xref: /freebsd/sys/kern/vfs_hash.c (revision fdafd315ad0d0f28a11b9fb4476a9ab059c62b92)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
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/param.h>
31 #include <sys/systm.h>
32 #include <sys/kernel.h>
33 #include <sys/malloc.h>
34 #include <sys/mount.h>
35 #include <sys/rwlock.h>
36 #include <sys/vnode.h>
37 
38 static MALLOC_DEFINE(M_VFS_HASH, "vfs_hash", "VFS hash table");
39 
40 static LIST_HEAD(vfs_hash_head, vnode)	*vfs_hash_tbl;
41 static LIST_HEAD(,vnode)		vfs_hash_side;
42 static u_long				vfs_hash_mask;
43 static struct rwlock __exclusive_cache_line vfs_hash_lock;
44 
45 static void
vfs_hashinit(void * dummy __unused)46 vfs_hashinit(void *dummy __unused)
47 {
48 
49 	vfs_hash_tbl = hashinit(desiredvnodes, M_VFS_HASH, &vfs_hash_mask);
50 	rw_init(&vfs_hash_lock, "vfs hash");
51 	LIST_INIT(&vfs_hash_side);
52 }
53 
54 /* Must be SI_ORDER_SECOND so desiredvnodes is available */
55 SYSINIT(vfs_hash, SI_SUB_VFS, SI_ORDER_SECOND, vfs_hashinit, NULL);
56 
57 u_int
vfs_hash_index(struct vnode * vp)58 vfs_hash_index(struct vnode *vp)
59 {
60 
61 	return (vp->v_hash + vp->v_mount->mnt_hashseed);
62 }
63 
64 static struct vfs_hash_head *
vfs_hash_bucket(const struct mount * mp,u_int hash)65 vfs_hash_bucket(const struct mount *mp, u_int hash)
66 {
67 
68 	return (&vfs_hash_tbl[(hash + mp->mnt_hashseed) & vfs_hash_mask]);
69 }
70 
71 int
vfs_hash_get(const struct mount * mp,u_int hash,int flags,struct thread * td,struct vnode ** vpp,vfs_hash_cmp_t * fn,void * arg)72 vfs_hash_get(const struct mount *mp, u_int hash, int flags, struct thread *td,
73     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
74 {
75 	struct vnode *vp;
76 	enum vgetstate vs;
77 	int error;
78 
79 	while (1) {
80 		rw_rlock(&vfs_hash_lock);
81 		LIST_FOREACH(vp, vfs_hash_bucket(mp, hash), v_hashlist) {
82 			if (vp->v_hash != hash)
83 				continue;
84 			if (vp->v_mount != mp)
85 				continue;
86 			if (fn != NULL && fn(vp, arg))
87 				continue;
88 			vs = vget_prep(vp);
89 			rw_runlock(&vfs_hash_lock);
90 			error = vget_finish(vp, flags, vs);
91 			if (error == ENOENT && (flags & LK_NOWAIT) == 0)
92 				break;
93 			if (error != 0)
94 				return (error);
95 			if (vp->v_hash != hash ||
96 			    (fn != NULL && fn(vp, arg))) {
97 				vput(vp);
98 				/* Restart the bucket walk. */
99 				break;
100 			}
101 			*vpp = vp;
102 			return (0);
103 		}
104 		if (vp == NULL) {
105 			rw_runlock(&vfs_hash_lock);
106 			*vpp = NULL;
107 			return (0);
108 		}
109 	}
110 }
111 
112 void
vfs_hash_ref(const struct mount * mp,u_int hash,struct thread * td,struct vnode ** vpp,vfs_hash_cmp_t * fn,void * arg)113 vfs_hash_ref(const struct mount *mp, u_int hash, struct thread *td,
114     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
115 {
116 	struct vnode *vp;
117 
118 	while (1) {
119 		rw_rlock(&vfs_hash_lock);
120 		LIST_FOREACH(vp, vfs_hash_bucket(mp, hash), v_hashlist) {
121 			if (vp->v_hash != hash)
122 				continue;
123 			if (vp->v_mount != mp)
124 				continue;
125 			if (fn != NULL && fn(vp, arg))
126 				continue;
127 			vhold(vp);
128 			rw_runlock(&vfs_hash_lock);
129 			vref(vp);
130 			vdrop(vp);
131 			*vpp = vp;
132 			return;
133 		}
134 		if (vp == NULL) {
135 			rw_runlock(&vfs_hash_lock);
136 			*vpp = NULL;
137 			return;
138 		}
139 	}
140 }
141 
142 void
vfs_hash_remove(struct vnode * vp)143 vfs_hash_remove(struct vnode *vp)
144 {
145 
146 	rw_wlock(&vfs_hash_lock);
147 	LIST_REMOVE(vp, v_hashlist);
148 	rw_wunlock(&vfs_hash_lock);
149 }
150 
151 int
vfs_hash_insert(struct vnode * vp,u_int hash,int flags,struct thread * td,struct vnode ** vpp,vfs_hash_cmp_t * fn,void * arg)152 vfs_hash_insert(struct vnode *vp, u_int hash, int flags, struct thread *td,
153     struct vnode **vpp, vfs_hash_cmp_t *fn, void *arg)
154 {
155 	struct vnode *vp2;
156 	enum vgetstate vs;
157 	int error;
158 
159 	*vpp = NULL;
160 	while (1) {
161 		rw_wlock(&vfs_hash_lock);
162 		LIST_FOREACH(vp2,
163 		    vfs_hash_bucket(vp->v_mount, hash), v_hashlist) {
164 			if (vp2->v_hash != hash)
165 				continue;
166 			if (vp2->v_mount != vp->v_mount)
167 				continue;
168 			if (fn != NULL && fn(vp2, arg))
169 				continue;
170 			vs = vget_prep(vp2);
171 			rw_wunlock(&vfs_hash_lock);
172 			error = vget_finish(vp2, flags, vs);
173 			if (error == ENOENT && (flags & LK_NOWAIT) == 0)
174 				break;
175 			rw_wlock(&vfs_hash_lock);
176 			LIST_INSERT_HEAD(&vfs_hash_side, vp, v_hashlist);
177 			rw_wunlock(&vfs_hash_lock);
178 			vgone(vp);
179 			vput(vp);
180 			if (!error)
181 				*vpp = vp2;
182 			return (error);
183 		}
184 		if (vp2 == NULL)
185 			break;
186 	}
187 	vp->v_hash = hash;
188 	LIST_INSERT_HEAD(vfs_hash_bucket(vp->v_mount, hash), vp, v_hashlist);
189 	rw_wunlock(&vfs_hash_lock);
190 	return (0);
191 }
192 
193 void
vfs_hash_rehash(struct vnode * vp,u_int hash)194 vfs_hash_rehash(struct vnode *vp, u_int hash)
195 {
196 	ASSERT_VOP_ELOCKED(vp, "rehash requires excl lock");
197 
198 	rw_wlock(&vfs_hash_lock);
199 	LIST_REMOVE(vp, v_hashlist);
200 	LIST_INSERT_HEAD(vfs_hash_bucket(vp->v_mount, hash), vp, v_hashlist);
201 	vp->v_hash = hash;
202 	rw_wunlock(&vfs_hash_lock);
203 }
204 
205 void
vfs_hash_changesize(u_long newmaxvnodes)206 vfs_hash_changesize(u_long newmaxvnodes)
207 {
208 	struct vfs_hash_head *vfs_hash_newtbl, *vfs_hash_oldtbl;
209 	u_long vfs_hash_newmask, vfs_hash_oldmask;
210 	struct vnode *vp;
211 	int i;
212 
213 	vfs_hash_newtbl = hashinit(newmaxvnodes, M_VFS_HASH,
214 		&vfs_hash_newmask);
215 	/* If same hash table size, nothing to do */
216 	if (vfs_hash_mask == vfs_hash_newmask) {
217 		free(vfs_hash_newtbl, M_VFS_HASH);
218 		return;
219 	}
220 	/*
221 	 * Move everything from the old hash table to the new table.
222 	 * None of the vnodes in the table can be recycled because to
223 	 * do so, they have to be removed from the hash table.
224 	 */
225 	rw_wlock(&vfs_hash_lock);
226 	vfs_hash_oldtbl = vfs_hash_tbl;
227 	vfs_hash_oldmask = vfs_hash_mask;
228 	vfs_hash_tbl = vfs_hash_newtbl;
229 	vfs_hash_mask = vfs_hash_newmask;
230 	for (i = 0; i <= vfs_hash_oldmask; i++) {
231 		while ((vp = LIST_FIRST(&vfs_hash_oldtbl[i])) != NULL) {
232 			LIST_REMOVE(vp, v_hashlist);
233 			LIST_INSERT_HEAD(
234 			    vfs_hash_bucket(vp->v_mount, vp->v_hash),
235 			    vp, v_hashlist);
236 		}
237 	}
238 	rw_wunlock(&vfs_hash_lock);
239 	free(vfs_hash_oldtbl, M_VFS_HASH);
240 }
241