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
18 #ifndef __XFS_BTREE_H__
19 #define __XFS_BTREE_H__
27 extern kmem_zone_t *xfs_btree_cur_zone;
30 * This nonsense is to make -wlint happy.
32 #define XFS_LOOKUP_EQ ((xfs_lookup_t)XFS_LOOKUP_EQi)
33 #define XFS_LOOKUP_LE ((xfs_lookup_t)XFS_LOOKUP_LEi)
34 #define XFS_LOOKUP_GE ((xfs_lookup_t)XFS_LOOKUP_GEi)
36 #define XFS_BTNUM_BNO ((xfs_btnum_t)XFS_BTNUM_BNOi)
37 #define XFS_BTNUM_CNT ((xfs_btnum_t)XFS_BTNUM_CNTi)
38 #define XFS_BTNUM_BMAP ((xfs_btnum_t)XFS_BTNUM_BMAPi)
39 #define XFS_BTNUM_INO ((xfs_btnum_t)XFS_BTNUM_INOi)
42 * Short form header: space allocation btrees.
44 typedef struct xfs_btree_sblock {
45 __be32 bb_magic; /* magic number for block type */
46 __be16 bb_level; /* 0 is a leaf */
47 __be16 bb_numrecs; /* current # of data records */
48 __be32 bb_leftsib; /* left sibling block or NULLAGBLOCK */
49 __be32 bb_rightsib; /* right sibling block or NULLAGBLOCK */
53 * Long form header: bmap btrees.
55 typedef struct xfs_btree_lblock {
56 __be32 bb_magic; /* magic number for block type */
57 __be16 bb_level; /* 0 is a leaf */
58 __be16 bb_numrecs; /* current # of data records */
59 __be64 bb_leftsib; /* left sibling block or NULLDFSBNO */
60 __be64 bb_rightsib; /* right sibling block or NULLDFSBNO */
64 * Combined header and structure, used by common code.
66 typedef struct xfs_btree_block {
67 __be32 bb_magic; /* magic number for block type */
68 __be16 bb_level; /* 0 is a leaf */
69 __be16 bb_numrecs; /* current # of data records */
74 } s; /* short form pointers */
78 } l; /* long form pointers */
83 * Generic key, ptr and record wrapper structures.
85 * These are disk format structures, and are converted where necessary
86 * by the btree specific code that needs to interpret them.
89 __be32 s; /* short form ptr */
90 __be64 l; /* long form ptr */
95 xfs_bmdr_key_t bmbr; /* bmbt root block */
96 xfs_alloc_key_t alloc;
97 xfs_inobt_key_t inobt;
100 union xfs_btree_rec {
102 xfs_bmdr_rec_t bmbr; /* bmbt root block */
103 xfs_alloc_rec_t alloc;
104 xfs_inobt_rec_t inobt;
108 * For logging record fields.
110 #define XFS_BB_MAGIC 0x01
111 #define XFS_BB_LEVEL 0x02
112 #define XFS_BB_NUMRECS 0x04
113 #define XFS_BB_LEFTSIB 0x08
114 #define XFS_BB_RIGHTSIB 0x10
115 #define XFS_BB_NUM_BITS 5
116 #define XFS_BB_ALL_BITS ((1 << XFS_BB_NUM_BITS) - 1)
119 * Boolean to select which form of xfs_btree_block_t.bb_u to use.
121 #define XFS_BTREE_LONG_PTRS(btnum) ((btnum) == XFS_BTNUM_BMAP)
124 * Magic numbers for btree blocks.
126 extern const __uint32_t xfs_magics[];
129 * Maximum and minimum records in a btree block.
130 * Given block size, type prefix, and leaf flag (0 or 1).
131 * The divisor below is equivalent to lf ? (e1) : (e2) but that produces
134 #define XFS_BTREE_BLOCK_MAXRECS(bsz,t,lf) \
135 ((int)(((bsz) - (uint)sizeof(t ## _block_t)) / \
136 (((lf) * (uint)sizeof(t ## _rec_t)) + \
138 ((uint)sizeof(t ## _key_t) + (uint)sizeof(t ## _ptr_t))))))
139 #define XFS_BTREE_BLOCK_MINRECS(bsz,t,lf) \
140 (XFS_BTREE_BLOCK_MAXRECS(bsz,t,lf) / 2)
143 * Record, key, and pointer address calculation macros.
144 * Given block size, type prefix, block pointer, and index of requested entry
145 * (first entry numbered 1).
147 #define XFS_BTREE_REC_ADDR(t,bb,i) \
148 ((t ## _rec_t *)((char *)(bb) + sizeof(t ## _block_t) + \
149 ((i) - 1) * sizeof(t ## _rec_t)))
150 #define XFS_BTREE_KEY_ADDR(t,bb,i) \
151 ((t ## _key_t *)((char *)(bb) + sizeof(t ## _block_t) + \
152 ((i) - 1) * sizeof(t ## _key_t)))
153 #define XFS_BTREE_PTR_ADDR(t,bb,i,mxr) \
154 ((t ## _ptr_t *)((char *)(bb) + sizeof(t ## _block_t) + \
155 (mxr) * sizeof(t ## _key_t) + ((i) - 1) * sizeof(t ## _ptr_t)))
157 #define XFS_BTREE_MAXLEVELS 8 /* max of all btrees */
159 struct xfs_btree_ops {
160 /* cursor operations */
161 struct xfs_btree_cur *(*dup_cursor)(struct xfs_btree_cur *);
165 * Btree cursor structure.
166 * This collects all information needed by the btree code in one place.
168 typedef struct xfs_btree_cur
170 struct xfs_trans *bc_tp; /* transaction we're in, if any */
171 struct xfs_mount *bc_mp; /* file system mount struct */
172 const struct xfs_btree_ops *bc_ops;
173 uint bc_flags; /* btree features - below */
175 xfs_alloc_rec_incore_t a;
177 xfs_inobt_rec_incore_t i;
178 } bc_rec; /* current insert/search record value */
179 struct xfs_buf *bc_bufs[XFS_BTREE_MAXLEVELS]; /* buf ptr per level */
180 int bc_ptrs[XFS_BTREE_MAXLEVELS]; /* key/record # */
181 __uint8_t bc_ra[XFS_BTREE_MAXLEVELS]; /* readahead bits */
182 #define XFS_BTCUR_LEFTRA 1 /* left sibling has been read-ahead */
183 #define XFS_BTCUR_RIGHTRA 2 /* right sibling has been read-ahead */
184 __uint8_t bc_nlevels; /* number of levels in the tree */
185 __uint8_t bc_blocklog; /* log2(blocksize) of btree blocks */
186 xfs_btnum_t bc_btnum; /* identifies which btree type */
188 struct { /* needed for BNO, CNT, INO */
189 struct xfs_buf *agbp; /* agf/agi buffer pointer */
190 xfs_agnumber_t agno; /* ag number */
192 struct { /* needed for BMAP */
193 struct xfs_inode *ip; /* pointer to our inode */
194 struct xfs_bmap_free *flist; /* list to free after */
195 xfs_fsblock_t firstblock; /* 1st blk allocated */
196 int allocated; /* count of alloced */
197 short forksize; /* fork's inode space */
198 char whichfork; /* data or attr fork */
199 char flags; /* flags */
200 #define XFS_BTCUR_BPRV_WASDEL 1 /* was delayed */
202 } bc_private; /* per-btree type data */
206 #define XFS_BTREE_ROOT_IN_INODE (1<<1) /* root may be variable size */
209 #define XFS_BTREE_NOERROR 0
210 #define XFS_BTREE_ERROR 1
213 * Convert from buffer to btree block header.
215 #define XFS_BUF_TO_BLOCK(bp) ((xfs_btree_block_t *)XFS_BUF_PTR(bp))
216 #define XFS_BUF_TO_LBLOCK(bp) ((xfs_btree_lblock_t *)XFS_BUF_PTR(bp))
217 #define XFS_BUF_TO_SBLOCK(bp) ((xfs_btree_sblock_t *)XFS_BUF_PTR(bp))
224 * Debug routine: check that block header is ok.
227 xfs_btree_check_block(
228 xfs_btree_cur_t *cur, /* btree cursor */
229 xfs_btree_block_t *block, /* generic btree block pointer */
230 int level, /* level of the btree block */
231 struct xfs_buf *bp); /* buffer containing block, if any */
234 * Debug routine: check that keys are in the right order.
238 xfs_btnum_t btnum, /* btree identifier */
239 void *ak1, /* pointer to left (lower) key */
240 void *ak2); /* pointer to right (higher) key */
243 * Debug routine: check that records are in the right order.
247 xfs_btnum_t btnum, /* btree identifier */
248 void *ar1, /* pointer to left (lower) record */
249 void *ar2); /* pointer to right (higher) record */
251 #define xfs_btree_check_block(a,b,c,d)
252 #define xfs_btree_check_key(a,b,c)
253 #define xfs_btree_check_rec(a,b,c)
257 * Checking routine: check that long form block header is ok.
259 int /* error (0 or EFSCORRUPTED) */
260 xfs_btree_check_lblock(
261 xfs_btree_cur_t *cur, /* btree cursor */
262 xfs_btree_lblock_t *block, /* btree long form block pointer */
263 int level, /* level of the btree block */
264 struct xfs_buf *bp); /* buffer containing block, if any */
267 * Checking routine: check that (long) pointer is ok.
269 int /* error (0 or EFSCORRUPTED) */
270 xfs_btree_check_lptr(
271 xfs_btree_cur_t *cur, /* btree cursor */
272 xfs_dfsbno_t ptr, /* btree block disk address */
273 int level); /* btree block level */
275 #define xfs_btree_check_lptr_disk(cur, ptr, level) \
276 xfs_btree_check_lptr(cur, be64_to_cpu(ptr), level)
279 * Checking routine: check that short form block header is ok.
281 int /* error (0 or EFSCORRUPTED) */
282 xfs_btree_check_sblock(
283 xfs_btree_cur_t *cur, /* btree cursor */
284 xfs_btree_sblock_t *block, /* btree short form block pointer */
285 int level, /* level of the btree block */
286 struct xfs_buf *bp); /* buffer containing block */
289 * Checking routine: check that (short) pointer is ok.
291 int /* error (0 or EFSCORRUPTED) */
292 xfs_btree_check_sptr(
293 xfs_btree_cur_t *cur, /* btree cursor */
294 xfs_agblock_t ptr, /* btree block disk address */
295 int level); /* btree block level */
298 * Delete the btree cursor.
301 xfs_btree_del_cursor(
302 xfs_btree_cur_t *cur, /* btree cursor */
303 int error); /* del because of error */
306 * Duplicate the btree cursor.
307 * Allocate a new one, copy the record, re-get the buffers.
310 xfs_btree_dup_cursor(
311 xfs_btree_cur_t *cur, /* input cursor */
312 xfs_btree_cur_t **ncur);/* output cursor */
315 * Change the cursor to point to the first record in the current block
316 * at the given level. Other levels are unaffected.
318 int /* success=1, failure=0 */
320 xfs_btree_cur_t *cur, /* btree cursor */
321 int level); /* level to change */
324 * Get a buffer for the block, return it with no data read.
325 * Long-form addressing.
327 struct xfs_buf * /* buffer for fsbno */
329 struct xfs_mount *mp, /* file system mount point */
330 struct xfs_trans *tp, /* transaction pointer */
331 xfs_fsblock_t fsbno, /* file system block number */
332 uint lock); /* lock flags for get_buf */
335 * Get a buffer for the block, return it with no data read.
336 * Short-form addressing.
338 struct xfs_buf * /* buffer for agno/agbno */
340 struct xfs_mount *mp, /* file system mount point */
341 struct xfs_trans *tp, /* transaction pointer */
342 xfs_agnumber_t agno, /* allocation group number */
343 xfs_agblock_t agbno, /* allocation group block number */
344 uint lock); /* lock flags for get_buf */
347 * Check for the cursor referring to the last block at the given level.
349 int /* 1=is last block, 0=not last block */
350 xfs_btree_islastblock(
351 xfs_btree_cur_t *cur, /* btree cursor */
352 int level); /* level to check */
355 * Change the cursor to point to the last record in the current block
356 * at the given level. Other levels are unaffected.
358 int /* success=1, failure=0 */
360 xfs_btree_cur_t *cur, /* btree cursor */
361 int level); /* level to change */
364 * Compute first and last byte offsets for the fields given.
365 * Interprets the offsets table, which contains struct field offsets.
369 __int64_t fields, /* bitmask of fields */
370 const short *offsets,/* table of field offsets */
371 int nbits, /* number of bits to inspect */
372 int *first, /* output: first byte offset */
373 int *last); /* output: last byte offset */
376 * Get a buffer for the block, return it read in.
377 * Long-form addressing.
381 struct xfs_mount *mp, /* file system mount point */
382 struct xfs_trans *tp, /* transaction pointer */
383 xfs_fsblock_t fsbno, /* file system block number */
384 uint lock, /* lock flags for read_buf */
385 struct xfs_buf **bpp, /* buffer for fsbno */
386 int refval);/* ref count value for buffer */
389 * Get a buffer for the block, return it read in.
390 * Short-form addressing.
394 struct xfs_mount *mp, /* file system mount point */
395 struct xfs_trans *tp, /* transaction pointer */
396 xfs_agnumber_t agno, /* allocation group number */
397 xfs_agblock_t agbno, /* allocation group block number */
398 uint lock, /* lock flags for read_buf */
399 struct xfs_buf **bpp, /* buffer for agno/agbno */
400 int refval);/* ref count value for buffer */
403 * Read-ahead the block, don't wait for it, don't return a buffer.
404 * Long-form addressing.
407 xfs_btree_reada_bufl(
408 struct xfs_mount *mp, /* file system mount point */
409 xfs_fsblock_t fsbno, /* file system block number */
410 xfs_extlen_t count); /* count of filesystem blocks */
413 * Read-ahead the block, don't wait for it, don't return a buffer.
414 * Short-form addressing.
417 xfs_btree_reada_bufs(
418 struct xfs_mount *mp, /* file system mount point */
419 xfs_agnumber_t agno, /* allocation group number */
420 xfs_agblock_t agbno, /* allocation group block number */
421 xfs_extlen_t count); /* count of filesystem blocks */
424 * Read-ahead btree blocks, at the given level.
425 * Bits in lr are set from XFS_BTCUR_{LEFT,RIGHT}RA.
427 int /* readahead block count */
428 xfs_btree_readahead_core(
429 xfs_btree_cur_t *cur, /* btree cursor */
430 int lev, /* level in btree */
431 int lr); /* left/right bits */
433 static inline int /* readahead block count */
435 xfs_btree_cur_t *cur, /* btree cursor */
436 int lev, /* level in btree */
437 int lr) /* left/right bits */
439 if ((cur->bc_ra[lev] | lr) == cur->bc_ra[lev])
442 return xfs_btree_readahead_core(cur, lev, lr);
447 * Set the buffer for level "lev" in the cursor to bp, releasing
448 * any previous buffer.
452 xfs_btree_cur_t *cur, /* btree cursor */
453 int lev, /* level in btree */
454 struct xfs_buf *bp); /* new buffer to set */
456 #endif /* __KERNEL__ */
460 * Min and max functions for extlen, agblock, fileoff, and filblks types.
462 #define XFS_EXTLEN_MIN(a,b) min_t(xfs_extlen_t, (a), (b))
463 #define XFS_EXTLEN_MAX(a,b) max_t(xfs_extlen_t, (a), (b))
464 #define XFS_AGBLOCK_MIN(a,b) min_t(xfs_agblock_t, (a), (b))
465 #define XFS_AGBLOCK_MAX(a,b) max_t(xfs_agblock_t, (a), (b))
466 #define XFS_FILEOFF_MIN(a,b) min_t(xfs_fileoff_t, (a), (b))
467 #define XFS_FILEOFF_MAX(a,b) max_t(xfs_fileoff_t, (a), (b))
468 #define XFS_FILBLKS_MIN(a,b) min_t(xfs_filblks_t, (a), (b))
469 #define XFS_FILBLKS_MAX(a,b) max_t(xfs_filblks_t, (a), (b))
471 #define XFS_FSB_SANITY_CHECK(mp,fsb) \
472 (XFS_FSB_TO_AGNO(mp, fsb) < mp->m_sb.sb_agcount && \
473 XFS_FSB_TO_AGBNO(mp, fsb) < mp->m_sb.sb_agblocks)
475 #endif /* __XFS_BTREE_H__ */