xref: /freebsd/sys/kern/vfs_hash.c (revision a521f2116473fbd8c09db395518f060a27d02334)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2005 Poul-Henning Kamp
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/kernel.h>
36 #include <sys/malloc.h>
37 #include <sys/mount.h>
38 #include <sys/rwlock.h>
39 #include <sys/vnode.h>
40 
41 static MALLOC_DEFINE(M_VFS_HASH, "vfs_hash", "VFS hash table");
42 
43 static LIST_HEAD(vfs_hash_head, vnode)	*vfs_hash_tbl;
44 static LIST_HEAD(,vnode)		vfs_hash_side;
45 static u_long				vfs_hash_mask;
46 static struct rwlock __exclusive_cache_line vfs_hash_lock;
47 
48 static void
49 vfs_hashinit(void *dummy __unused)
50 {
51 
52 	vfs_hash_tbl = hashinit(desiredvnodes, M_VFS_HASH, &vfs_hash_mask);
53 	rw_init(&vfs_hash_lock, "vfs hash");
54 	LIST_INIT(&vfs_hash_side);
55 }
56 
57 /* Must be SI_ORDER_SECOND so desiredvnodes is available */
58 SYSINIT(vfs_hash, SI_SUB_VFS, SI_ORDER_SECOND, vfs_hashinit, NULL);
59 
60 u_int
61 vfs_hash_index(struct vnode *vp)
62 {
63 
64 	return (vp->v_hash + vp->v_mount->mnt_hashseed);
65 }
66 
67 static struct vfs_hash_head *
68 vfs_hash_bucket(const struct mount *mp, u_int hash)
69 {
70 
71 	return (&vfs_hash_tbl[(hash + mp->mnt_hashseed) & vfs_hash_mask]);
72 }
73 
74 int
75 vfs_hash_get(const struct mount *mp, u_int hash, int flags, struct thread *td,
76     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
77 {
78 	struct vnode *vp;
79 	enum vgetstate vs;
80 	int error;
81 
82 	while (1) {
83 		rw_rlock(&vfs_hash_lock);
84 		LIST_FOREACH(vp, vfs_hash_bucket(mp, hash), v_hashlist) {
85 			if (vp->v_hash != hash)
86 				continue;
87 			if (vp->v_mount != mp)
88 				continue;
89 			if (fn != NULL && fn(vp, arg))
90 				continue;
91 			vs = vget_prep(vp);
92 			rw_runlock(&vfs_hash_lock);
93 			error = vget_finish(vp, flags, vs);
94 			if (error == ENOENT && (flags & LK_NOWAIT) == 0)
95 				break;
96 			if (error)
97 				return (error);
98 			*vpp = vp;
99 			return (0);
100 		}
101 		if (vp == NULL) {
102 			rw_runlock(&vfs_hash_lock);
103 			*vpp = NULL;
104 			return (0);
105 		}
106 	}
107 }
108 
109 void
110 vfs_hash_ref(const struct mount *mp, u_int hash, struct thread *td,
111     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
112 {
113 	struct vnode *vp;
114 
115 	while (1) {
116 		rw_rlock(&vfs_hash_lock);
117 		LIST_FOREACH(vp, vfs_hash_bucket(mp, hash), v_hashlist) {
118 			if (vp->v_hash != hash)
119 				continue;
120 			if (vp->v_mount != mp)
121 				continue;
122 			if (fn != NULL && fn(vp, arg))
123 				continue;
124 			vhold(vp);
125 			rw_runlock(&vfs_hash_lock);
126 			vref(vp);
127 			vdrop(vp);
128 			*vpp = vp;
129 			return;
130 		}
131 		if (vp == NULL) {
132 			rw_runlock(&vfs_hash_lock);
133 			*vpp = NULL;
134 			return;
135 		}
136 	}
137 }
138 
139 void
140 vfs_hash_remove(struct vnode *vp)
141 {
142 
143 	rw_wlock(&vfs_hash_lock);
144 	LIST_REMOVE(vp, v_hashlist);
145 	rw_wunlock(&vfs_hash_lock);
146 }
147 
148 int
149 vfs_hash_insert(struct vnode *vp, u_int hash, int flags, struct thread *td,
150     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
151 {
152 	struct vnode *vp2;
153 	enum vgetstate vs;
154 	int error;
155 
156 	*vpp = NULL;
157 	while (1) {
158 		rw_wlock(&vfs_hash_lock);
159 		LIST_FOREACH(vp2,
160 		    vfs_hash_bucket(vp->v_mount, hash), v_hashlist) {
161 			if (vp2->v_hash != hash)
162 				continue;
163 			if (vp2->v_mount != vp->v_mount)
164 				continue;
165 			if (fn != NULL && fn(vp2, arg))
166 				continue;
167 			vs = vget_prep(vp2);
168 			rw_wunlock(&vfs_hash_lock);
169 			error = vget_finish(vp2, flags, vs);
170 			if (error == ENOENT && (flags & LK_NOWAIT) == 0)
171 				break;
172 			rw_wlock(&vfs_hash_lock);
173 			LIST_INSERT_HEAD(&vfs_hash_side, vp, v_hashlist);
174 			rw_wunlock(&vfs_hash_lock);
175 			vgone(vp);
176 			vput(vp);
177 			if (!error)
178 				*vpp = vp2;
179 			return (error);
180 		}
181 		if (vp2 == NULL)
182 			break;
183 	}
184 	vp->v_hash = hash;
185 	LIST_INSERT_HEAD(vfs_hash_bucket(vp->v_mount, hash), vp, v_hashlist);
186 	rw_wunlock(&vfs_hash_lock);
187 	return (0);
188 }
189 
190 void
191 vfs_hash_rehash(struct vnode *vp, u_int hash)
192 {
193 
194 	rw_wlock(&vfs_hash_lock);
195 	LIST_REMOVE(vp, v_hashlist);
196 	LIST_INSERT_HEAD(vfs_hash_bucket(vp->v_mount, hash), vp, v_hashlist);
197 	vp->v_hash = hash;
198 	rw_wunlock(&vfs_hash_lock);
199 }
200 
201 void
202 vfs_hash_changesize(u_long newmaxvnodes)
203 {
204 	struct vfs_hash_head *vfs_hash_newtbl, *vfs_hash_oldtbl;
205 	u_long vfs_hash_newmask, vfs_hash_oldmask;
206 	struct vnode *vp;
207 	int i;
208 
209 	vfs_hash_newtbl = hashinit(newmaxvnodes, M_VFS_HASH,
210 		&vfs_hash_newmask);
211 	/* If same hash table size, nothing to do */
212 	if (vfs_hash_mask == vfs_hash_newmask) {
213 		free(vfs_hash_newtbl, M_VFS_HASH);
214 		return;
215 	}
216 	/*
217 	 * Move everything from the old hash table to the new table.
218 	 * None of the vnodes in the table can be recycled because to
219 	 * do so, they have to be removed from the hash table.
220 	 */
221 	rw_wlock(&vfs_hash_lock);
222 	vfs_hash_oldtbl = vfs_hash_tbl;
223 	vfs_hash_oldmask = vfs_hash_mask;
224 	vfs_hash_tbl = vfs_hash_newtbl;
225 	vfs_hash_mask = vfs_hash_newmask;
226 	for (i = 0; i <= vfs_hash_oldmask; i++) {
227 		while ((vp = LIST_FIRST(&vfs_hash_oldtbl[i])) != NULL) {
228 			LIST_REMOVE(vp, v_hashlist);
229 			LIST_INSERT_HEAD(
230 			    vfs_hash_bucket(vp->v_mount, vp->v_hash),
231 			    vp, v_hashlist);
232 		}
233 	}
234 	rw_wunlock(&vfs_hash_lock);
235 	free(vfs_hash_oldtbl, M_VFS_HASH);
236 }
237