null_subr.c revision 330897
1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1992, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * This code is derived from software donated to Berkeley by
8 * Jan-Simon Pendry.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	@(#)null_subr.c	8.7 (Berkeley) 5/14/95
35 *
36 * $FreeBSD: stable/11/sys/fs/nullfs/null_subr.c 330897 2018-03-14 03:19:51Z eadler $
37 */
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/lock.h>
43#include <sys/rwlock.h>
44#include <sys/malloc.h>
45#include <sys/mount.h>
46#include <sys/proc.h>
47#include <sys/vnode.h>
48
49#include <fs/nullfs/null.h>
50
51/*
52 * Null layer cache:
53 * Each cache entry holds a reference to the lower vnode
54 * along with a pointer to the alias vnode.  When an
55 * entry is added the lower vnode is VREF'd.  When the
56 * alias is removed the lower vnode is vrele'd.
57 */
58
59#define	NULL_NHASH(vp) (&null_node_hashtbl[vfs_hash_index(vp) & null_hash_mask])
60
61static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl;
62static struct rwlock null_hash_lock;
63static u_long null_hash_mask;
64
65static MALLOC_DEFINE(M_NULLFSHASH, "nullfs_hash", "NULLFS hash table");
66MALLOC_DEFINE(M_NULLFSNODE, "nullfs_node", "NULLFS vnode private part");
67
68static struct vnode * null_hashins(struct mount *, struct null_node *);
69
70/*
71 * Initialise cache headers
72 */
73int
74nullfs_init(vfsp)
75	struct vfsconf *vfsp;
76{
77
78	null_node_hashtbl = hashinit(desiredvnodes, M_NULLFSHASH,
79	    &null_hash_mask);
80	rw_init(&null_hash_lock, "nullhs");
81	return (0);
82}
83
84int
85nullfs_uninit(vfsp)
86	struct vfsconf *vfsp;
87{
88
89	rw_destroy(&null_hash_lock);
90	hashdestroy(null_node_hashtbl, M_NULLFSHASH, null_hash_mask);
91	return (0);
92}
93
94/*
95 * Return a VREF'ed alias for lower vnode if already exists, else 0.
96 * Lower vnode should be locked on entry and will be left locked on exit.
97 */
98struct vnode *
99null_hashget(mp, lowervp)
100	struct mount *mp;
101	struct vnode *lowervp;
102{
103	struct null_node_hashhead *hd;
104	struct null_node *a;
105	struct vnode *vp;
106
107	ASSERT_VOP_LOCKED(lowervp, "null_hashget");
108
109	/*
110	 * Find hash base, and then search the (two-way) linked
111	 * list looking for a null_node structure which is referencing
112	 * the lower vnode.  If found, the increment the null_node
113	 * reference count (but NOT the lower vnode's VREF counter).
114	 */
115	hd = NULL_NHASH(lowervp);
116	rw_rlock(&null_hash_lock);
117	LIST_FOREACH(a, hd, null_hash) {
118		if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) {
119			/*
120			 * Since we have the lower node locked the nullfs
121			 * node can not be in the process of recycling.  If
122			 * it had been recycled before we grabed the lower
123			 * lock it would not have been found on the hash.
124			 */
125			vp = NULLTOV(a);
126			vref(vp);
127			rw_runlock(&null_hash_lock);
128			return (vp);
129		}
130	}
131	rw_runlock(&null_hash_lock);
132	return (NULLVP);
133}
134
135/*
136 * Act like null_hashget, but add passed null_node to hash if no existing
137 * node found.
138 */
139static struct vnode *
140null_hashins(mp, xp)
141	struct mount *mp;
142	struct null_node *xp;
143{
144	struct null_node_hashhead *hd;
145	struct null_node *oxp;
146	struct vnode *ovp;
147
148	hd = NULL_NHASH(xp->null_lowervp);
149	rw_wlock(&null_hash_lock);
150	LIST_FOREACH(oxp, hd, null_hash) {
151		if (oxp->null_lowervp == xp->null_lowervp &&
152		    NULLTOV(oxp)->v_mount == mp) {
153			/*
154			 * See null_hashget for a description of this
155			 * operation.
156			 */
157			ovp = NULLTOV(oxp);
158			vref(ovp);
159			rw_wunlock(&null_hash_lock);
160			return (ovp);
161		}
162	}
163	LIST_INSERT_HEAD(hd, xp, null_hash);
164	rw_wunlock(&null_hash_lock);
165	return (NULLVP);
166}
167
168static void
169null_destroy_proto(struct vnode *vp, void *xp)
170{
171
172	lockmgr(&vp->v_lock, LK_EXCLUSIVE, NULL);
173	VI_LOCK(vp);
174	vp->v_data = NULL;
175	vp->v_vnlock = &vp->v_lock;
176	vp->v_op = &dead_vnodeops;
177	VI_UNLOCK(vp);
178	vgone(vp);
179	vput(vp);
180	free(xp, M_NULLFSNODE);
181}
182
183static void
184null_insmntque_dtr(struct vnode *vp, void *xp)
185{
186
187	vput(((struct null_node *)xp)->null_lowervp);
188	null_destroy_proto(vp, xp);
189}
190
191/*
192 * Make a new or get existing nullfs node.
193 * Vp is the alias vnode, lowervp is the lower vnode.
194 *
195 * The lowervp assumed to be locked and having "spare" reference. This routine
196 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
197 * the caller's "spare" reference to created nullfs vnode.
198 */
199int
200null_nodeget(mp, lowervp, vpp)
201	struct mount *mp;
202	struct vnode *lowervp;
203	struct vnode **vpp;
204{
205	struct null_node *xp;
206	struct vnode *vp;
207	int error;
208
209	ASSERT_VOP_LOCKED(lowervp, "lowervp");
210	KASSERT(lowervp->v_usecount >= 1, ("Unreferenced vnode %p", lowervp));
211
212	/* Lookup the hash firstly. */
213	*vpp = null_hashget(mp, lowervp);
214	if (*vpp != NULL) {
215		vrele(lowervp);
216		return (0);
217	}
218
219	/*
220	 * The insmntque1() call below requires the exclusive lock on
221	 * the nullfs vnode.  Upgrade the lock now if hash failed to
222	 * provide ready to use vnode.
223	 */
224	if (VOP_ISLOCKED(lowervp) != LK_EXCLUSIVE) {
225		KASSERT((MOUNTTONULLMOUNT(mp)->nullm_flags & NULLM_CACHE) != 0,
226		    ("lowervp %p is not excl locked and cache is disabled",
227		    lowervp));
228		vn_lock(lowervp, LK_UPGRADE | LK_RETRY);
229		if ((lowervp->v_iflag & VI_DOOMED) != 0) {
230			vput(lowervp);
231			return (ENOENT);
232		}
233	}
234
235	/*
236	 * We do not serialize vnode creation, instead we will check for
237	 * duplicates later, when adding new vnode to hash.
238	 * Note that duplicate can only appear in hash if the lowervp is
239	 * locked LK_SHARED.
240	 */
241	xp = malloc(sizeof(struct null_node), M_NULLFSNODE, M_WAITOK);
242
243	error = getnewvnode("nullfs", mp, &null_vnodeops, &vp);
244	if (error) {
245		vput(lowervp);
246		free(xp, M_NULLFSNODE);
247		return (error);
248	}
249
250	xp->null_vnode = vp;
251	xp->null_lowervp = lowervp;
252	xp->null_flags = 0;
253	vp->v_type = lowervp->v_type;
254	vp->v_data = xp;
255	vp->v_vnlock = lowervp->v_vnlock;
256	error = insmntque1(vp, mp, null_insmntque_dtr, xp);
257	if (error != 0)
258		return (error);
259	/*
260	 * Atomically insert our new node into the hash or vget existing
261	 * if someone else has beaten us to it.
262	 */
263	*vpp = null_hashins(mp, xp);
264	if (*vpp != NULL) {
265		vrele(lowervp);
266		null_destroy_proto(vp, xp);
267		return (0);
268	}
269	*vpp = vp;
270
271	return (0);
272}
273
274/*
275 * Remove node from hash.
276 */
277void
278null_hashrem(xp)
279	struct null_node *xp;
280{
281
282	rw_wlock(&null_hash_lock);
283	LIST_REMOVE(xp, null_hash);
284	rw_wunlock(&null_hash_lock);
285}
286
287#ifdef DIAGNOSTIC
288
289struct vnode *
290null_checkvp(vp, fil, lno)
291	struct vnode *vp;
292	char *fil;
293	int lno;
294{
295	struct null_node *a = VTONULL(vp);
296
297#ifdef notyet
298	/*
299	 * Can't do this check because vop_reclaim runs
300	 * with a funny vop vector.
301	 */
302	if (vp->v_op != null_vnodeop_p) {
303		printf ("null_checkvp: on non-null-node\n");
304		panic("null_checkvp");
305	}
306#endif
307	if (a->null_lowervp == NULLVP) {
308		/* Should never happen */
309		panic("null_checkvp %p", vp);
310	}
311	VI_LOCK_FLAGS(a->null_lowervp, MTX_DUPOK);
312	if (a->null_lowervp->v_usecount < 1)
313		panic ("null with unref'ed lowervp, vp %p lvp %p",
314		    vp, a->null_lowervp);
315	VI_UNLOCK(a->null_lowervp);
316#ifdef notyet
317	printf("null %x/%d -> %x/%d [%s, %d]\n",
318	        NULLTOV(a), vrefcnt(NULLTOV(a)),
319		a->null_lowervp, vrefcnt(a->null_lowervp),
320		fil, lno);
321#endif
322	return (a->null_lowervp);
323}
324#endif
325