1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
6 #include "libxfs_priv.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
13 #include "xfs_mount.h"
14 #include "xfs_inode.h"
15 #include "xfs_btree.h"
16 #include "xfs_ialloc.h"
17 #include "xfs_ialloc_btree.h"
18 #include "xfs_alloc.h"
19 #include "xfs_trace.h"
20 #include "xfs_cksum.h"
21 #include "xfs_trans.h"
26 xfs_inobt_get_minrecs(
27 struct xfs_btree_cur
*cur
,
30 return cur
->bc_mp
->m_inobt_mnr
[level
!= 0];
33 STATIC
struct xfs_btree_cur
*
35 struct xfs_btree_cur
*cur
)
37 return xfs_inobt_init_cursor(cur
->bc_mp
, cur
->bc_tp
,
38 cur
->bc_private
.a
.agbp
, cur
->bc_private
.a
.agno
,
44 struct xfs_btree_cur
*cur
,
45 union xfs_btree_ptr
*nptr
,
46 int inc
) /* level change */
48 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
49 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
51 agi
->agi_root
= nptr
->s
;
52 be32_add_cpu(&agi
->agi_level
, inc
);
53 xfs_ialloc_log_agi(cur
->bc_tp
, agbp
, XFS_AGI_ROOT
| XFS_AGI_LEVEL
);
58 struct xfs_btree_cur
*cur
,
59 union xfs_btree_ptr
*nptr
,
60 int inc
) /* level change */
62 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
63 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
65 agi
->agi_free_root
= nptr
->s
;
66 be32_add_cpu(&agi
->agi_free_level
, inc
);
67 xfs_ialloc_log_agi(cur
->bc_tp
, agbp
,
68 XFS_AGI_FREE_ROOT
| XFS_AGI_FREE_LEVEL
);
72 __xfs_inobt_alloc_block(
73 struct xfs_btree_cur
*cur
,
74 union xfs_btree_ptr
*start
,
75 union xfs_btree_ptr
*new,
77 enum xfs_ag_resv_type resv
)
79 xfs_alloc_arg_t args
; /* block allocation args */
80 int error
; /* error return value */
81 xfs_agblock_t sbno
= be32_to_cpu(start
->s
);
83 memset(&args
, 0, sizeof(args
));
86 xfs_rmap_ag_owner(&args
.oinfo
, XFS_RMAP_OWN_INOBT
);
87 args
.fsbno
= XFS_AGB_TO_FSB(args
.mp
, cur
->bc_private
.a
.agno
, sbno
);
91 args
.type
= XFS_ALLOCTYPE_NEAR_BNO
;
94 error
= xfs_alloc_vextent(&args
);
98 if (args
.fsbno
== NULLFSBLOCK
) {
102 ASSERT(args
.len
== 1);
104 new->s
= cpu_to_be32(XFS_FSB_TO_AGBNO(args
.mp
, args
.fsbno
));
110 xfs_inobt_alloc_block(
111 struct xfs_btree_cur
*cur
,
112 union xfs_btree_ptr
*start
,
113 union xfs_btree_ptr
*new,
116 return __xfs_inobt_alloc_block(cur
, start
, new, stat
, XFS_AG_RESV_NONE
);
120 xfs_finobt_alloc_block(
121 struct xfs_btree_cur
*cur
,
122 union xfs_btree_ptr
*start
,
123 union xfs_btree_ptr
*new,
126 if (cur
->bc_mp
->m_inotbt_nores
)
127 return xfs_inobt_alloc_block(cur
, start
, new, stat
);
128 return __xfs_inobt_alloc_block(cur
, start
, new, stat
,
129 XFS_AG_RESV_METADATA
);
133 __xfs_inobt_free_block(
134 struct xfs_btree_cur
*cur
,
136 enum xfs_ag_resv_type resv
)
138 struct xfs_owner_info oinfo
;
140 xfs_rmap_ag_owner(&oinfo
, XFS_RMAP_OWN_INOBT
);
141 return xfs_free_extent(cur
->bc_tp
,
142 XFS_DADDR_TO_FSB(cur
->bc_mp
, XFS_BUF_ADDR(bp
)), 1,
147 xfs_inobt_free_block(
148 struct xfs_btree_cur
*cur
,
151 return __xfs_inobt_free_block(cur
, bp
, XFS_AG_RESV_NONE
);
155 xfs_finobt_free_block(
156 struct xfs_btree_cur
*cur
,
159 if (cur
->bc_mp
->m_inotbt_nores
)
160 return xfs_inobt_free_block(cur
, bp
);
161 return __xfs_inobt_free_block(cur
, bp
, XFS_AG_RESV_METADATA
);
165 xfs_inobt_get_maxrecs(
166 struct xfs_btree_cur
*cur
,
169 return cur
->bc_mp
->m_inobt_mxr
[level
!= 0];
173 xfs_inobt_init_key_from_rec(
174 union xfs_btree_key
*key
,
175 union xfs_btree_rec
*rec
)
177 key
->inobt
.ir_startino
= rec
->inobt
.ir_startino
;
181 xfs_inobt_init_high_key_from_rec(
182 union xfs_btree_key
*key
,
183 union xfs_btree_rec
*rec
)
187 x
= be32_to_cpu(rec
->inobt
.ir_startino
);
188 x
+= XFS_INODES_PER_CHUNK
- 1;
189 key
->inobt
.ir_startino
= cpu_to_be32(x
);
193 xfs_inobt_init_rec_from_cur(
194 struct xfs_btree_cur
*cur
,
195 union xfs_btree_rec
*rec
)
197 rec
->inobt
.ir_startino
= cpu_to_be32(cur
->bc_rec
.i
.ir_startino
);
198 if (xfs_sb_version_hassparseinodes(&cur
->bc_mp
->m_sb
)) {
199 rec
->inobt
.ir_u
.sp
.ir_holemask
=
200 cpu_to_be16(cur
->bc_rec
.i
.ir_holemask
);
201 rec
->inobt
.ir_u
.sp
.ir_count
= cur
->bc_rec
.i
.ir_count
;
202 rec
->inobt
.ir_u
.sp
.ir_freecount
= cur
->bc_rec
.i
.ir_freecount
;
204 /* ir_holemask/ir_count not supported on-disk */
205 rec
->inobt
.ir_u
.f
.ir_freecount
=
206 cpu_to_be32(cur
->bc_rec
.i
.ir_freecount
);
208 rec
->inobt
.ir_free
= cpu_to_be64(cur
->bc_rec
.i
.ir_free
);
212 * initial value of ptr for lookup
215 xfs_inobt_init_ptr_from_cur(
216 struct xfs_btree_cur
*cur
,
217 union xfs_btree_ptr
*ptr
)
219 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(cur
->bc_private
.a
.agbp
);
221 ASSERT(cur
->bc_private
.a
.agno
== be32_to_cpu(agi
->agi_seqno
));
223 ptr
->s
= agi
->agi_root
;
227 xfs_finobt_init_ptr_from_cur(
228 struct xfs_btree_cur
*cur
,
229 union xfs_btree_ptr
*ptr
)
231 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(cur
->bc_private
.a
.agbp
);
233 ASSERT(cur
->bc_private
.a
.agno
== be32_to_cpu(agi
->agi_seqno
));
234 ptr
->s
= agi
->agi_free_root
;
239 struct xfs_btree_cur
*cur
,
240 union xfs_btree_key
*key
)
242 return (int64_t)be32_to_cpu(key
->inobt
.ir_startino
) -
243 cur
->bc_rec
.i
.ir_startino
;
247 xfs_inobt_diff_two_keys(
248 struct xfs_btree_cur
*cur
,
249 union xfs_btree_key
*k1
,
250 union xfs_btree_key
*k2
)
252 return (int64_t)be32_to_cpu(k1
->inobt
.ir_startino
) -
253 be32_to_cpu(k2
->inobt
.ir_startino
);
256 static xfs_failaddr_t
260 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
261 struct xfs_btree_block
*block
= XFS_BUF_TO_BLOCK(bp
);
266 * During growfs operations, we can't verify the exact owner as the
267 * perag is not fully initialised and hence not attached to the buffer.
269 * Similarly, during log recovery we will have a perag structure
270 * attached, but the agi information will not yet have been initialised
271 * from the on disk AGI. We don't currently use any of this information,
272 * but beware of the landmine (i.e. need to check pag->pagi_init) if we
275 switch (block
->bb_magic
) {
276 case cpu_to_be32(XFS_IBT_CRC_MAGIC
):
277 case cpu_to_be32(XFS_FIBT_CRC_MAGIC
):
278 fa
= xfs_btree_sblock_v5hdr_verify(bp
);
282 case cpu_to_be32(XFS_IBT_MAGIC
):
283 case cpu_to_be32(XFS_FIBT_MAGIC
):
286 return __this_address
;
289 /* level verification */
290 level
= be16_to_cpu(block
->bb_level
);
291 if (level
>= mp
->m_in_maxlevels
)
292 return __this_address
;
294 return xfs_btree_sblock_verify(bp
, mp
->m_inobt_mxr
[level
!= 0]);
298 xfs_inobt_read_verify(
303 if (!xfs_btree_sblock_verify_crc(bp
))
304 xfs_verifier_error(bp
, -EFSBADCRC
, __this_address
);
306 fa
= xfs_inobt_verify(bp
);
308 xfs_verifier_error(bp
, -EFSCORRUPTED
, fa
);
312 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
316 xfs_inobt_write_verify(
321 fa
= xfs_inobt_verify(bp
);
323 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
324 xfs_verifier_error(bp
, -EFSCORRUPTED
, fa
);
327 xfs_btree_sblock_calc_crc(bp
);
331 const struct xfs_buf_ops xfs_inobt_buf_ops
= {
333 .verify_read
= xfs_inobt_read_verify
,
334 .verify_write
= xfs_inobt_write_verify
,
335 .verify_struct
= xfs_inobt_verify
,
339 xfs_inobt_keys_inorder(
340 struct xfs_btree_cur
*cur
,
341 union xfs_btree_key
*k1
,
342 union xfs_btree_key
*k2
)
344 return be32_to_cpu(k1
->inobt
.ir_startino
) <
345 be32_to_cpu(k2
->inobt
.ir_startino
);
349 xfs_inobt_recs_inorder(
350 struct xfs_btree_cur
*cur
,
351 union xfs_btree_rec
*r1
,
352 union xfs_btree_rec
*r2
)
354 return be32_to_cpu(r1
->inobt
.ir_startino
) + XFS_INODES_PER_CHUNK
<=
355 be32_to_cpu(r2
->inobt
.ir_startino
);
358 static const struct xfs_btree_ops xfs_inobt_ops
= {
359 .rec_len
= sizeof(xfs_inobt_rec_t
),
360 .key_len
= sizeof(xfs_inobt_key_t
),
362 .dup_cursor
= xfs_inobt_dup_cursor
,
363 .set_root
= xfs_inobt_set_root
,
364 .alloc_block
= xfs_inobt_alloc_block
,
365 .free_block
= xfs_inobt_free_block
,
366 .get_minrecs
= xfs_inobt_get_minrecs
,
367 .get_maxrecs
= xfs_inobt_get_maxrecs
,
368 .init_key_from_rec
= xfs_inobt_init_key_from_rec
,
369 .init_high_key_from_rec
= xfs_inobt_init_high_key_from_rec
,
370 .init_rec_from_cur
= xfs_inobt_init_rec_from_cur
,
371 .init_ptr_from_cur
= xfs_inobt_init_ptr_from_cur
,
372 .key_diff
= xfs_inobt_key_diff
,
373 .buf_ops
= &xfs_inobt_buf_ops
,
374 .diff_two_keys
= xfs_inobt_diff_two_keys
,
375 .keys_inorder
= xfs_inobt_keys_inorder
,
376 .recs_inorder
= xfs_inobt_recs_inorder
,
379 static const struct xfs_btree_ops xfs_finobt_ops
= {
380 .rec_len
= sizeof(xfs_inobt_rec_t
),
381 .key_len
= sizeof(xfs_inobt_key_t
),
383 .dup_cursor
= xfs_inobt_dup_cursor
,
384 .set_root
= xfs_finobt_set_root
,
385 .alloc_block
= xfs_finobt_alloc_block
,
386 .free_block
= xfs_finobt_free_block
,
387 .get_minrecs
= xfs_inobt_get_minrecs
,
388 .get_maxrecs
= xfs_inobt_get_maxrecs
,
389 .init_key_from_rec
= xfs_inobt_init_key_from_rec
,
390 .init_high_key_from_rec
= xfs_inobt_init_high_key_from_rec
,
391 .init_rec_from_cur
= xfs_inobt_init_rec_from_cur
,
392 .init_ptr_from_cur
= xfs_finobt_init_ptr_from_cur
,
393 .key_diff
= xfs_inobt_key_diff
,
394 .buf_ops
= &xfs_inobt_buf_ops
,
395 .diff_two_keys
= xfs_inobt_diff_two_keys
,
396 .keys_inorder
= xfs_inobt_keys_inorder
,
397 .recs_inorder
= xfs_inobt_recs_inorder
,
401 * Allocate a new inode btree cursor.
403 struct xfs_btree_cur
* /* new inode btree cursor */
404 xfs_inobt_init_cursor(
405 struct xfs_mount
*mp
, /* file system mount point */
406 struct xfs_trans
*tp
, /* transaction pointer */
407 struct xfs_buf
*agbp
, /* buffer for agi structure */
408 xfs_agnumber_t agno
, /* allocation group number */
409 xfs_btnum_t btnum
) /* ialloc or free ino btree */
411 struct xfs_agi
*agi
= XFS_BUF_TO_AGI(agbp
);
412 struct xfs_btree_cur
*cur
;
414 cur
= kmem_zone_zalloc(xfs_btree_cur_zone
, KM_NOFS
);
418 cur
->bc_btnum
= btnum
;
419 if (btnum
== XFS_BTNUM_INO
) {
420 cur
->bc_nlevels
= be32_to_cpu(agi
->agi_level
);
421 cur
->bc_ops
= &xfs_inobt_ops
;
422 cur
->bc_statoff
= XFS_STATS_CALC_INDEX(xs_ibt_2
);
424 cur
->bc_nlevels
= be32_to_cpu(agi
->agi_free_level
);
425 cur
->bc_ops
= &xfs_finobt_ops
;
426 cur
->bc_statoff
= XFS_STATS_CALC_INDEX(xs_fibt_2
);
429 cur
->bc_blocklog
= mp
->m_sb
.sb_blocklog
;
431 if (xfs_sb_version_hascrc(&mp
->m_sb
))
432 cur
->bc_flags
|= XFS_BTREE_CRC_BLOCKS
;
434 cur
->bc_private
.a
.agbp
= agbp
;
435 cur
->bc_private
.a
.agno
= agno
;
441 * Calculate number of records in an inobt btree block.
445 struct xfs_mount
*mp
,
449 blocklen
-= XFS_INOBT_BLOCK_LEN(mp
);
452 return blocklen
/ sizeof(xfs_inobt_rec_t
);
453 return blocklen
/ (sizeof(xfs_inobt_key_t
) + sizeof(xfs_inobt_ptr_t
));
457 * Convert the inode record holemask to an inode allocation bitmap. The inode
458 * allocation bitmap is inode granularity and specifies whether an inode is
459 * physically allocated on disk (not whether the inode is considered allocated
460 * or free by the fs).
462 * A bit value of 1 means the inode is allocated, a value of 0 means it is free.
465 xfs_inobt_irec_to_allocmask(
466 struct xfs_inobt_rec_incore
*rec
)
474 * The holemask has 16-bits for a 64 inode record. Therefore each
475 * holemask bit represents multiple inodes. Create a mask of bits to set
476 * in the allocmask for each holemask bit.
478 inodespbit
= (1 << XFS_INODES_PER_HOLEMASK_BIT
) - 1;
481 * Allocated inodes are represented by 0 bits in holemask. Invert the 0
482 * bits to 1 and convert to a uint so we can use xfs_next_bit(). Mask
483 * anything beyond the 16 holemask bits since this casts to a larger
486 allocbitmap
= ~rec
->ir_holemask
& ((1 << XFS_INOBT_HOLEMASK_BITS
) - 1);
489 * allocbitmap is the inverted holemask so every set bit represents
490 * allocated inodes. To expand from 16-bit holemask granularity to
491 * 64-bit (e.g., bit-per-inode), set inodespbit bits in the target
492 * bitmap for every holemask bit.
494 nextbit
= xfs_next_bit(&allocbitmap
, 1, 0);
495 while (nextbit
!= -1) {
496 ASSERT(nextbit
< (sizeof(rec
->ir_holemask
) * NBBY
));
498 bitmap
|= (inodespbit
<<
499 (nextbit
* XFS_INODES_PER_HOLEMASK_BIT
));
501 nextbit
= xfs_next_bit(&allocbitmap
, 1, nextbit
+ 1);
507 #if defined(DEBUG) || defined(XFS_WARN)
509 * Verify that an in-core inode record has a valid inode count.
512 xfs_inobt_rec_check_count(
513 struct xfs_mount
*mp
,
514 struct xfs_inobt_rec_incore
*rec
)
521 wordsz
= sizeof(allocbmap
) / sizeof(unsigned int);
522 allocbmap
= xfs_inobt_irec_to_allocmask(rec
);
524 nextbit
= xfs_next_bit((uint
*) &allocbmap
, wordsz
, nextbit
);
525 while (nextbit
!= -1) {
527 nextbit
= xfs_next_bit((uint
*) &allocbmap
, wordsz
,
531 if (inocount
!= rec
->ir_count
)
532 return -EFSCORRUPTED
;
540 struct xfs_mount
*mp
,
543 xfs_agblock_t agblocks
= xfs_ag_block_count(mp
, agno
);
545 /* Bail out if we're uninitialized, which can happen in mkfs. */
546 if (mp
->m_inobt_mxr
[0] == 0)
549 return xfs_btree_calc_size(mp
->m_inobt_mnr
,
550 (uint64_t)agblocks
* mp
->m_sb
.sb_inopblock
/
551 XFS_INODES_PER_CHUNK
);
555 xfs_inobt_count_blocks(
556 struct xfs_mount
*mp
,
557 struct xfs_trans
*tp
,
560 xfs_extlen_t
*tree_blocks
)
562 struct xfs_buf
*agbp
;
563 struct xfs_btree_cur
*cur
;
566 error
= xfs_ialloc_read_agi(mp
, tp
, agno
, &agbp
);
570 cur
= xfs_inobt_init_cursor(mp
, tp
, agbp
, agno
, btnum
);
571 error
= xfs_btree_count_blocks(cur
, tree_blocks
);
572 xfs_btree_del_cursor(cur
, error
);
573 xfs_trans_brelse(tp
, agbp
);
579 * Figure out how many blocks to reserve and how many are used by this btree.
582 xfs_finobt_calc_reserves(
583 struct xfs_mount
*mp
,
584 struct xfs_trans
*tp
,
589 xfs_extlen_t tree_len
= 0;
592 if (!xfs_sb_version_hasfinobt(&mp
->m_sb
))
595 error
= xfs_inobt_count_blocks(mp
, tp
, agno
, XFS_BTNUM_FINO
, &tree_len
);
599 *ask
+= xfs_inobt_max_size(mp
, agno
);
604 /* Calculate the inobt btree size for some records. */
606 xfs_iallocbt_calc_size(
607 struct xfs_mount
*mp
,
608 unsigned long long len
)
610 return xfs_btree_calc_size(mp
->m_inobt_mnr
, len
);