null_subr.c revision 229600
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 229600 2012-01-05 17:06:04Z dim $
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	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 */
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_insmntque_dtr(struct vnode *vp, void *xp)
173{
174
175	vput(((struct null_node *)xp)->null_lowervp);
176	vp->v_data = NULL;
177	vp->v_vnlock = &vp->v_lock;
178	free(xp, M_NULLFSNODE);
179	vp->v_op = &dead_vnodeops;
180	(void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
181	vgone(vp);
182	vput(vp);
183}
184
185/*
186 * Make a new or get existing nullfs node.
187 * Vp is the alias vnode, lowervp is the lower vnode.
188 *
189 * The lowervp assumed to be locked and having "spare" reference. This routine
190 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
191 * the caller's "spare" reference to created nullfs vnode.
192 */
193int
194null_nodeget(mp, lowervp, vpp)
195	struct mount *mp;
196	struct vnode *lowervp;
197	struct vnode **vpp;
198{
199	struct null_node *xp;
200	struct vnode *vp;
201	int error;
202
203	ASSERT_VOP_LOCKED(lowervp, "lowervp");
204	KASSERT(lowervp->v_usecount >= 1, ("Unreferenced vnode %p\n", lowervp));
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	xp = malloc(sizeof(struct null_node),
227	    M_NULLFSNODE, M_WAITOK);
228
229	error = getnewvnode("null", mp, &null_vnodeops, &vp);
230	if (error) {
231		vput(lowervp);
232		free(xp, M_NULLFSNODE);
233		return (error);
234	}
235
236	xp->null_vnode = vp;
237	xp->null_lowervp = lowervp;
238	vp->v_type = lowervp->v_type;
239	vp->v_data = xp;
240	vp->v_vnlock = lowervp->v_vnlock;
241	if (vp->v_vnlock == NULL)
242		panic("null_nodeget: Passed a NULL vnlock.\n");
243	error = insmntque1(vp, mp, null_insmntque_dtr, xp);
244	if (error != 0)
245		return (error);
246	/*
247	 * Atomically insert our new node into the hash or vget existing
248	 * if someone else has beaten us to it.
249	 */
250	*vpp = null_hashins(mp, xp);
251	if (*vpp != NULL) {
252		vrele(lowervp);
253		vp->v_vnlock = &vp->v_lock;
254		xp->null_lowervp = NULL;
255		vrele(vp);
256		return (0);
257	}
258	*vpp = vp;
259
260	return (0);
261}
262
263/*
264 * Remove node from hash.
265 */
266void
267null_hashrem(xp)
268	struct null_node *xp;
269{
270
271	mtx_lock(&null_hashmtx);
272	LIST_REMOVE(xp, null_hash);
273	mtx_unlock(&null_hashmtx);
274}
275
276#ifdef DIAGNOSTIC
277
278struct vnode *
279null_checkvp(vp, fil, lno)
280	struct vnode *vp;
281	char *fil;
282	int lno;
283{
284	struct null_node *a = VTONULL(vp);
285
286#ifdef notyet
287	/*
288	 * Can't do this check because vop_reclaim runs
289	 * with a funny vop vector.
290	 */
291	if (vp->v_op != null_vnodeop_p) {
292		printf ("null_checkvp: on non-null-node\n");
293		panic("null_checkvp");
294	}
295#endif
296	if (a->null_lowervp == NULLVP) {
297		/* Should never happen */
298		panic("null_checkvp %p", vp);
299	}
300	VI_LOCK_FLAGS(a->null_lowervp, MTX_DUPOK);
301	if (a->null_lowervp->v_usecount < 1)
302		panic ("null with unref'ed lowervp, vp %p lvp %p",
303		    vp, a->null_lowervp);
304	VI_UNLOCK(a->null_lowervp);
305#ifdef notyet
306	printf("null %x/%d -> %x/%d [%s, %d]\n",
307	        NULLTOV(a), vrefcnt(NULLTOV(a)),
308		a->null_lowervp, vrefcnt(a->null_lowervp),
309		fil, lno);
310#endif
311	return (a->null_lowervp);
312}
313#endif
314