xref: /openbsd/sys/ufs/ufs/dirhash.h (revision da325f0b)
1 /* $OpenBSD: dirhash.h,v 1.9 2024/10/14 02:20:01 jsg Exp $	*/
2 /*
3  * Copyright (c) 2001 Ian Dowse.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD: src/sys/ufs/ufs/dirhash.h,v 1.4 2003/01/01 18:48:59 schweikh Exp $
27  */
28 
29 #ifndef _UFS_UFS_DIRHASH_H_
30 #define _UFS_UFS_DIRHASH_H_
31 
32 #include <sys/rwlock.h>
33 
34 /*
35  * For fast operations on large directories, we maintain a hash
36  * that maps the file name to the offset of the directory entry within
37  * the directory file.
38  *
39  * The hashing uses a dumb spillover to the next free slot on
40  * collisions, so we must keep the utilisation low to avoid
41  * long linear searches. Deleted entries that are not the last
42  * in a chain must be marked DIRHASH_DEL.
43  *
44  * We also maintain information about free space in each block
45  * to speed up creations.
46  */
47 #define DIRHASH_EMPTY	(-1)	/* entry unused */
48 #define DIRHASH_DEL	(-2)	/* deleted entry; may be part of chain */
49 
50 #define DIRALIGN	4
51 #define DH_NFSTATS	(DIRECTSIZ(MAXNAMLEN + 1) / DIRALIGN)
52 				 /* max DIRALIGN words in a directory entry */
53 
54 /*
55  * Dirhash uses a score mechanism to achieve a hybrid between a
56  * least-recently-used and a least-often-used algorithm for entry
57  * recycling. The score is incremented when a directory is used, and
58  * decremented when the directory is a candidate for recycling. When
59  * the score reaches zero, the hash is recycled. Hashes are linked
60  * together on a TAILQ list, and hashes with higher scores filter
61  * towards the tail (most recently used) end of the list.
62  *
63  * New hash entries are given an initial score of DH_SCOREINIT and are
64  * placed at the most-recently-used end of the list. This helps a lot
65  * in the worst-case case scenario where every directory access is
66  * to a directory that is not hashed (i.e. the working set of hash
67  * candidates is much larger than the configured memory limit). In this
68  * case it limits the number of hash builds to 1/DH_SCOREINIT of the
69  * number of accesses.
70  */
71 #define DH_SCOREINIT	8	/* initial dh_score when dirhash built */
72 #define DH_SCOREMAX	64	/* max dh_score value */
73 
74 /*
75  * The main hash table has 2 levels. It is an array of pointers to
76  * blocks of DH_NBLKOFF offsets.
77  */
78 #define DH_BLKOFFSHIFT	8
79 #define DH_NBLKOFF	(1 << DH_BLKOFFSHIFT)
80 #define DH_BLKOFFMASK	(DH_NBLKOFF - 1)
81 
82 #define DH_ENTRY(dh, slot) \
83     ((dh)->dh_hash[(slot) >> DH_BLKOFFSHIFT][(slot) & DH_BLKOFFMASK])
84 
85 struct dirhash {
86 	struct rwlock dh_mtx;	/* protects all fields except dh_list */
87 	doff_t	**dh_hash;	/* the hash array (2-level) */
88 	int	dh_narrays;	/* number of entries in dh_hash */
89 	int	dh_hlen;	/* total slots in the 2-level hash array */
90 	int	dh_hused;	/* entries in use */
91 
92 	/* Free space statistics. XXX assumes DIRBLKSIZ is 512. */
93 	u_int8_t *dh_blkfree;	/* free DIRALIGN words in each dir block */
94 	int	dh_nblk;	/* size of dh_blkfree array */
95 	int	dh_dirblks;	/* number of DIRBLKSIZ blocks in dir */
96 	int	dh_firstfree[DH_NFSTATS + 1]; /* first blk with N words free */
97 
98 	int	dh_seqopt;	/* sequential access optimisation enabled */
99 	doff_t	dh_seqoff;	/* sequential access optimisation offset */
100 
101 	int	dh_score;	/* access count for this dirhash */
102 
103 	int	dh_onlist;	/* true if on the ufsdirhash_list chain */
104 
105 	/* Protected by ufsdirhash_mtx. */
106 	TAILQ_ENTRY(dirhash) dh_list;	/* chain of all dirhashes */
107 };
108 
109 extern	int ufs_mindirhashsize;
110 extern	int ufs_dirhashmaxmem;
111 extern	int ufs_dirhashmem;
112 
113 /*
114  * Dirhash functions.
115  */
116 void	ufsdirhash_init(void);
117 void	ufsdirhash_uninit(void);
118 int	ufsdirhash_build(struct inode *);
119 doff_t	ufsdirhash_findfree(struct inode *, int, int *);
120 doff_t	ufsdirhash_enduseful(struct inode *);
121 int	ufsdirhash_lookup(struct inode *, char *, int, doff_t *, struct buf **,
122 	    doff_t *);
123 void	ufsdirhash_newblk(struct inode *, doff_t);
124 void	ufsdirhash_add(struct inode *, struct direct *, doff_t);
125 void	ufsdirhash_remove(struct inode *, struct direct *, doff_t);
126 void	ufsdirhash_move(struct inode *, struct direct *, doff_t, doff_t);
127 void	ufsdirhash_dirtrunc(struct inode *, doff_t);
128 void	ufsdirhash_free(struct inode *);
129 
130 void	ufsdirhash_checkblock(struct inode *, char *, doff_t);
131 
132 #endif /* !_UFS_UFS_DIRHASH_H_ */
133