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 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95
37 *
38 * $FreeBSD: releng/5.0/sys/fs/nullfs/null_subr.c 105077 2002-10-14 03:20:36Z mckusick $
39 */
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/mutex.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/proc.h>
49 #include <sys/vnode.h>
50
51 #include <fs/nullfs/null.h>
52
53 #define LOG2_SIZEVNODE 7 /* log2(sizeof struct vnode) */
54 #define NNULLNODECACHE 16
55
56 /*
57 * Null layer cache:
58 * Each cache entry holds a reference to the lower vnode
59 * along with a pointer to the alias vnode. When an
60 * entry is added the lower vnode is VREF'd. When the
61 * alias is removed the lower vnode is vrele'd.
62 */
63
64 #define NULL_NHASH(vp) \
65 (&null_node_hashtbl[(((uintptr_t)vp)>>LOG2_SIZEVNODE) & null_node_hash])
66
67 static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl;
68 static u_long null_node_hash;
69 struct mtx null_hashmtx;
70
71 static MALLOC_DEFINE(M_NULLFSHASH, "NULLFS hash", "NULLFS hash table");
72 MALLOC_DEFINE(M_NULLFSNODE, "NULLFS node", "NULLFS vnode private part");
73
74 static struct vnode * null_hashget(struct vnode *);
75 static struct vnode * null_hashins(struct null_node *);
76
77 /*
78 * Initialise cache headers
79 */
80 int
81 nullfs_init(vfsp)
82 struct vfsconf *vfsp;
83 {
84
85 NULLFSDEBUG("nullfs_init\n"); /* printed during system boot */
86 null_node_hashtbl = hashinit(NNULLNODECACHE, M_NULLFSHASH, &null_node_hash);
87 mtx_init(&null_hashmtx, "nullhs", NULL, MTX_DEF);
88 return (0);
89 }
90
91 int
92 nullfs_uninit(vfsp)
93 struct vfsconf *vfsp;
94 {
95
96 mtx_destroy(&null_hashmtx);
97 free(null_node_hashtbl, M_NULLFSHASH);
98 return (0);
99 }
100
101 /*
102 * Return a VREF'ed alias for lower vnode if already exists, else 0.
103 * Lower vnode should be locked on entry and will be left locked on exit.
104 */
105 static struct vnode *
106 null_hashget(lowervp)
107 struct vnode *lowervp;
108 {
109 struct thread *td = curthread; /* XXX */
110 struct null_node_hashhead *hd;
111 struct null_node *a;
112 struct vnode *vp;
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 loop:
122 mtx_lock(&null_hashmtx);
123 LIST_FOREACH(a, hd, null_hash) {
124 if (a->null_lowervp == lowervp) {
125 vp = NULLTOV(a);
126 mtx_lock(&vp->v_interlock);
127 mtx_unlock(&null_hashmtx);
128 /*
129 * We need vget for the VXLOCK
130 * stuff, but we don't want to lock
131 * the lower node.
132 */
133 if (vget(vp, LK_EXCLUSIVE | LK_THISLAYER | LK_INTERLOCK, td))
134 goto loop;
135
136 return (vp);
137 }
138 }
139 mtx_unlock(&null_hashmtx);
140 return (NULLVP);
141 }
142
143 /*
144 * Act like null_hashget, but add passed null_node to hash if no existing
145 * node found.
146 */
147 static struct vnode *
148 null_hashins(xp)
149 struct null_node *xp;
150 {
151 struct thread *td = curthread; /* XXX */
152 struct null_node_hashhead *hd;
153 struct null_node *oxp;
154 struct vnode *ovp;
155
156 hd = NULL_NHASH(xp->null_lowervp);
157 loop:
158 mtx_lock(&null_hashmtx);
159 LIST_FOREACH(oxp, hd, null_hash) {
160 if (oxp->null_lowervp == xp->null_lowervp) {
161 ovp = NULLTOV(oxp);
162 mtx_lock(&ovp->v_interlock);
163 mtx_unlock(&null_hashmtx);
164 if (vget(ovp, LK_EXCLUSIVE | LK_THISLAYER | LK_INTERLOCK, td))
165 goto loop;
166
167 return (ovp);
168 }
169 }
170 LIST_INSERT_HEAD(hd, xp, null_hash);
171 mtx_unlock(&null_hashmtx);
172 return (NULLVP);
173 }
174
175 /*
176 * Make a new or get existing nullfs node.
177 * Vp is the alias vnode, lowervp is the lower vnode.
178 *
179 * The lowervp assumed to be locked and having "spare" reference. This routine
180 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
181 * the caller's "spare" reference to created nullfs vnode.
182 */
183 int
184 null_nodeget(mp, lowervp, vpp)
185 struct mount *mp;
186 struct vnode *lowervp;
187 struct vnode **vpp;
188 {
189 struct thread *td = curthread; /* XXX */
190 struct null_node *xp;
191 struct vnode *vp;
192 int error;
193
194 /* Lookup the hash firstly */
195 *vpp = null_hashget(lowervp);
196 if (*vpp != NULL) {
197 vrele(lowervp);
198 return (0);
199 }
200
201 /*
202 * We do not serialize vnode creation, instead we will check for
203 * duplicates later, when adding new vnode to hash.
204 *
205 * Note that duplicate can only appear in hash if the lowervp is
206 * locked LK_SHARED.
207 */
208
209 /*
210 * Do the MALLOC before the getnewvnode since doing so afterward
211 * might cause a bogus v_data pointer to get dereferenced
212 * elsewhere if MALLOC should block.
213 */
214 MALLOC(xp, struct null_node *, sizeof(struct null_node),
215 M_NULLFSNODE, M_WAITOK);
216
217 error = getnewvnode("null", mp, null_vnodeop_p, &vp);
218 if (error) {
219 FREE(xp, M_NULLFSNODE);
220 return (error);
221 }
222
223 xp->null_vnode = vp;
224 xp->null_lowervp = lowervp;
225
226 vp->v_type = lowervp->v_type;
227 vp->v_data = xp;
228
229 /*
230 * From NetBSD:
231 * Now lock the new node. We rely on the fact that we were passed
232 * a locked vnode. If the lower node is exporting a struct lock
233 * (v_vnlock != NULL) then we just set the upper v_vnlock to the
234 * lower one, and both are now locked. If the lower node is exporting
235 * NULL, then we copy that up and manually lock the new vnode.
236 */
237
238 vp->v_vnlock = lowervp->v_vnlock;
239 error = VOP_LOCK(vp, LK_EXCLUSIVE | LK_THISLAYER, td);
240 if (error)
241 panic("null_nodeget: can't lock new vnode\n");
242
243 /*
244 * Atomically insert our new node into the hash or vget existing
245 * if someone else has beaten us to it.
246 */
247 *vpp = null_hashins(xp);
248 if (*vpp != NULL) {
249 vrele(lowervp);
250 VOP_UNLOCK(vp, LK_THISLAYER, td);
251 vp->v_vnlock = NULL;
252 xp->null_lowervp = NULL;
253 vrele(vp);
254 return (0);
255 }
256
257 /*
258 * XXX We take extra vref just to workaround UFS's XXX:
259 * UFS can vrele() vnode in VOP_CLOSE() in some cases. Luckily, this
260 * can only happen if v_usecount == 1. To workaround, we just don't
261 * let v_usecount be 1, it will be 2 or more.
262 */
263 VREF(lowervp);
264
265 *vpp = vp;
266
267 return (0);
268 }
269
270 /*
271 * Remove node from hash.
272 */
273 void
274 null_hashrem(xp)
275 struct null_node *xp;
276 {
277
278 mtx_lock(&null_hashmtx);
279 LIST_REMOVE(xp, null_hash);
280 mtx_unlock(&null_hashmtx);
281 }
282
283 #ifdef DIAGNOSTIC
284 #include "opt_ddb.h"
285
286 #ifdef DDB
287 #define null_checkvp_barrier 1
288 #else
289 #define null_checkvp_barrier 0
290 #endif
291
292 struct vnode *
293 null_checkvp(vp, fil, lno)
294 struct vnode *vp;
295 char *fil;
296 int lno;
297 {
298 struct null_node *a = VTONULL(vp);
299 #ifdef notyet
300 /*
301 * Can't do this check because vop_reclaim runs
302 * with a funny vop vector.
303 */
304 if (vp->v_op != null_vnodeop_p) {
305 printf ("null_checkvp: on non-null-node\n");
306 while (null_checkvp_barrier) /*WAIT*/ ;
307 panic("null_checkvp");
308 };
309 #endif
310 if (a->null_lowervp == NULLVP) {
311 /* Should never happen */
312 int i; u_long *p;
313 printf("vp = %p, ZERO ptr\n", (void *)vp);
314 for (p = (u_long *) a, i = 0; i < 8; i++)
315 printf(" %lx", p[i]);
316 printf("\n");
317 /* wait for debugger */
318 while (null_checkvp_barrier) /*WAIT*/ ;
319 panic("null_checkvp");
320 }
321 if (vrefcnt(a->null_lowervp) < 1) {
322 int i; u_long *p;
323 printf("vp = %p, unref'ed lowervp\n", (void *)vp);
324 for (p = (u_long *) a, i = 0; i < 8; i++)
325 printf(" %lx", p[i]);
326 printf("\n");
327 /* wait for debugger */
328 while (null_checkvp_barrier) /*WAIT*/ ;
329 panic ("null with unref'ed lowervp");
330 };
331 #ifdef notyet
332 printf("null %x/%d -> %x/%d [%s, %d]\n",
333 NULLTOV(a), vrefcnt(NULLTOV(a)),
334 a->null_lowervp, vrefcnt(a->null_lowervp),
335 fil, lno);
336 #endif
337 return a->null_lowervp;
338 }
339 #endif
Cache object: 63da6217562dbf905eb978a93f85ed93
|