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"
26 #include "xfs_mount.h"
27 #include "xfs_btree.h"
28 #include "xfs_alloc_btree.h"
29 #include "xfs_alloc.h"
30 #include "xfs_extent_busy.h"
31 #include "xfs_error.h"
32 #include "xfs_trace.h"
33 #include "xfs_cksum.h"
34 #include "xfs_trans.h"
37 STATIC
struct xfs_btree_cur
*
38 xfs_allocbt_dup_cursor(
39 struct xfs_btree_cur
*cur
)
41 return xfs_allocbt_init_cursor(cur
->bc_mp
, cur
->bc_tp
,
42 cur
->bc_private
.a
.agbp
, cur
->bc_private
.a
.agno
,
48 struct xfs_btree_cur
*cur
,
49 union xfs_btree_ptr
*ptr
,
52 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
53 struct xfs_agf
*agf
= XFS_BUF_TO_AGF(agbp
);
54 xfs_agnumber_t seqno
= be32_to_cpu(agf
->agf_seqno
);
55 int btnum
= cur
->bc_btnum
;
56 struct xfs_perag
*pag
= xfs_perag_get(cur
->bc_mp
, seqno
);
60 agf
->agf_roots
[btnum
] = ptr
->s
;
61 be32_add_cpu(&agf
->agf_levels
[btnum
], inc
);
62 pag
->pagf_levels
[btnum
] += inc
;
65 xfs_alloc_log_agf(cur
->bc_tp
, agbp
, XFS_AGF_ROOTS
| XFS_AGF_LEVELS
);
69 xfs_allocbt_alloc_block(
70 struct xfs_btree_cur
*cur
,
71 union xfs_btree_ptr
*start
,
72 union xfs_btree_ptr
*new,
79 XFS_BTREE_TRACE_CURSOR(cur
, XBT_ENTRY
);
81 /* Allocate the new block from the freelist. If we can't, give up. */
82 error
= xfs_alloc_get_freelist(cur
->bc_tp
, cur
->bc_private
.a
.agbp
,
85 XFS_BTREE_TRACE_CURSOR(cur
, XBT_ERROR
);
89 if (bno
== NULLAGBLOCK
) {
90 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
95 xfs_extent_busy_reuse(cur
->bc_mp
, cur
->bc_private
.a
.agno
, bno
, 1, false);
97 xfs_trans_agbtree_delta(cur
->bc_tp
, 1);
98 new->s
= cpu_to_be32(bno
);
100 XFS_BTREE_TRACE_CURSOR(cur
, XBT_EXIT
);
106 xfs_allocbt_free_block(
107 struct xfs_btree_cur
*cur
,
110 struct xfs_buf
*agbp
= cur
->bc_private
.a
.agbp
;
111 struct xfs_agf
*agf
= XFS_BUF_TO_AGF(agbp
);
115 bno
= xfs_daddr_to_agbno(cur
->bc_mp
, XFS_BUF_ADDR(bp
));
116 error
= xfs_alloc_put_freelist(cur
->bc_tp
, agbp
, NULL
, bno
, 1);
120 xfs_extent_busy_insert(cur
->bc_tp
, be32_to_cpu(agf
->agf_seqno
), bno
, 1,
121 XFS_EXTENT_BUSY_SKIP_DISCARD
);
122 xfs_trans_agbtree_delta(cur
->bc_tp
, -1);
124 xfs_trans_binval(cur
->bc_tp
, bp
);
129 * Update the longest extent in the AGF
132 xfs_allocbt_update_lastrec(
133 struct xfs_btree_cur
*cur
,
134 struct xfs_btree_block
*block
,
135 union xfs_btree_rec
*rec
,
139 struct xfs_agf
*agf
= XFS_BUF_TO_AGF(cur
->bc_private
.a
.agbp
);
140 xfs_agnumber_t seqno
= be32_to_cpu(agf
->agf_seqno
);
141 struct xfs_perag
*pag
;
145 ASSERT(cur
->bc_btnum
== XFS_BTNUM_CNT
);
150 * If this is the last leaf block and it's the last record,
151 * then update the size of the longest extent in the AG.
153 if (ptr
!= xfs_btree_get_numrecs(block
))
155 len
= rec
->alloc
.ar_blockcount
;
158 if (be32_to_cpu(rec
->alloc
.ar_blockcount
) <=
159 be32_to_cpu(agf
->agf_longest
))
161 len
= rec
->alloc
.ar_blockcount
;
164 numrecs
= xfs_btree_get_numrecs(block
);
167 ASSERT(ptr
== numrecs
+ 1);
170 xfs_alloc_rec_t
*rrp
;
172 rrp
= XFS_ALLOC_REC_ADDR(cur
->bc_mp
, block
, numrecs
);
173 len
= rrp
->ar_blockcount
;
184 agf
->agf_longest
= len
;
185 pag
= xfs_perag_get(cur
->bc_mp
, seqno
);
186 pag
->pagf_longest
= be32_to_cpu(len
);
188 xfs_alloc_log_agf(cur
->bc_tp
, cur
->bc_private
.a
.agbp
, XFS_AGF_LONGEST
);
192 xfs_allocbt_get_minrecs(
193 struct xfs_btree_cur
*cur
,
196 return cur
->bc_mp
->m_alloc_mnr
[level
!= 0];
200 xfs_allocbt_get_maxrecs(
201 struct xfs_btree_cur
*cur
,
204 return cur
->bc_mp
->m_alloc_mxr
[level
!= 0];
208 xfs_allocbt_init_key_from_rec(
209 union xfs_btree_key
*key
,
210 union xfs_btree_rec
*rec
)
212 ASSERT(rec
->alloc
.ar_startblock
!= 0);
214 key
->alloc
.ar_startblock
= rec
->alloc
.ar_startblock
;
215 key
->alloc
.ar_blockcount
= rec
->alloc
.ar_blockcount
;
219 xfs_allocbt_init_rec_from_key(
220 union xfs_btree_key
*key
,
221 union xfs_btree_rec
*rec
)
223 ASSERT(key
->alloc
.ar_startblock
!= 0);
225 rec
->alloc
.ar_startblock
= key
->alloc
.ar_startblock
;
226 rec
->alloc
.ar_blockcount
= key
->alloc
.ar_blockcount
;
230 xfs_allocbt_init_rec_from_cur(
231 struct xfs_btree_cur
*cur
,
232 union xfs_btree_rec
*rec
)
234 ASSERT(cur
->bc_rec
.a
.ar_startblock
!= 0);
236 rec
->alloc
.ar_startblock
= cpu_to_be32(cur
->bc_rec
.a
.ar_startblock
);
237 rec
->alloc
.ar_blockcount
= cpu_to_be32(cur
->bc_rec
.a
.ar_blockcount
);
241 xfs_allocbt_init_ptr_from_cur(
242 struct xfs_btree_cur
*cur
,
243 union xfs_btree_ptr
*ptr
)
245 struct xfs_agf
*agf
= XFS_BUF_TO_AGF(cur
->bc_private
.a
.agbp
);
247 ASSERT(cur
->bc_private
.a
.agno
== be32_to_cpu(agf
->agf_seqno
));
248 ASSERT(agf
->agf_roots
[cur
->bc_btnum
] != 0);
250 ptr
->s
= agf
->agf_roots
[cur
->bc_btnum
];
254 xfs_allocbt_key_diff(
255 struct xfs_btree_cur
*cur
,
256 union xfs_btree_key
*key
)
258 xfs_alloc_rec_incore_t
*rec
= &cur
->bc_rec
.a
;
259 xfs_alloc_key_t
*kp
= &key
->alloc
;
262 if (cur
->bc_btnum
== XFS_BTNUM_BNO
) {
263 return (__int64_t
)be32_to_cpu(kp
->ar_startblock
) -
267 diff
= (__int64_t
)be32_to_cpu(kp
->ar_blockcount
) - rec
->ar_blockcount
;
271 return (__int64_t
)be32_to_cpu(kp
->ar_startblock
) - rec
->ar_startblock
;
278 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
279 struct xfs_btree_block
*block
= XFS_BUF_TO_BLOCK(bp
);
280 struct xfs_perag
*pag
= bp
->b_pag
;
284 * magic number and level verification
286 * During growfs operations, we can't verify the exact level or owner as
287 * the perag is not fully initialised and hence not attached to the
288 * buffer. In this case, check against the maximum tree depth.
290 * Similarly, during log recovery we will have a perag structure
291 * attached, but the agf information will not yet have been initialised
292 * from the on disk AGF. Again, we can only check against maximum limits
295 level
= be16_to_cpu(block
->bb_level
);
296 switch (block
->bb_magic
) {
297 case cpu_to_be32(XFS_ABTB_CRC_MAGIC
):
298 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
300 if (!uuid_equal(&block
->bb_u
.s
.bb_uuid
, &mp
->m_sb
.sb_uuid
))
302 if (block
->bb_u
.s
.bb_blkno
!= cpu_to_be64(bp
->b_bn
))
305 be32_to_cpu(block
->bb_u
.s
.bb_owner
) != pag
->pag_agno
)
308 case cpu_to_be32(XFS_ABTB_MAGIC
):
309 if (pag
&& pag
->pagf_init
) {
310 if (level
>= pag
->pagf_levels
[XFS_BTNUM_BNOi
])
312 } else if (level
>= mp
->m_ag_maxlevels
)
315 case cpu_to_be32(XFS_ABTC_CRC_MAGIC
):
316 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
318 if (!uuid_equal(&block
->bb_u
.s
.bb_uuid
, &mp
->m_sb
.sb_uuid
))
320 if (block
->bb_u
.s
.bb_blkno
!= cpu_to_be64(bp
->b_bn
))
323 be32_to_cpu(block
->bb_u
.s
.bb_owner
) != pag
->pag_agno
)
326 case cpu_to_be32(XFS_ABTC_MAGIC
):
327 if (pag
&& pag
->pagf_init
) {
328 if (level
>= pag
->pagf_levels
[XFS_BTNUM_CNTi
])
330 } else if (level
>= mp
->m_ag_maxlevels
)
337 /* numrecs verification */
338 if (be16_to_cpu(block
->bb_numrecs
) > mp
->m_alloc_mxr
[level
!= 0])
341 /* sibling pointer verification */
342 if (!block
->bb_u
.s
.bb_leftsib
||
343 (be32_to_cpu(block
->bb_u
.s
.bb_leftsib
) >= mp
->m_sb
.sb_agblocks
&&
344 block
->bb_u
.s
.bb_leftsib
!= cpu_to_be32(NULLAGBLOCK
)))
346 if (!block
->bb_u
.s
.bb_rightsib
||
347 (be32_to_cpu(block
->bb_u
.s
.bb_rightsib
) >= mp
->m_sb
.sb_agblocks
&&
348 block
->bb_u
.s
.bb_rightsib
!= cpu_to_be32(NULLAGBLOCK
)))
355 xfs_allocbt_read_verify(
358 if (!(xfs_btree_sblock_verify_crc(bp
) &&
359 xfs_allocbt_verify(bp
))) {
360 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
361 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
,
362 bp
->b_target
->bt_mount
, bp
->b_addr
);
363 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
368 xfs_allocbt_write_verify(
371 if (!xfs_allocbt_verify(bp
)) {
372 trace_xfs_btree_corrupt(bp
, _RET_IP_
);
373 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
,
374 bp
->b_target
->bt_mount
, bp
->b_addr
);
375 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
377 xfs_btree_sblock_calc_crc(bp
);
381 const struct xfs_buf_ops xfs_allocbt_buf_ops
= {
382 .verify_read
= xfs_allocbt_read_verify
,
383 .verify_write
= xfs_allocbt_write_verify
,
387 #if defined(DEBUG) || defined(XFS_WARN)
389 xfs_allocbt_keys_inorder(
390 struct xfs_btree_cur
*cur
,
391 union xfs_btree_key
*k1
,
392 union xfs_btree_key
*k2
)
394 if (cur
->bc_btnum
== XFS_BTNUM_BNO
) {
395 return be32_to_cpu(k1
->alloc
.ar_startblock
) <
396 be32_to_cpu(k2
->alloc
.ar_startblock
);
398 return be32_to_cpu(k1
->alloc
.ar_blockcount
) <
399 be32_to_cpu(k2
->alloc
.ar_blockcount
) ||
400 (k1
->alloc
.ar_blockcount
== k2
->alloc
.ar_blockcount
&&
401 be32_to_cpu(k1
->alloc
.ar_startblock
) <
402 be32_to_cpu(k2
->alloc
.ar_startblock
));
407 xfs_allocbt_recs_inorder(
408 struct xfs_btree_cur
*cur
,
409 union xfs_btree_rec
*r1
,
410 union xfs_btree_rec
*r2
)
412 if (cur
->bc_btnum
== XFS_BTNUM_BNO
) {
413 return be32_to_cpu(r1
->alloc
.ar_startblock
) +
414 be32_to_cpu(r1
->alloc
.ar_blockcount
) <=
415 be32_to_cpu(r2
->alloc
.ar_startblock
);
417 return be32_to_cpu(r1
->alloc
.ar_blockcount
) <
418 be32_to_cpu(r2
->alloc
.ar_blockcount
) ||
419 (r1
->alloc
.ar_blockcount
== r2
->alloc
.ar_blockcount
&&
420 be32_to_cpu(r1
->alloc
.ar_startblock
) <
421 be32_to_cpu(r2
->alloc
.ar_startblock
));
426 static const struct xfs_btree_ops xfs_allocbt_ops
= {
427 .rec_len
= sizeof(xfs_alloc_rec_t
),
428 .key_len
= sizeof(xfs_alloc_key_t
),
430 .dup_cursor
= xfs_allocbt_dup_cursor
,
431 .set_root
= xfs_allocbt_set_root
,
432 .alloc_block
= xfs_allocbt_alloc_block
,
433 .free_block
= xfs_allocbt_free_block
,
434 .update_lastrec
= xfs_allocbt_update_lastrec
,
435 .get_minrecs
= xfs_allocbt_get_minrecs
,
436 .get_maxrecs
= xfs_allocbt_get_maxrecs
,
437 .init_key_from_rec
= xfs_allocbt_init_key_from_rec
,
438 .init_rec_from_key
= xfs_allocbt_init_rec_from_key
,
439 .init_rec_from_cur
= xfs_allocbt_init_rec_from_cur
,
440 .init_ptr_from_cur
= xfs_allocbt_init_ptr_from_cur
,
441 .key_diff
= xfs_allocbt_key_diff
,
442 .buf_ops
= &xfs_allocbt_buf_ops
,
443 #if defined(DEBUG) || defined(XFS_WARN)
444 .keys_inorder
= xfs_allocbt_keys_inorder
,
445 .recs_inorder
= xfs_allocbt_recs_inorder
,
450 * Allocate a new allocation btree cursor.
452 struct xfs_btree_cur
* /* new alloc btree cursor */
453 xfs_allocbt_init_cursor(
454 struct xfs_mount
*mp
, /* file system mount point */
455 struct xfs_trans
*tp
, /* transaction pointer */
456 struct xfs_buf
*agbp
, /* buffer for agf structure */
457 xfs_agnumber_t agno
, /* allocation group number */
458 xfs_btnum_t btnum
) /* btree identifier */
460 struct xfs_agf
*agf
= XFS_BUF_TO_AGF(agbp
);
461 struct xfs_btree_cur
*cur
;
463 ASSERT(btnum
== XFS_BTNUM_BNO
|| btnum
== XFS_BTNUM_CNT
);
465 cur
= kmem_zone_zalloc(xfs_btree_cur_zone
, KM_SLEEP
);
469 cur
->bc_btnum
= btnum
;
470 cur
->bc_blocklog
= mp
->m_sb
.sb_blocklog
;
471 cur
->bc_ops
= &xfs_allocbt_ops
;
473 if (btnum
== XFS_BTNUM_CNT
) {
474 cur
->bc_nlevels
= be32_to_cpu(agf
->agf_levels
[XFS_BTNUM_CNT
]);
475 cur
->bc_flags
= XFS_BTREE_LASTREC_UPDATE
;
477 cur
->bc_nlevels
= be32_to_cpu(agf
->agf_levels
[XFS_BTNUM_BNO
]);
480 cur
->bc_private
.a
.agbp
= agbp
;
481 cur
->bc_private
.a
.agno
= agno
;
483 if (xfs_sb_version_hascrc(&mp
->m_sb
))
484 cur
->bc_flags
|= XFS_BTREE_CRC_BLOCKS
;
490 * Calculate number of records in an alloc btree block.
494 struct xfs_mount
*mp
,
498 blocklen
-= XFS_ALLOC_BLOCK_LEN(mp
);
501 return blocklen
/ sizeof(xfs_alloc_rec_t
);
502 return blocklen
/ (sizeof(xfs_alloc_key_t
) + sizeof(xfs_alloc_ptr_t
));