur
,
360 .init_ptr_from_cur
= xfs_inobt_init_ptr_from_cur
,
361 .key_diff
= xfs_inobt_key_diff
,
362 .buf_ops
= &xfs_inobt_buf_ops
,
363 .diff_two_keys
= xfs_inobt_diff_two_keys
,
364 .keys_inorder
= xfs_inobt_keys_inorder
,
365 .recs_inorder
= xfs_inobt_recs_inorder
,
368 static const struct xfs_btree_ops xfs_finobt_ops
= {
369 .rec_len
= sizeof(xfs_inobt_rec_t
),
370 .key_len
= sizeof(xfs_inobt_key_t
),
372 .dup_cursor
= xfs_inobt_dup_cursor
,
373 .set_root
= xfs_finobt_set_root
,
374 .alloc_block
= xfs_finobt_alloc_block
,
375 .free_block
= xfs_inobt_free_block
,
376 .get_minrecs
= xfs_inobt_get_minrecs
,
377 .get_maxrecs
= xfs_inobt_get_maxrecs
,
378 .init_key_from_rec
= xfs_inobt_init_key_from_rec
,
379 .init_high_key_from_rec
= xfs_inobt_init_high_key_from_rec
,
380 .init_rec_from_cur
= xfs_inobt_init_rec_from_cur
,
381 .init_ptr_from_cur
= xfs_finobt_init_ptr_from_cur
,
382 .key_diff
= xfs_inobt_key_diff
,
383 .buf_ops
= &xfs_inobt_buf_ops
,
384 .diff_two_keys
= xfs_inobt_diff_two_keys
,
385 .keys_inorder
= xfs_inobt_keys_inorder
,
386 .recs_inorder
= xfs_inobt_recs_inorder
,
390 * Allocate a new inode btree cursor.
392 struct xfs_btree_cur
* /* new inode btree cursor */
393 xfs_inobt_init_cursor(
394 struct xfs_mount
*mp
, /* file system mount point */
395 struct xfs_trans
*tp
, /* transaction pointer */
396 struct xfs_buf
*agbp
, /* buffer for agi structure */
397 xfs_agnumber_t agno
, /* allocation group number */
398 xfs_btnum_t btnum
) /* ialloc or free ino btree */
400 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
401 struct xfs_btree_cur
*cur
;
403 cur
= kmem_zone_zalloc(xfs_btree_cur_zone
, KM_NOFS
);
407 cur
->bc_btnum
= btnum
;
408 if (btnum
== XFS_BTNUM_INO
) {
409 cur
->bc_nlevels
= be32_to_cpu(agi
->agi_level
);
410 cur
->bc_ops
= &xfs_inobt_ops
;
411 cur
->bc_statoff
= XFS_STATS_CALC_INDEX(xs_ibt_2
);
413 cur
->bc_nlevels
= be32_to_cpu(agi
->agi_free_level
);
414 cur
->bc_ops
= &xfs_finobt_ops
;
415 cur
->bc_statoff
= XFS_STATS_CALC_INDEX(xs_fibt_2
);
418 cur
->bc_blocklog
= mp
->m_sb
.sb_blocklog
;
420 if (xfs_sb_version_hascrc(&mp
->m_sb
))
421 cur
->bc_flags
|= XFS_BTREE_CRC_BLOCKS
;
423 cur
->bc_private
.a
.agbp
= agbp
;
424 cur
->bc_private
.a
.agno
= agno
;
430 * Calculate number of records in an inobt btree block.
434 struct xfs_mount
*mp
,
438 blocklen
-= XFS_INOBT_BLOCK_LEN(mp
);
441 return blocklen
/ sizeof(xfs_inobt_rec_t
);
442 return blocklen
/ (sizeof(xfs_inobt_key_t
) + sizeof(xfs_inobt_ptr_t
));
446 * Convert the inode record holemask to an inode allocation bitmap. The inode
447 * allocation bitmap is inode granularity and specifies whether an inode is
448 * physically allocated on disk (not whether the inode is considered allocated
449 * or free by the fs).
451 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
454 xfs_inobt_irec_to_allocmask(
455 struct xfs_inobt_rec_incore
*rec
)
463 * The holemask has 16-bits for a 64 inode record. Therefore each
464 * holemask bit represents multiple inodes. Create a mask of bits to set
465 * in the allocmask for each holemask bit.
467 inodespbit
= (1 << XFS_INODES_PER_HOLEMASK_BIT
) - 1;
470 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
471 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
472 * anything beyond the 16 holemask bits since this casts to a larger
475 allocbitmap
= ~rec
->ir_holemask
& ((1 << XFS_INOBT_HOLEMASK_BITS
) - 1);
478 * allocbitmap is the inverted holemask so every set bit represents
479 * allocated inodes. To expand from 16-bit holemask granularity to
480 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
481 * bitmap for every holemask bit.
483 nextbit
= xfs_next_bit(&allocbitmap
, 1, 0);
484 while (nextbit
!= -1) {
485 ASSERT(nextbit
< (sizeof(rec
->ir_holemask
) * NBBY
));
487 bitmap
|= (inodespbit
<<
488 (nextbit
* XFS_INODES_PER_HOLEMASK_BIT
));
490 nextbit
= xfs_next_bit(&allocbitmap
, 1, nextbit
+ 1);
496 #if defined(DEBUG) || defined(XFS_WARN)
498 * Verify that an in-core inode record has a valid inode count.
501 xfs_inobt_rec_check_count(
502 struct xfs_mount
*mp
,
503 struct xfs_inobt_rec_incore
*rec
)
510 wordsz
= sizeof(allocbmap
) / sizeof(unsigned int);
511 allocbmap
= xfs_inobt_irec_to_allocmask(rec
);
513 nextbit
= xfs_next_bit((uint
*) &allocbmap
, wordsz
, nextbit
);
514 while (nextbit
!= -1) {
516 nextbit
= xfs_next_bit((uint
*) &allocbmap
, wordsz
,
520 if (inocount
!= rec
->ir_count
)
521 return -EFSCORRUPTED
;
529 struct xfs_mount
*mp
)
531 /* Bail out if we're uninitialized, which can happen in mkfs. */
532 if (mp
->m_inobt_mxr
[0] == 0)
535 return xfs_btree_calc_size(mp
, mp
->m_inobt_mnr
,
536 (uint64_t)mp
->m_sb
.sb_agblocks
* mp
->m_sb
.sb_inopblock
/
537 XFS_INODES_PER_CHUNK
);
541 xfs_inobt_count_blocks(
542 struct xfs_mount
*mp
,
545 xfs_extlen_t
*tree_blocks
)
547 struct xfs_buf
*agbp
;
548 struct xfs_btree_cur
*cur
;
551 error
= xfs_ialloc_read_agi(mp
, NULL
, agno
, &agbp
);
555 cur
= xfs_inobt_init_cursor(mp
, NULL
, agbp
, agno
, btnum
);
556 error
= xfs_btree_count_blocks(cur
, tree_blocks
);
557 xfs_btree_del_cursor(cur
, error
? XFS_BTREE_ERROR
: XFS_BTREE_NOERROR
);
564 * Figure out how many blocks to reserve and how many are used by this btree.
567 xfs_finobt_calc_reserves(
568 struct xfs_mount
*mp
,
573 xfs_extlen_t tree_len
= 0;
576 if (!xfs_sb_version_hasfinobt(&mp
->m_sb
))
579 error
= xfs_inobt_count_blocks(mp
, agno
, XFS_BTNUM_FINO
, &tree_len
);
583 *ask
+= xfs_inobt_max_size(mp
);