1 /*-
2 * Copyright (c) 1992, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software donated to Berkeley by
6 * Jan-Simon Pendry.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95
33 *
34 * $FreeBSD$
35 */
36
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/lock.h>
41 #include <sys/mutex.h>
42 #include <sys/malloc.h>
43 #include <sys/mount.h>
44 #include <sys/proc.h>
45 #include <sys/vnode.h>
46
47 #include <fs/nullfs/null.h>
48
49 #define LOG2_SIZEVNODE 8 /* log2(sizeof struct vnode) */
50 #define NNULLNODECACHE 16
51
52 /*
53 * Null layer cache:
54 * Each cache entry holds a reference to the lower vnode
55 * along with a pointer to the alias vnode. When an
56 * entry is added the lower vnode is VREF'd. When the
57 * alias is removed the lower vnode is vrele'd.
58 */
59
60 #define NULL_NHASH(vp) \
61 (&null_node_hashtbl[(((uintptr_t)vp)>>LOG2_SIZEVNODE) & null_node_hash])
62
63 static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl;
64 static u_long null_node_hash;
65 struct mtx null_hashmtx;
66
67 static MALLOC_DEFINE(M_NULLFSHASH, "NULLFS hash", "NULLFS hash table");
68 MALLOC_DEFINE(M_NULLFSNODE, "NULLFS node", "NULLFS vnode private part");
69
70 static struct vnode * null_hashget(struct mount *, struct vnode *);
71 static struct vnode * null_hashins(struct mount *, struct null_node *);
72
73 /*
74 * Initialise cache headers
75 */
76 int
77 nullfs_init(vfsp)
78 struct vfsconf *vfsp;
79 {
80
81 NULLFSDEBUG("nullfs_init\n"); /* printed during system boot */
82 null_node_hashtbl = hashinit(NNULLNODECACHE, M_NULLFSHASH, &null_node_hash);
83 mtx_init(&null_hashmtx, "nullhs", NULL, MTX_DEF);
84 return (0);
85 }
86
87 int
88 nullfs_uninit(vfsp)
89 struct vfsconf *vfsp;
90 {
91
92 mtx_destroy(&null_hashmtx);
93 free(null_node_hashtbl, M_NULLFSHASH);
94 return (0);
95 }
96
97 /*
98 * Return a VREF'ed alias for lower vnode if already exists, else 0.
99 * Lower vnode should be locked on entry and will be left locked on exit.
100 */
101 static struct vnode *
102 null_hashget(mp, lowervp)
103 struct mount *mp;
104 struct vnode *lowervp;
105 {
106 struct thread *td = curthread; /* XXX */
107 struct null_node_hashhead *hd;
108 struct null_node *a;
109 struct vnode *vp;
110 int error;
111
112 ASSERT_VOP_LOCKED(lowervp, "null_hashget");
113
114 /*
115 * Find hash base, and then search the (two-way) linked
116 * list looking for a null_node structure which is referencing
117 * the lower vnode. If found, the increment the null_node
118 * reference count (but NOT the lower vnode's VREF counter).
119 */
120 hd = NULL_NHASH(lowervp);
121 mtx_lock(&null_hashmtx);
122 LIST_FOREACH(a, hd, null_hash) {
123 if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) {
124 vp = NULLTOV(a);
125 VI_LOCK(vp);
126 mtx_unlock(&null_hashmtx);
127 /*
128 * We need to clear the OWEINACT flag here as this
129 * may lead vget() to try to lock our vnode which
130 * is already locked via lowervp.
131 */
132 vp->v_iflag &= ~VI_OWEINACT;
133 error = vget(vp, LK_INTERLOCK, td);
134 /*
135 * Since we have the lower node locked the nullfs
136 * node can not be in the process of recycling. If
137 * it had been recycled before we grabed the lower
138 * lock it would not have been found on the hash.
139 */
140 if (error)
141 panic("null_hashget: vget error %d", error);
142 return (vp);
143 }
144 }
145 mtx_unlock(&null_hashmtx);
146 return (NULLVP);
147 }
148
149 /*
150 * Act like null_hashget, but add passed null_node to hash if no existing
151 * node found.
152 */
153 static struct vnode *
154 null_hashins(mp, xp)
155 struct mount *mp;
156 struct null_node *xp;
157 {
158 struct thread *td = curthread; /* XXX */
159 struct null_node_hashhead *hd;
160 struct null_node *oxp;
161 struct vnode *ovp;
162 int error;
163
164 hd = NULL_NHASH(xp->null_lowervp);
165 mtx_lock(&null_hashmtx);
166 LIST_FOREACH(oxp, hd, null_hash) {
167 if (oxp->null_lowervp == xp->null_lowervp &&
168 NULLTOV(oxp)->v_mount == mp) {
169 /*
170 * See null_hashget for a description of this
171 * operation.
172 */
173 ovp = NULLTOV(oxp);
174 VI_LOCK(ovp);
175 mtx_unlock(&null_hashmtx);
176 ovp->v_iflag &= ~VI_OWEINACT;
177 error = vget(ovp, LK_INTERLOCK, td);
178 if (error)
179 panic("null_hashins: vget error %d", error);
180 return (ovp);
181 }
182 }
183 LIST_INSERT_HEAD(hd, xp, null_hash);
184 mtx_unlock(&null_hashmtx);
185 return (NULLVP);
186 }
187
188 /*
189 * Make a new or get existing nullfs node.
190 * Vp is the alias vnode, lowervp is the lower vnode.
191 *
192 * The lowervp assumed to be locked and having "spare" reference. This routine
193 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
194 * the caller's "spare" reference to created nullfs vnode.
195 */
196 int
197 null_nodeget(mp, lowervp, vpp)
198 struct mount *mp;
199 struct vnode *lowervp;
200 struct vnode **vpp;
201 {
202 struct null_node *xp;
203 struct vnode *vp;
204 int error;
205
206 /* Lookup the hash firstly */
207 *vpp = null_hashget(mp, lowervp);
208 if (*vpp != NULL) {
209 vrele(lowervp);
210 return (0);
211 }
212
213 /*
214 * We do not serialize vnode creation, instead we will check for
215 * duplicates later, when adding new vnode to hash.
216 *
217 * Note that duplicate can only appear in hash if the lowervp is
218 * locked LK_SHARED.
219 */
220
221 /*
222 * Do the MALLOC before the getnewvnode since doing so afterward
223 * might cause a bogus v_data pointer to get dereferenced
224 * elsewhere if MALLOC should block.
225 */
226 MALLOC(xp, struct null_node *, sizeof(struct null_node),
227 M_NULLFSNODE, M_WAITOK);
228
229 error = getnewvnode("null", mp, &null_vnodeops, &vp);
230 if (error) {
231 FREE(xp, M_NULLFSNODE);
232 return (error);
233 }
234
235 xp->null_vnode = vp;
236 xp->null_lowervp = lowervp;
237 vp->v_type = lowervp->v_type;
238 vp->v_data = xp;
239 vp->v_vnlock = lowervp->v_vnlock;
240 if (vp->v_vnlock == NULL)
241 panic("null_nodeget: Passed a NULL vnlock.\n");
242 /*
243 * Atomically insert our new node into the hash or vget existing
244 * if someone else has beaten us to it.
245 */
246 *vpp = null_hashins(mp, xp);
247 if (*vpp != NULL) {
248 vrele(lowervp);
249 vp->v_vnlock = &vp->v_lock;
250 xp->null_lowervp = NULL;
251 vrele(vp);
252 return (0);
253 }
254 *vpp = vp;
255
256 return (0);
257 }
258
259 /*
260 * Remove node from hash.
261 */
262 void
263 null_hashrem(xp)
264 struct null_node *xp;
265 {
266
267 mtx_lock(&null_hashmtx);
268 LIST_REMOVE(xp, null_hash);
269 mtx_unlock(&null_hashmtx);
270 }
271
272 #ifdef DIAGNOSTIC
273
274 #ifdef KDB
275 #define null_checkvp_barrier 1
276 #else
277 #define null_checkvp_barrier 0
278 #endif
279
280 struct vnode *
281 null_checkvp(vp, fil, lno)
282 struct vnode *vp;
283 char *fil;
284 int lno;
285 {
286 int interlock = 0;
287 struct null_node *a = VTONULL(vp);
288 #ifdef notyet
289 /*
290 * Can't do this check because vop_reclaim runs
291 * with a funny vop vector.
292 */
293 if (vp->v_op != null_vnodeop_p) {
294 printf ("null_checkvp: on non-null-node\n");
295 while (null_checkvp_barrier) /*WAIT*/ ;
296 panic("null_checkvp");
297 };
298 #endif
299 if (a->null_lowervp == NULLVP) {
300 /* Should never happen */
301 int i; u_long *p;
302 printf("vp = %p, ZERO ptr\n", (void *)vp);
303 for (p = (u_long *) a, i = 0; i < 8; i++)
304 printf(" %lx", p[i]);
305 printf("\n");
306 /* wait for debugger */
307 while (null_checkvp_barrier) /*WAIT*/ ;
308 panic("null_checkvp");
309 }
310 if (mtx_owned(VI_MTX(vp)) != 0) {
311 VI_UNLOCK(vp);
312 interlock = 1;
313 }
314 if (vrefcnt(a->null_lowervp) < 1) {
315 int i; u_long *p;
316 printf("vp = %p, unref'ed lowervp\n", (void *)vp);
317 for (p = (u_long *) a, i = 0; i < 8; i++)
318 printf(" %lx", p[i]);
319 printf("\n");
320 /* wait for debugger */
321 while (null_checkvp_barrier) /*WAIT*/ ;
322 panic ("null with unref'ed lowervp");
323 };
324 if (interlock != 0)
325 VI_LOCK(vp);
326 #ifdef notyet
327 printf("null %x/%d -> %x/%d [%s, %d]\n",
328 NULLTOV(a), vrefcnt(NULLTOV(a)),
329 a->null_lowervp, vrefcnt(a->null_lowervp),
330 fil, lno);
331 #endif
332 return a->null_lowervp;
333 }
334 #endif
Cache object: c861e7bd22870db126473f5b7cddc829
|