1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
6 #ifndef __XFS_BTREE_H__
7 #define __XFS_BTREE_H__
16 extern kmem_zone_t *xfs_btree_cur_zone;
19 * Generic key, ptr and record wrapper structures.
21 * These are disk format structures, and are converted where necessary
22 * by the btree specific code that needs to interpret them.
25 __be32 s; /* short form ptr */
26 __be64 l; /* long form ptr */
30 * The in-core btree key. Overlapping btrees actually store two keys
31 * per pointer, so we reserve enough memory to hold both. The __*bigkey
32 * items should never be accessed directly.
35 struct xfs_bmbt_key bmbt;
36 xfs_bmdr_key_t bmbr; /* bmbt root block */
37 xfs_alloc_key_t alloc;
38 struct xfs_inobt_key inobt;
39 struct xfs_rmap_key rmap;
40 struct xfs_rmap_key __rmap_bigkey[2];
41 struct xfs_refcount_key refc;
45 struct xfs_bmbt_rec bmbt;
46 xfs_bmdr_rec_t bmbr; /* bmbt root block */
47 struct xfs_alloc_rec alloc;
48 struct xfs_inobt_rec inobt;
49 struct xfs_rmap_rec rmap;
50 struct xfs_refcount_rec refc;
54 * This nonsense is to make -wlint happy.
56 #define XFS_LOOKUP_EQ ((xfs_lookup_t)XFS_LOOKUP_EQi)
57 #define XFS_LOOKUP_LE ((xfs_lookup_t)XFS_LOOKUP_LEi)
58 #define XFS_LOOKUP_GE ((xfs_lookup_t)XFS_LOOKUP_GEi)
60 #define XFS_BTNUM_BNO ((xfs_btnum_t)XFS_BTNUM_BNOi)
61 #define XFS_BTNUM_CNT ((xfs_btnum_t)XFS_BTNUM_CNTi)
62 #define XFS_BTNUM_BMAP ((xfs_btnum_t)XFS_BTNUM_BMAPi)
63 #define XFS_BTNUM_INO ((xfs_btnum_t)XFS_BTNUM_INOi)
64 #define XFS_BTNUM_FINO ((xfs_btnum_t)XFS_BTNUM_FINOi)
65 #define XFS_BTNUM_RMAP ((xfs_btnum_t)XFS_BTNUM_RMAPi)
66 #define XFS_BTNUM_REFC ((xfs_btnum_t)XFS_BTNUM_REFCi)
68 uint32_t xfs_btree_magic(int crc, xfs_btnum_t btnum);
71 * For logging record fields.
73 #define XFS_BB_MAGIC (1 << 0)
74 #define XFS_BB_LEVEL (1 << 1)
75 #define XFS_BB_NUMRECS (1 << 2)
76 #define XFS_BB_LEFTSIB (1 << 3)
77 #define XFS_BB_RIGHTSIB (1 << 4)
78 #define XFS_BB_BLKNO (1 << 5)
79 #define XFS_BB_LSN (1 << 6)
80 #define XFS_BB_UUID (1 << 7)
81 #define XFS_BB_OWNER (1 << 8)
82 #define XFS_BB_NUM_BITS 5
83 #define XFS_BB_ALL_BITS ((1 << XFS_BB_NUM_BITS) - 1)
84 #define XFS_BB_NUM_BITS_CRC 9
85 #define XFS_BB_ALL_BITS_CRC ((1 << XFS_BB_NUM_BITS_CRC) - 1)
88 * Generic stats interface
90 #define XFS_BTREE_STATS_INC(cur, stat) \
91 XFS_STATS_INC_OFF((cur)->bc_mp, (cur)->bc_statoff + __XBTS_ ## stat)
92 #define XFS_BTREE_STATS_ADD(cur, stat, val) \
93 XFS_STATS_ADD_OFF((cur)->bc_mp, (cur)->bc_statoff + __XBTS_ ## stat, val)
95 #define XFS_BTREE_MAXLEVELS 9 /* max of all btrees */
97 struct xfs_btree_ops {
98 /* size of the key and record structures */
102 /* cursor operations */
103 struct xfs_btree_cur *(*dup_cursor)(struct xfs_btree_cur *);
104 void (*update_cursor)(struct xfs_btree_cur *src,
105 struct xfs_btree_cur *dst);
107 /* update btree root pointer */
108 void (*set_root)(struct xfs_btree_cur *cur,
109 union xfs_btree_ptr *nptr, int level_change);
111 /* block allocation / freeing */
112 int (*alloc_block)(struct xfs_btree_cur *cur,
113 union xfs_btree_ptr *start_bno,
114 union xfs_btree_ptr *new_bno,
116 int (*free_block)(struct xfs_btree_cur *cur, struct xfs_buf *bp);
118 /* update last record information */
119 void (*update_lastrec)(struct xfs_btree_cur *cur,
120 struct xfs_btree_block *block,
121 union xfs_btree_rec *rec,
122 int ptr, int reason);
124 /* records in block/level */
125 int (*get_minrecs)(struct xfs_btree_cur *cur, int level);
126 int (*get_maxrecs)(struct xfs_btree_cur *cur, int level);
128 /* records on disk. Matter for the root in inode case. */
129 int (*get_dmaxrecs)(struct xfs_btree_cur *cur, int level);
131 /* init values of btree structures */
132 void (*init_key_from_rec)(union xfs_btree_key *key,
133 union xfs_btree_rec *rec);
134 void (*init_rec_from_cur)(struct xfs_btree_cur *cur,
135 union xfs_btree_rec *rec);
136 void (*init_ptr_from_cur)(struct xfs_btree_cur *cur,
137 union xfs_btree_ptr *ptr);
138 void (*init_high_key_from_rec)(union xfs_btree_key *key,
139 union xfs_btree_rec *rec);
141 /* difference between key value and cursor value */
142 int64_t (*key_diff)(struct xfs_btree_cur *cur,
143 union xfs_btree_key *key);
146 * Difference between key2 and key1 -- positive if key1 > key2,
147 * negative if key1 < key2, and zero if equal.
149 int64_t (*diff_two_keys)(struct xfs_btree_cur *cur,
150 union xfs_btree_key *key1,
151 union xfs_btree_key *key2);
153 const struct xfs_buf_ops *buf_ops;
155 /* check that k1 is lower than k2 */
156 int (*keys_inorder)(struct xfs_btree_cur *cur,
157 union xfs_btree_key *k1,
158 union xfs_btree_key *k2);
160 /* check that r1 is lower than r2 */
161 int (*recs_inorder)(struct xfs_btree_cur *cur,
162 union xfs_btree_rec *r1,
163 union xfs_btree_rec *r2);
167 * Reasons for the update_lastrec method to be called.
169 #define LASTREC_UPDATE 0
170 #define LASTREC_INSREC 1
171 #define LASTREC_DELREC 2
174 union xfs_btree_irec {
175 struct xfs_alloc_rec_incore a;
176 struct xfs_bmbt_irec b;
177 struct xfs_inobt_rec_incore i;
178 struct xfs_rmap_irec r;
179 struct xfs_refcount_irec rc;
182 /* Per-AG btree information. */
183 struct xfs_btree_cur_ag {
184 struct xfs_perag *pag;
186 struct xfs_buf *agbp;
187 struct xbtree_afakeroot *afake; /* for staging cursor */
191 unsigned long nr_ops; /* # record updates */
192 int shape_changes; /* # of extent splits */
195 bool active; /* allocation cursor state */
200 /* Btree-in-inode cursor information */
201 struct xfs_btree_cur_ino {
202 struct xfs_inode *ip;
203 struct xbtree_ifakeroot *ifake; /* for staging cursor */
208 /* We are converting a delalloc reservation */
209 #define XFS_BTCUR_BMBT_WASDEL (1 << 0)
211 /* For extent swap, ignore owner check in verifier */
212 #define XFS_BTCUR_BMBT_INVALID_OWNER (1 << 1)
216 * Btree cursor structure.
217 * This collects all information needed by the btree code in one place.
219 typedef struct xfs_btree_cur
221 struct xfs_trans *bc_tp; /* transaction we're in, if any */
222 struct xfs_mount *bc_mp; /* file system mount struct */
223 const struct xfs_btree_ops *bc_ops;
224 uint bc_flags; /* btree features - below */
225 union xfs_btree_irec bc_rec; /* current insert/search record value */
226 struct xfs_buf *bc_bufs[XFS_BTREE_MAXLEVELS]; /* buf ptr per level */
227 int bc_ptrs[XFS_BTREE_MAXLEVELS]; /* key/record # */
228 uint8_t bc_ra[XFS_BTREE_MAXLEVELS]; /* readahead bits */
229 #define XFS_BTCUR_LEFTRA 1 /* left sibling has been read-ahead */
230 #define XFS_BTCUR_RIGHTRA 2 /* right sibling has been read-ahead */
231 uint8_t bc_nlevels; /* number of levels in the tree */
232 uint8_t bc_blocklog; /* log2(blocksize) of btree blocks */
233 xfs_btnum_t bc_btnum; /* identifies which btree type */
234 int bc_statoff; /* offset of btre stats array */
237 * Short btree pointers need an agno to be able to turn the pointers
238 * into physical addresses for IO, so the btree cursor switches between
239 * bc_ino and bc_ag based on whether XFS_BTREE_LONG_PTRS is set for the
243 struct xfs_btree_cur_ag bc_ag;
244 struct xfs_btree_cur_ino bc_ino;
249 #define XFS_BTREE_LONG_PTRS (1<<0) /* pointers are 64bits long */
250 #define XFS_BTREE_ROOT_IN_INODE (1<<1) /* root may be variable size */
251 #define XFS_BTREE_LASTREC_UPDATE (1<<2) /* track last rec externally */
252 #define XFS_BTREE_CRC_BLOCKS (1<<3) /* uses extended btree blocks */
253 #define XFS_BTREE_OVERLAPPING (1<<4) /* overlapping intervals */
255 * The root of this btree is a fakeroot structure so that we can stage a btree
256 * rebuild without leaving it accessible via primary metadata. The ops struct
257 * is dynamically allocated and must be freed when the cursor is deleted.
259 #define XFS_BTREE_STAGING (1<<5)
262 #define XFS_BTREE_NOERROR 0
263 #define XFS_BTREE_ERROR 1
266 * Convert from buffer to btree block header.
268 #define XFS_BUF_TO_BLOCK(bp) ((struct xfs_btree_block *)((bp)->b_addr))
271 * Internal long and short btree block checks. They return NULL if the
272 * block is ok or the address of the failed check otherwise.
274 xfs_failaddr_t __xfs_btree_check_lblock(struct xfs_btree_cur *cur,
275 struct xfs_btree_block *block, int level, struct xfs_buf *bp);
276 xfs_failaddr_t __xfs_btree_check_sblock(struct xfs_btree_cur *cur,
277 struct xfs_btree_block *block, int level, struct xfs_buf *bp);
280 * Check that block header is ok.
283 xfs_btree_check_block(
284 struct xfs_btree_cur *cur, /* btree cursor */
285 struct xfs_btree_block *block, /* generic btree block pointer */
286 int level, /* level of the btree block */
287 struct xfs_buf *bp); /* buffer containing block, if any */
290 * Check that (long) pointer is ok.
292 bool /* error (0 or EFSCORRUPTED) */
293 xfs_btree_check_lptr(
294 struct xfs_btree_cur *cur, /* btree cursor */
295 xfs_fsblock_t fsbno, /* btree block disk address */
296 int level); /* btree block level */
299 * Check that (short) pointer is ok.
301 bool /* error (0 or EFSCORRUPTED) */
302 xfs_btree_check_sptr(
303 struct xfs_btree_cur *cur, /* btree cursor */
304 xfs_agblock_t agbno, /* btree block disk address */
305 int level); /* btree block level */
308 * Delete the btree cursor.
311 xfs_btree_del_cursor(
312 xfs_btree_cur_t *cur, /* btree cursor */
313 int error); /* del because of error */
316 * Duplicate the btree cursor.
317 * Allocate a new one, copy the record, re-get the buffers.
320 xfs_btree_dup_cursor(
321 xfs_btree_cur_t *cur, /* input cursor */
322 xfs_btree_cur_t **ncur);/* output cursor */
325 * Compute first and last byte offsets for the fields given.
326 * Interprets the offsets table, which contains struct field offsets.
330 int64_t fields, /* bitmask of fields */
331 const short *offsets,/* table of field offsets */
332 int nbits, /* number of bits to inspect */
333 int *first, /* output: first byte offset */
334 int *last); /* output: last byte offset */
337 * Get a buffer for the block, return it read in.
338 * Long-form addressing.
342 struct xfs_mount *mp, /* file system mount point */
343 struct xfs_trans *tp, /* transaction pointer */
344 xfs_fsblock_t fsbno, /* file system block number */
345 struct xfs_buf **bpp, /* buffer for fsbno */
346 int refval, /* ref count value for buffer */
347 const struct xfs_buf_ops *ops);
350 * Read-ahead the block, don't wait for it, don't return a buffer.
351 * Long-form addressing.
354 xfs_btree_reada_bufl(
355 struct xfs_mount *mp, /* file system mount point */
356 xfs_fsblock_t fsbno, /* file system block number */
357 xfs_extlen_t count, /* count of filesystem blocks */
358 const struct xfs_buf_ops *ops);
361 * Read-ahead the block, don't wait for it, don't return a buffer.
362 * Short-form addressing.
365 xfs_btree_reada_bufs(
366 struct xfs_mount *mp, /* file system mount point */
367 xfs_agnumber_t agno, /* allocation group number */
368 xfs_agblock_t agbno, /* allocation group block number */
369 xfs_extlen_t count, /* count of filesystem blocks */
370 const struct xfs_buf_ops *ops);
373 * Initialise a new btree block header
376 xfs_btree_init_block(
377 struct xfs_mount *mp,
385 xfs_btree_init_block_int(
386 struct xfs_mount *mp,
387 struct xfs_btree_block *buf,
396 * Common btree core entry points.
398 int xfs_btree_increment(struct xfs_btree_cur *, int, int *);
399 int xfs_btree_decrement(struct xfs_btree_cur *, int, int *);
400 int xfs_btree_lookup(struct xfs_btree_cur *, xfs_lookup_t, int *);
401 int xfs_btree_update(struct xfs_btree_cur *, union xfs_btree_rec *);
402 int xfs_btree_new_iroot(struct xfs_btree_cur *, int *, int *);
403 int xfs_btree_insert(struct xfs_btree_cur *, int *);
404 int xfs_btree_delete(struct xfs_btree_cur *, int *);
405 int xfs_btree_get_rec(struct xfs_btree_cur *, union xfs_btree_rec **, int *);
406 int xfs_btree_change_owner(struct xfs_btree_cur *cur, uint64_t new_owner,
407 struct list_head *buffer_list);
410 * btree block CRC helpers
412 void xfs_btree_lblock_calc_crc(struct xfs_buf *);
413 bool xfs_btree_lblock_verify_crc(struct xfs_buf *);
414 void xfs_btree_sblock_calc_crc(struct xfs_buf *);
415 bool xfs_btree_sblock_verify_crc(struct xfs_buf *);
418 * Internal btree helpers also used by xfs_bmap.c.
420 void xfs_btree_log_block(struct xfs_btree_cur *, struct xfs_buf *, int);
421 void xfs_btree_log_recs(struct xfs_btree_cur *, struct xfs_buf *, int, int);
426 static inline int xfs_btree_get_numrecs(struct xfs_btree_block *block)
428 return be16_to_cpu(block->bb_numrecs);
431 static inline void xfs_btree_set_numrecs(struct xfs_btree_block *block,
434 block->bb_numrecs = cpu_to_be16(numrecs);
437 static inline int xfs_btree_get_level(struct xfs_btree_block *block)
439 return be16_to_cpu(block->bb_level);
444 * Min and max functions for extlen, agblock, fileoff, and filblks types.
446 #define XFS_EXTLEN_MIN(a,b) min_t(xfs_extlen_t, (a), (b))
447 #define XFS_EXTLEN_MAX(a,b) max_t(xfs_extlen_t, (a), (b))
448 #define XFS_AGBLOCK_MIN(a,b) min_t(xfs_agblock_t, (a), (b))
449 #define XFS_AGBLOCK_MAX(a,b) max_t(xfs_agblock_t, (a), (b))
450 #define XFS_FILEOFF_MIN(a,b) min_t(xfs_fileoff_t, (a), (b))
451 #define XFS_FILEOFF_MAX(a,b) max_t(xfs_fileoff_t, (a), (b))
452 #define XFS_FILBLKS_MIN(a,b) min_t(xfs_filblks_t, (a), (b))
453 #define XFS_FILBLKS_MAX(a,b) max_t(xfs_filblks_t, (a), (b))
455 xfs_failaddr_t xfs_btree_sblock_v5hdr_verify(struct xfs_buf *bp);
456 xfs_failaddr_t xfs_btree_sblock_verify(struct xfs_buf *bp,
457 unsigned int max_recs);
458 xfs_failaddr_t xfs_btree_lblock_v5hdr_verify(struct xfs_buf *bp,
460 xfs_failaddr_t xfs_btree_lblock_verify(struct xfs_buf *bp,
461 unsigned int max_recs);
463 uint xfs_btree_compute_maxlevels(uint *limits, unsigned long len);
464 unsigned long long xfs_btree_calc_size(uint *limits, unsigned long long len);
467 * Return codes for the query range iterator function are 0 to continue
468 * iterating, and non-zero to stop iterating. Any non-zero value will be
469 * passed up to the _query_range caller. The special value -ECANCELED can be
470 * used to stop iteration, because _query_range never generates that error
473 typedef int (*xfs_btree_query_range_fn)(struct xfs_btree_cur *cur,
474 union xfs_btree_rec *rec, void *priv);
476 int xfs_btree_query_range(struct xfs_btree_cur *cur,
477 union xfs_btree_irec *low_rec, union xfs_btree_irec *high_rec,
478 xfs_btree_query_range_fn fn, void *priv);
479 int xfs_btree_query_all(struct xfs_btree_cur *cur, xfs_btree_query_range_fn fn,
482 typedef int (*xfs_btree_visit_blocks_fn)(struct xfs_btree_cur *cur, int level,
484 /* Visit record blocks. */
485 #define XFS_BTREE_VISIT_RECORDS (1 << 0)
486 /* Visit leaf blocks. */
487 #define XFS_BTREE_VISIT_LEAVES (1 << 1)
488 /* Visit all blocks. */
489 #define XFS_BTREE_VISIT_ALL (XFS_BTREE_VISIT_RECORDS | \
490 XFS_BTREE_VISIT_LEAVES)
491 int xfs_btree_visit_blocks(struct xfs_btree_cur *cur,
492 xfs_btree_visit_blocks_fn fn, unsigned int flags, void *data);
494 int xfs_btree_count_blocks(struct xfs_btree_cur *cur, xfs_extlen_t *blocks);
496 union xfs_btree_rec *xfs_btree_rec_addr(struct xfs_btree_cur *cur, int n,
497 struct xfs_btree_block *block);
498 union xfs_btree_key *xfs_btree_key_addr(struct xfs_btree_cur *cur, int n,
499 struct xfs_btree_block *block);
500 union xfs_btree_key *xfs_btree_high_key_addr(struct xfs_btree_cur *cur, int n,
501 struct xfs_btree_block *block);
502 union xfs_btree_ptr *xfs_btree_ptr_addr(struct xfs_btree_cur *cur, int n,
503 struct xfs_btree_block *block);
504 int xfs_btree_lookup_get_block(struct xfs_btree_cur *cur, int level,
505 union xfs_btree_ptr *pp, struct xfs_btree_block **blkp);
506 struct xfs_btree_block *xfs_btree_get_block(struct xfs_btree_cur *cur,
507 int level, struct xfs_buf **bpp);
508 bool xfs_btree_ptr_is_null(struct xfs_btree_cur *cur, union xfs_btree_ptr *ptr);
509 int64_t xfs_btree_diff_two_ptrs(struct xfs_btree_cur *cur,
510 const union xfs_btree_ptr *a,
511 const union xfs_btree_ptr *b);
512 void xfs_btree_get_sibling(struct xfs_btree_cur *cur,
513 struct xfs_btree_block *block,
514 union xfs_btree_ptr *ptr, int lr);
515 void xfs_btree_get_keys(struct xfs_btree_cur *cur,
516 struct xfs_btree_block *block, union xfs_btree_key *key);
517 union xfs_btree_key *xfs_btree_high_key_from_key(struct xfs_btree_cur *cur,
518 union xfs_btree_key *key);
519 int xfs_btree_has_record(struct xfs_btree_cur *cur, union xfs_btree_irec *low,
520 union xfs_btree_irec *high, bool *exists);
521 bool xfs_btree_has_more_records(struct xfs_btree_cur *cur);
522 struct xfs_ifork *xfs_btree_ifork_ptr(struct xfs_btree_cur *cur);
524 /* Does this cursor point to the last block in the given level? */
526 xfs_btree_islastblock(
527 xfs_btree_cur_t *cur,
530 struct xfs_btree_block *block;
533 block = xfs_btree_get_block(cur, level, &bp);
534 ASSERT(block && xfs_btree_check_block(cur, block, level, bp) == 0);
536 if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
537 return block->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK);
538 return block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK);
541 void xfs_btree_set_ptr_null(struct xfs_btree_cur *cur,
542 union xfs_btree_ptr *ptr);
543 int xfs_btree_get_buf_block(struct xfs_btree_cur *cur, union xfs_btree_ptr *ptr,
544 struct xfs_btree_block **block, struct xfs_buf **bpp);
545 void xfs_btree_set_sibling(struct xfs_btree_cur *cur,
546 struct xfs_btree_block *block, union xfs_btree_ptr *ptr,
548 void xfs_btree_init_block_cur(struct xfs_btree_cur *cur,
549 struct xfs_buf *bp, int level, int numrecs);
550 void xfs_btree_copy_ptrs(struct xfs_btree_cur *cur,
551 union xfs_btree_ptr *dst_ptr,
552 const union xfs_btree_ptr *src_ptr, int numptrs);
553 void xfs_btree_copy_keys(struct xfs_btree_cur *cur,
554 union xfs_btree_key *dst_key, union xfs_btree_key *src_key,
557 #endif /* __XFS_BTREE_H__ */