null_subr.c revision 232304
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: head/sys/fs/nullfs/null_subr.c 232304 2012-02-29 15:18:04Z kib $
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
63static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl;
64static u_long null_node_hash;
65struct mtx null_hashmtx;
66
67static MALLOC_DEFINE(M_NULLFSHASH, "nullfs_hash", "NULLFS hash table");
68MALLOC_DEFINE(M_NULLFSNODE, "nullfs_node", "NULLFS vnode private part");
69
70static struct vnode * null_hashget(struct mount *, struct vnode *);
71static struct vnode * null_hashins(struct mount *, struct null_node *);
72
73/*
74 * Initialise cache headers
75 */
76int
77nullfs_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
87int
88nullfs_uninit(vfsp)
89	struct vfsconf *vfsp;
90{
91
92	mtx_destroy(&null_hashmtx);
93	hashdestroy(null_node_hashtbl, M_NULLFSHASH, null_node_hash);
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 */
101static struct vnode *
102null_hashget(mp, lowervp)
103	struct mount *mp;
104	struct vnode *lowervp;
105{
106	struct null_node_hashhead *hd;
107	struct null_node *a;
108	struct vnode *vp;
109
110	ASSERT_VOP_LOCKED(lowervp, "null_hashget");
111
112	/*
113	 * Find hash base, and then search the (two-way) linked
114	 * list looking for a null_node structure which is referencing
115	 * the lower vnode.  If found, the increment the null_node
116	 * reference count (but NOT the lower vnode's VREF counter).
117	 */
118	hd = NULL_NHASH(lowervp);
119	mtx_lock(&null_hashmtx);
120	LIST_FOREACH(a, hd, null_hash) {
121		if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) {
122			/*
123			 * Since we have the lower node locked the nullfs
124			 * node can not be in the process of recycling.  If
125			 * it had been recycled before we grabed the lower
126			 * lock it would not have been found on the hash.
127			 */
128			vp = NULLTOV(a);
129			vref(vp);
130			mtx_unlock(&null_hashmtx);
131			return (vp);
132		}
133	}
134	mtx_unlock(&null_hashmtx);
135	return (NULLVP);
136}
137
138/*
139 * Act like null_hashget, but add passed null_node to hash if no existing
140 * node found.
141 */
142static struct vnode *
143null_hashins(mp, xp)
144	struct mount *mp;
145	struct null_node *xp;
146{
147	struct null_node_hashhead *hd;
148	struct null_node *oxp;
149	struct vnode *ovp;
150
151	hd = NULL_NHASH(xp->null_lowervp);
152	mtx_lock(&null_hashmtx);
153	LIST_FOREACH(oxp, hd, null_hash) {
154		if (oxp->null_lowervp == xp->null_lowervp &&
155		    NULLTOV(oxp)->v_mount == mp) {
156			/*
157			 * See null_hashget for a description of this
158			 * operation.
159			 */
160			ovp = NULLTOV(oxp);
161			vref(ovp);
162			mtx_unlock(&null_hashmtx);
163			return (ovp);
164		}
165	}
166	LIST_INSERT_HEAD(hd, xp, null_hash);
167	mtx_unlock(&null_hashmtx);
168	return (NULLVP);
169}
170
171static void
172null_destroy_proto(struct vnode *vp, void *xp)
173{
174
175	VI_LOCK(vp);
176	vp->v_data = NULL;
177	vp->v_vnlock = &vp->v_lock;
178	vp->v_op = &dead_vnodeops;
179	VI_UNLOCK(vp);
180	(void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
181	vgone(vp);
182	vput(vp);
183	free(xp, M_NULLFSNODE);
184}
185
186static void
187null_insmntque_dtr(struct vnode *vp, void *xp)
188{
189
190	vput(((struct null_node *)xp)->null_lowervp);
191	null_destroy_proto(vp, xp);
192}
193
194/*
195 * Make a new or get existing nullfs node.
196 * Vp is the alias vnode, lowervp is the lower vnode.
197 *
198 * The lowervp assumed to be locked and having "spare" reference. This routine
199 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
200 * the caller's "spare" reference to created nullfs vnode.
201 */
202int
203null_nodeget(mp, lowervp, vpp)
204	struct mount *mp;
205	struct vnode *lowervp;
206	struct vnode **vpp;
207{
208	struct null_node *xp;
209	struct vnode *vp;
210	int error;
211
212	/*
213	 * The insmntque1() call below requires the exclusive lock on
214	 * the nullfs vnode.
215	 */
216	ASSERT_VOP_ELOCKED(lowervp, "lowervp");
217	KASSERT(lowervp->v_usecount >= 1, ("Unreferenced vnode %p\n", lowervp));
218
219	/* Lookup the hash firstly */
220	*vpp = null_hashget(mp, lowervp);
221	if (*vpp != NULL) {
222		vrele(lowervp);
223		return (0);
224	}
225
226	/*
227	 * We do not serialize vnode creation, instead we will check for
228	 * duplicates later, when adding new vnode to hash.
229	 * Note that duplicate can only appear in hash if the lowervp is
230	 * locked LK_SHARED.
231	 *
232	 * Do the MALLOC before the getnewvnode since doing so afterward
233	 * might cause a bogus v_data pointer to get dereferenced
234	 * elsewhere if MALLOC should block.
235	 */
236	xp = malloc(sizeof(struct null_node),
237	    M_NULLFSNODE, M_WAITOK);
238
239	error = getnewvnode("null", mp, &null_vnodeops, &vp);
240	if (error) {
241		vput(lowervp);
242		free(xp, M_NULLFSNODE);
243		return (error);
244	}
245
246	xp->null_vnode = vp;
247	xp->null_lowervp = lowervp;
248	vp->v_type = lowervp->v_type;
249	vp->v_data = xp;
250	vp->v_vnlock = lowervp->v_vnlock;
251	if (vp->v_vnlock == NULL)
252		panic("null_nodeget: Passed a NULL vnlock.\n");
253	error = insmntque1(vp, mp, null_insmntque_dtr, xp);
254	if (error != 0)
255		return (error);
256	/*
257	 * Atomically insert our new node into the hash or vget existing
258	 * if someone else has beaten us to it.
259	 */
260	*vpp = null_hashins(mp, xp);
261	if (*vpp != NULL) {
262		vrele(lowervp);
263		null_destroy_proto(vp, xp);
264		return (0);
265	}
266	*vpp = vp;
267
268	return (0);
269}
270
271/*
272 * Remove node from hash.
273 */
274void
275null_hashrem(xp)
276	struct null_node *xp;
277{
278
279	mtx_lock(&null_hashmtx);
280	LIST_REMOVE(xp, null_hash);
281	mtx_unlock(&null_hashmtx);
282}
283
284#ifdef DIAGNOSTIC
285
286struct vnode *
287null_checkvp(vp, fil, lno)
288	struct vnode *vp;
289	char *fil;
290	int lno;
291{
292	struct null_node *a = VTONULL(vp);
293
294#ifdef notyet
295	/*
296	 * Can't do this check because vop_reclaim runs
297	 * with a funny vop vector.
298	 */
299	if (vp->v_op != null_vnodeop_p) {
300		printf ("null_checkvp: on non-null-node\n");
301		panic("null_checkvp");
302	}
303#endif
304	if (a->null_lowervp == NULLVP) {
305		/* Should never happen */
306		panic("null_checkvp %p", vp);
307	}
308	VI_LOCK_FLAGS(a->null_lowervp, MTX_DUPOK);
309	if (a->null_lowervp->v_usecount < 1)
310		panic ("null with unref'ed lowervp, vp %p lvp %p",
311		    vp, a->null_lowervp);
312	VI_UNLOCK(a->null_lowervp);
313#ifdef notyet
314	printf("null %x/%d -> %x/%d [%s, %d]\n",
315	        NULLTOV(a), vrefcnt(NULLTOV(a)),
316		a->null_lowervp, vrefcnt(a->null_lowervp),
317		fil, lno);
318#endif
319	return (a->null_lowervp);
320}
321#endif
322