xref: /openbmc/linux/fs/xfs/libxfs/xfs_btree.h (revision d774a589)
1 /*
2  * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
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.
8  *
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.
13  *
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
17  */
18 #ifndef __XFS_BTREE_H__
19 #define	__XFS_BTREE_H__
20 
21 struct xfs_buf;
22 struct xfs_defer_ops;
23 struct xfs_inode;
24 struct xfs_mount;
25 struct xfs_trans;
26 
27 extern kmem_zone_t	*xfs_btree_cur_zone;
28 
29 /*
30  * Generic key, ptr and record wrapper structures.
31  *
32  * These are disk format structures, and are converted where necessary
33  * by the btree specific code that needs to interpret them.
34  */
35 union xfs_btree_ptr {
36 	__be32			s;	/* short form ptr */
37 	__be64			l;	/* long form ptr */
38 };
39 
40 /*
41  * The in-core btree key.  Overlapping btrees actually store two keys
42  * per pointer, so we reserve enough memory to hold both.  The __*bigkey
43  * items should never be accessed directly.
44  */
45 union xfs_btree_key {
46 	struct xfs_bmbt_key		bmbt;
47 	xfs_bmdr_key_t			bmbr;	/* bmbt root block */
48 	xfs_alloc_key_t			alloc;
49 	struct xfs_inobt_key		inobt;
50 	struct xfs_rmap_key		rmap;
51 	struct xfs_rmap_key		__rmap_bigkey[2];
52 	struct xfs_refcount_key		refc;
53 };
54 
55 union xfs_btree_rec {
56 	struct xfs_bmbt_rec		bmbt;
57 	xfs_bmdr_rec_t			bmbr;	/* bmbt root block */
58 	struct xfs_alloc_rec		alloc;
59 	struct xfs_inobt_rec		inobt;
60 	struct xfs_rmap_rec		rmap;
61 	struct xfs_refcount_rec		refc;
62 };
63 
64 /*
65  * This nonsense is to make -wlint happy.
66  */
67 #define	XFS_LOOKUP_EQ	((xfs_lookup_t)XFS_LOOKUP_EQi)
68 #define	XFS_LOOKUP_LE	((xfs_lookup_t)XFS_LOOKUP_LEi)
69 #define	XFS_LOOKUP_GE	((xfs_lookup_t)XFS_LOOKUP_GEi)
70 
71 #define	XFS_BTNUM_BNO	((xfs_btnum_t)XFS_BTNUM_BNOi)
72 #define	XFS_BTNUM_CNT	((xfs_btnum_t)XFS_BTNUM_CNTi)
73 #define	XFS_BTNUM_BMAP	((xfs_btnum_t)XFS_BTNUM_BMAPi)
74 #define	XFS_BTNUM_INO	((xfs_btnum_t)XFS_BTNUM_INOi)
75 #define	XFS_BTNUM_FINO	((xfs_btnum_t)XFS_BTNUM_FINOi)
76 #define	XFS_BTNUM_RMAP	((xfs_btnum_t)XFS_BTNUM_RMAPi)
77 #define	XFS_BTNUM_REFC	((xfs_btnum_t)XFS_BTNUM_REFCi)
78 
79 /*
80  * For logging record fields.
81  */
82 #define	XFS_BB_MAGIC		(1 << 0)
83 #define	XFS_BB_LEVEL		(1 << 1)
84 #define	XFS_BB_NUMRECS		(1 << 2)
85 #define	XFS_BB_LEFTSIB		(1 << 3)
86 #define	XFS_BB_RIGHTSIB		(1 << 4)
87 #define	XFS_BB_BLKNO		(1 << 5)
88 #define	XFS_BB_LSN		(1 << 6)
89 #define	XFS_BB_UUID		(1 << 7)
90 #define	XFS_BB_OWNER		(1 << 8)
91 #define	XFS_BB_NUM_BITS		5
92 #define	XFS_BB_ALL_BITS		((1 << XFS_BB_NUM_BITS) - 1)
93 #define	XFS_BB_NUM_BITS_CRC	9
94 #define	XFS_BB_ALL_BITS_CRC	((1 << XFS_BB_NUM_BITS_CRC) - 1)
95 
96 /*
97  * Generic stats interface
98  */
99 #define XFS_BTREE_STATS_INC(cur, stat)	\
100 	XFS_STATS_INC_OFF((cur)->bc_mp, (cur)->bc_statoff + __XBTS_ ## stat)
101 #define XFS_BTREE_STATS_ADD(cur, stat, val)	\
102 	XFS_STATS_ADD_OFF((cur)->bc_mp, (cur)->bc_statoff + __XBTS_ ## stat, val)
103 
104 #define	XFS_BTREE_MAXLEVELS	9	/* max of all btrees */
105 
106 struct xfs_btree_ops {
107 	/* size of the key and record structures */
108 	size_t	key_len;
109 	size_t	rec_len;
110 
111 	/* cursor operations */
112 	struct xfs_btree_cur *(*dup_cursor)(struct xfs_btree_cur *);
113 	void	(*update_cursor)(struct xfs_btree_cur *src,
114 				 struct xfs_btree_cur *dst);
115 
116 	/* update btree root pointer */
117 	void	(*set_root)(struct xfs_btree_cur *cur,
118 			    union xfs_btree_ptr *nptr, int level_change);
119 
120 	/* block allocation / freeing */
121 	int	(*alloc_block)(struct xfs_btree_cur *cur,
122 			       union xfs_btree_ptr *start_bno,
123 			       union xfs_btree_ptr *new_bno,
124 			       int *stat);
125 	int	(*free_block)(struct xfs_btree_cur *cur, struct xfs_buf *bp);
126 
127 	/* update last record information */
128 	void	(*update_lastrec)(struct xfs_btree_cur *cur,
129 				  struct xfs_btree_block *block,
130 				  union xfs_btree_rec *rec,
131 				  int ptr, int reason);
132 
133 	/* records in block/level */
134 	int	(*get_minrecs)(struct xfs_btree_cur *cur, int level);
135 	int	(*get_maxrecs)(struct xfs_btree_cur *cur, int level);
136 
137 	/* records on disk.  Matter for the root in inode case. */
138 	int	(*get_dmaxrecs)(struct xfs_btree_cur *cur, int level);
139 
140 	/* init values of btree structures */
141 	void	(*init_key_from_rec)(union xfs_btree_key *key,
142 				     union xfs_btree_rec *rec);
143 	void	(*init_rec_from_cur)(struct xfs_btree_cur *cur,
144 				     union xfs_btree_rec *rec);
145 	void	(*init_ptr_from_cur)(struct xfs_btree_cur *cur,
146 				     union xfs_btree_ptr *ptr);
147 	void	(*init_high_key_from_rec)(union xfs_btree_key *key,
148 					  union xfs_btree_rec *rec);
149 
150 	/* difference between key value and cursor value */
151 	__int64_t (*key_diff)(struct xfs_btree_cur *cur,
152 			      union xfs_btree_key *key);
153 
154 	/*
155 	 * Difference between key2 and key1 -- positive if key1 > key2,
156 	 * negative if key1 < key2, and zero if equal.
157 	 */
158 	__int64_t (*diff_two_keys)(struct xfs_btree_cur *cur,
159 				   union xfs_btree_key *key1,
160 				   union xfs_btree_key *key2);
161 
162 	const struct xfs_buf_ops	*buf_ops;
163 
164 #if defined(DEBUG) || defined(XFS_WARN)
165 	/* check that k1 is lower than k2 */
166 	int	(*keys_inorder)(struct xfs_btree_cur *cur,
167 				union xfs_btree_key *k1,
168 				union xfs_btree_key *k2);
169 
170 	/* check that r1 is lower than r2 */
171 	int	(*recs_inorder)(struct xfs_btree_cur *cur,
172 				union xfs_btree_rec *r1,
173 				union xfs_btree_rec *r2);
174 #endif
175 };
176 
177 /*
178  * Reasons for the update_lastrec method to be called.
179  */
180 #define LASTREC_UPDATE	0
181 #define LASTREC_INSREC	1
182 #define LASTREC_DELREC	2
183 
184 
185 union xfs_btree_irec {
186 	struct xfs_alloc_rec_incore	a;
187 	struct xfs_bmbt_irec		b;
188 	struct xfs_inobt_rec_incore	i;
189 	struct xfs_rmap_irec		r;
190 	struct xfs_refcount_irec	rc;
191 };
192 
193 /* Per-AG btree private information. */
194 union xfs_btree_cur_private {
195 	struct {
196 		unsigned long	nr_ops;		/* # record updates */
197 		int		shape_changes;	/* # of extent splits */
198 	} refc;
199 };
200 
201 /*
202  * Btree cursor structure.
203  * This collects all information needed by the btree code in one place.
204  */
205 typedef struct xfs_btree_cur
206 {
207 	struct xfs_trans	*bc_tp;	/* transaction we're in, if any */
208 	struct xfs_mount	*bc_mp;	/* file system mount struct */
209 	const struct xfs_btree_ops *bc_ops;
210 	uint			bc_flags; /* btree features - below */
211 	union xfs_btree_irec	bc_rec;	/* current insert/search record value */
212 	struct xfs_buf	*bc_bufs[XFS_BTREE_MAXLEVELS];	/* buf ptr per level */
213 	int		bc_ptrs[XFS_BTREE_MAXLEVELS];	/* key/record # */
214 	__uint8_t	bc_ra[XFS_BTREE_MAXLEVELS];	/* readahead bits */
215 #define	XFS_BTCUR_LEFTRA	1	/* left sibling has been read-ahead */
216 #define	XFS_BTCUR_RIGHTRA	2	/* right sibling has been read-ahead */
217 	__uint8_t	bc_nlevels;	/* number of levels in the tree */
218 	__uint8_t	bc_blocklog;	/* log2(blocksize) of btree blocks */
219 	xfs_btnum_t	bc_btnum;	/* identifies which btree type */
220 	int		bc_statoff;	/* offset of btre stats array */
221 	union {
222 		struct {			/* needed for BNO, CNT, INO */
223 			struct xfs_buf	*agbp;	/* agf/agi buffer pointer */
224 			struct xfs_defer_ops *dfops;	/* deferred updates */
225 			xfs_agnumber_t	agno;	/* ag number */
226 			union xfs_btree_cur_private	priv;
227 		} a;
228 		struct {			/* needed for BMAP */
229 			struct xfs_inode *ip;	/* pointer to our inode */
230 			struct xfs_defer_ops *dfops;	/* deferred updates */
231 			xfs_fsblock_t	firstblock;	/* 1st blk allocated */
232 			int		allocated;	/* count of alloced */
233 			short		forksize;	/* fork's inode space */
234 			char		whichfork;	/* data or attr fork */
235 			char		flags;		/* flags */
236 #define	XFS_BTCUR_BPRV_WASDEL	1			/* was delayed */
237 		} b;
238 	}		bc_private;	/* per-btree type data */
239 } xfs_btree_cur_t;
240 
241 /* cursor flags */
242 #define XFS_BTREE_LONG_PTRS		(1<<0)	/* pointers are 64bits long */
243 #define XFS_BTREE_ROOT_IN_INODE		(1<<1)	/* root may be variable size */
244 #define XFS_BTREE_LASTREC_UPDATE	(1<<2)	/* track last rec externally */
245 #define XFS_BTREE_CRC_BLOCKS		(1<<3)	/* uses extended btree blocks */
246 #define XFS_BTREE_OVERLAPPING		(1<<4)	/* overlapping intervals */
247 
248 
249 #define	XFS_BTREE_NOERROR	0
250 #define	XFS_BTREE_ERROR		1
251 
252 /*
253  * Convert from buffer to btree block header.
254  */
255 #define	XFS_BUF_TO_BLOCK(bp)	((struct xfs_btree_block *)((bp)->b_addr))
256 
257 
258 /*
259  * Check that block header is ok.
260  */
261 int
262 xfs_btree_check_block(
263 	struct xfs_btree_cur	*cur,	/* btree cursor */
264 	struct xfs_btree_block	*block,	/* generic btree block pointer */
265 	int			level,	/* level of the btree block */
266 	struct xfs_buf		*bp);	/* buffer containing block, if any */
267 
268 /*
269  * Check that (long) pointer is ok.
270  */
271 int					/* error (0 or EFSCORRUPTED) */
272 xfs_btree_check_lptr(
273 	struct xfs_btree_cur	*cur,	/* btree cursor */
274 	xfs_fsblock_t		ptr,	/* btree block disk address */
275 	int			level);	/* btree block level */
276 
277 /*
278  * Delete the btree cursor.
279  */
280 void
281 xfs_btree_del_cursor(
282 	xfs_btree_cur_t		*cur,	/* btree cursor */
283 	int			error);	/* del because of error */
284 
285 /*
286  * Duplicate the btree cursor.
287  * Allocate a new one, copy the record, re-get the buffers.
288  */
289 int					/* error */
290 xfs_btree_dup_cursor(
291 	xfs_btree_cur_t		*cur,	/* input cursor */
292 	xfs_btree_cur_t		**ncur);/* output cursor */
293 
294 /*
295  * Get a buffer for the block, return it with no data read.
296  * Long-form addressing.
297  */
298 struct xfs_buf *				/* buffer for fsbno */
299 xfs_btree_get_bufl(
300 	struct xfs_mount	*mp,	/* file system mount point */
301 	struct xfs_trans	*tp,	/* transaction pointer */
302 	xfs_fsblock_t		fsbno,	/* file system block number */
303 	uint			lock);	/* lock flags for get_buf */
304 
305 /*
306  * Get a buffer for the block, return it with no data read.
307  * Short-form addressing.
308  */
309 struct xfs_buf *				/* buffer for agno/agbno */
310 xfs_btree_get_bufs(
311 	struct xfs_mount	*mp,	/* file system mount point */
312 	struct xfs_trans	*tp,	/* transaction pointer */
313 	xfs_agnumber_t		agno,	/* allocation group number */
314 	xfs_agblock_t		agbno,	/* allocation group block number */
315 	uint			lock);	/* lock flags for get_buf */
316 
317 /*
318  * Check for the cursor referring to the last block at the given level.
319  */
320 int					/* 1=is last block, 0=not last block */
321 xfs_btree_islastblock(
322 	xfs_btree_cur_t		*cur,	/* btree cursor */
323 	int			level);	/* level to check */
324 
325 /*
326  * Compute first and last byte offsets for the fields given.
327  * Interprets the offsets table, which contains struct field offsets.
328  */
329 void
330 xfs_btree_offsets(
331 	__int64_t		fields,	/* bitmask of fields */
332 	const short		*offsets,/* table of field offsets */
333 	int			nbits,	/* number of bits to inspect */
334 	int			*first,	/* output: first byte offset */
335 	int			*last);	/* output: last byte offset */
336 
337 /*
338  * Get a buffer for the block, return it read in.
339  * Long-form addressing.
340  */
341 int					/* error */
342 xfs_btree_read_bufl(
343 	struct xfs_mount	*mp,	/* file system mount point */
344 	struct xfs_trans	*tp,	/* transaction pointer */
345 	xfs_fsblock_t		fsbno,	/* file system block number */
346 	uint			lock,	/* lock flags for read_buf */
347 	struct xfs_buf		**bpp,	/* buffer for fsbno */
348 	int			refval,	/* ref count value for buffer */
349 	const struct xfs_buf_ops *ops);
350 
351 /*
352  * Read-ahead the block, don't wait for it, don't return a buffer.
353  * Long-form addressing.
354  */
355 void					/* error */
356 xfs_btree_reada_bufl(
357 	struct xfs_mount	*mp,	/* file system mount point */
358 	xfs_fsblock_t		fsbno,	/* file system block number */
359 	xfs_extlen_t		count,	/* count of filesystem blocks */
360 	const struct xfs_buf_ops *ops);
361 
362 /*
363  * Read-ahead the block, don't wait for it, don't return a buffer.
364  * Short-form addressing.
365  */
366 void					/* error */
367 xfs_btree_reada_bufs(
368 	struct xfs_mount	*mp,	/* file system mount point */
369 	xfs_agnumber_t		agno,	/* allocation group number */
370 	xfs_agblock_t		agbno,	/* allocation group block number */
371 	xfs_extlen_t		count,	/* count of filesystem blocks */
372 	const struct xfs_buf_ops *ops);
373 
374 /*
375  * Initialise a new btree block header
376  */
377 void
378 xfs_btree_init_block(
379 	struct xfs_mount *mp,
380 	struct xfs_buf	*bp,
381 	__u32		magic,
382 	__u16		level,
383 	__u16		numrecs,
384 	__u64		owner,
385 	unsigned int	flags);
386 
387 void
388 xfs_btree_init_block_int(
389 	struct xfs_mount	*mp,
390 	struct xfs_btree_block	*buf,
391 	xfs_daddr_t		blkno,
392 	__u32			magic,
393 	__u16			level,
394 	__u16			numrecs,
395 	__u64			owner,
396 	unsigned int		flags);
397 
398 /*
399  * Common btree core entry points.
400  */
401 int xfs_btree_increment(struct xfs_btree_cur *, int, int *);
402 int xfs_btree_decrement(struct xfs_btree_cur *, int, int *);
403 int xfs_btree_lookup(struct xfs_btree_cur *, xfs_lookup_t, int *);
404 int xfs_btree_update(struct xfs_btree_cur *, union xfs_btree_rec *);
405 int xfs_btree_new_iroot(struct xfs_btree_cur *, int *, int *);
406 int xfs_btree_insert(struct xfs_btree_cur *, int *);
407 int xfs_btree_delete(struct xfs_btree_cur *, int *);
408 int xfs_btree_get_rec(struct xfs_btree_cur *, union xfs_btree_rec **, int *);
409 int xfs_btree_change_owner(struct xfs_btree_cur *cur, __uint64_t new_owner,
410 			   struct list_head *buffer_list);
411 
412 /*
413  * btree block CRC helpers
414  */
415 void xfs_btree_lblock_calc_crc(struct xfs_buf *);
416 bool xfs_btree_lblock_verify_crc(struct xfs_buf *);
417 void xfs_btree_sblock_calc_crc(struct xfs_buf *);
418 bool xfs_btree_sblock_verify_crc(struct xfs_buf *);
419 
420 /*
421  * Internal btree helpers also used by xfs_bmap.c.
422  */
423 void xfs_btree_log_block(struct xfs_btree_cur *, struct xfs_buf *, int);
424 void xfs_btree_log_recs(struct xfs_btree_cur *, struct xfs_buf *, int, int);
425 
426 /*
427  * Helpers.
428  */
429 static inline int xfs_btree_get_numrecs(struct xfs_btree_block *block)
430 {
431 	return be16_to_cpu(block->bb_numrecs);
432 }
433 
434 static inline void xfs_btree_set_numrecs(struct xfs_btree_block *block,
435 		__uint16_t numrecs)
436 {
437 	block->bb_numrecs = cpu_to_be16(numrecs);
438 }
439 
440 static inline int xfs_btree_get_level(struct xfs_btree_block *block)
441 {
442 	return be16_to_cpu(block->bb_level);
443 }
444 
445 
446 /*
447  * Min and max functions for extlen, agblock, fileoff, and filblks types.
448  */
449 #define	XFS_EXTLEN_MIN(a,b)	min_t(xfs_extlen_t, (a), (b))
450 #define	XFS_EXTLEN_MAX(a,b)	max_t(xfs_extlen_t, (a), (b))
451 #define	XFS_AGBLOCK_MIN(a,b)	min_t(xfs_agblock_t, (a), (b))
452 #define	XFS_AGBLOCK_MAX(a,b)	max_t(xfs_agblock_t, (a), (b))
453 #define	XFS_FILEOFF_MIN(a,b)	min_t(xfs_fileoff_t, (a), (b))
454 #define	XFS_FILEOFF_MAX(a,b)	max_t(xfs_fileoff_t, (a), (b))
455 #define	XFS_FILBLKS_MIN(a,b)	min_t(xfs_filblks_t, (a), (b))
456 #define	XFS_FILBLKS_MAX(a,b)	max_t(xfs_filblks_t, (a), (b))
457 
458 #define	XFS_FSB_SANITY_CHECK(mp,fsb)	\
459 	(XFS_FSB_TO_AGNO(mp, fsb) < mp->m_sb.sb_agcount && \
460 		XFS_FSB_TO_AGBNO(mp, fsb) < mp->m_sb.sb_agblocks)
461 
462 /*
463  * Trace hooks.  Currently not implemented as they need to be ported
464  * over to the generic tracing functionality, which is some effort.
465  *
466  * i,j = integer (32 bit)
467  * b = btree block buffer (xfs_buf_t)
468  * p = btree ptr
469  * r = btree record
470  * k = btree key
471  */
472 #define	XFS_BTREE_TRACE_ARGBI(c, b, i)
473 #define	XFS_BTREE_TRACE_ARGBII(c, b, i, j)
474 #define	XFS_BTREE_TRACE_ARGI(c, i)
475 #define	XFS_BTREE_TRACE_ARGIPK(c, i, p, s)
476 #define	XFS_BTREE_TRACE_ARGIPR(c, i, p, r)
477 #define	XFS_BTREE_TRACE_ARGIK(c, i, k)
478 #define XFS_BTREE_TRACE_ARGR(c, r)
479 #define	XFS_BTREE_TRACE_CURSOR(c, t)
480 
481 bool xfs_btree_sblock_v5hdr_verify(struct xfs_buf *bp);
482 bool xfs_btree_sblock_verify(struct xfs_buf *bp, unsigned int max_recs);
483 uint xfs_btree_compute_maxlevels(struct xfs_mount *mp, uint *limits,
484 				 unsigned long len);
485 xfs_extlen_t xfs_btree_calc_size(struct xfs_mount *mp, uint *limits,
486 		unsigned long long len);
487 
488 /* return codes */
489 #define XFS_BTREE_QUERY_RANGE_CONTINUE	0	/* keep iterating */
490 #define XFS_BTREE_QUERY_RANGE_ABORT	1	/* stop iterating */
491 typedef int (*xfs_btree_query_range_fn)(struct xfs_btree_cur *cur,
492 		union xfs_btree_rec *rec, void *priv);
493 
494 int xfs_btree_query_range(struct xfs_btree_cur *cur,
495 		union xfs_btree_irec *low_rec, union xfs_btree_irec *high_rec,
496 		xfs_btree_query_range_fn fn, void *priv);
497 
498 typedef int (*xfs_btree_visit_blocks_fn)(struct xfs_btree_cur *cur, int level,
499 		void *data);
500 int xfs_btree_visit_blocks(struct xfs_btree_cur *cur,
501 		xfs_btree_visit_blocks_fn fn, void *data);
502 
503 int xfs_btree_count_blocks(struct xfs_btree_cur *cur, xfs_extlen_t *blocks);
504 
505 #endif	/* __XFS_BTREE_H__ */
506