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 * 3. 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
35#include <sys/param.h>
36#include <sys/systm.h>
37#include <sys/kernel.h>
38#include <sys/lock.h>
39#include <sys/rwlock.h>
40#include <sys/malloc.h>
41#include <sys/mount.h>
42#include <sys/proc.h>
43#include <sys/vnode.h>
44
45#include <fs/nullfs/null.h>
46
47/*
48 * Null layer cache:
49 * Each cache entry holds a reference to the lower vnode
50 * along with a pointer to the alias vnode.  When an
51 * entry is added the lower vnode is VREF'd.  When the
52 * alias is removed the lower vnode is vrele'd.
53 */
54
55#define	NULL_NHASH(vp) (&null_node_hashtbl[vfs_hash_index(vp) & null_hash_mask])
56
57static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl;
58static struct rwlock null_hash_lock;
59static u_long null_hash_mask;
60
61static MALLOC_DEFINE(M_NULLFSHASH, "nullfs_hash", "NULLFS hash table");
62MALLOC_DEFINE(M_NULLFSNODE, "nullfs_node", "NULLFS vnode private part");
63
64static void null_hashins(struct mount *, struct null_node *);
65
66/*
67 * Initialise cache headers
68 */
69int
70nullfs_init(struct vfsconf *vfsp)
71{
72
73	null_node_hashtbl = hashinit(desiredvnodes, M_NULLFSHASH,
74	    &null_hash_mask);
75	rw_init(&null_hash_lock, "nullhs");
76	return (0);
77}
78
79int
80nullfs_uninit(struct vfsconf *vfsp)
81{
82
83	rw_destroy(&null_hash_lock);
84	hashdestroy(null_node_hashtbl, M_NULLFSHASH, null_hash_mask);
85	return (0);
86}
87
88/*
89 * Return a VREF'ed alias for lower vnode if already exists, else 0.
90 * Lower vnode should be locked on entry and will be left locked on exit.
91 */
92static struct vnode *
93null_hashget_locked(struct mount *mp, struct vnode *lowervp)
94{
95	struct null_node_hashhead *hd;
96	struct null_node *a;
97	struct vnode *vp;
98
99	ASSERT_VOP_LOCKED(lowervp, "null_hashget");
100	rw_assert(&null_hash_lock, RA_LOCKED);
101
102	/*
103	 * Find hash base, and then search the (two-way) linked
104	 * list looking for a null_node structure which is referencing
105	 * the lower vnode.  If found, the increment the null_node
106	 * reference count (but NOT the lower vnode's VREF counter).
107	 */
108	hd = NULL_NHASH(lowervp);
109	LIST_FOREACH(a, hd, null_hash) {
110		if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) {
111			/*
112			 * Since we have the lower node locked the nullfs
113			 * node can not be in the process of recycling.  If
114			 * it had been recycled before we grabed the lower
115			 * lock it would not have been found on the hash.
116			 */
117			vp = NULLTOV(a);
118			vref(vp);
119			return (vp);
120		}
121	}
122	return (NULLVP);
123}
124
125struct vnode *
126null_hashget(struct mount *mp, struct vnode *lowervp)
127{
128	struct null_node_hashhead *hd;
129	struct vnode *vp;
130
131	hd = NULL_NHASH(lowervp);
132	if (LIST_EMPTY(hd))
133		return (NULLVP);
134
135	rw_rlock(&null_hash_lock);
136	vp = null_hashget_locked(mp, lowervp);
137	rw_runlock(&null_hash_lock);
138
139	return (vp);
140}
141
142static void
143null_hashins(struct mount *mp, struct null_node *xp)
144{
145	struct null_node_hashhead *hd;
146#ifdef INVARIANTS
147	struct null_node *oxp;
148#endif
149
150	rw_assert(&null_hash_lock, RA_WLOCKED);
151
152	hd = NULL_NHASH(xp->null_lowervp);
153#ifdef INVARIANTS
154	LIST_FOREACH(oxp, hd, null_hash) {
155		if (oxp->null_lowervp == xp->null_lowervp &&
156		    NULLTOV(oxp)->v_mount == mp) {
157			VNASSERT(0, NULLTOV(oxp),
158			    ("vnode already in hash"));
159		}
160	}
161#endif
162	LIST_INSERT_HEAD(hd, xp, null_hash);
163}
164
165static void
166null_destroy_proto(struct vnode *vp, void *xp)
167{
168
169	lockmgr(&vp->v_lock, LK_EXCLUSIVE, NULL);
170	VI_LOCK(vp);
171	vp->v_data = NULL;
172	vp->v_vnlock = &vp->v_lock;
173	vp->v_op = &dead_vnodeops;
174	VI_UNLOCK(vp);
175	vgone(vp);
176	vput(vp);
177	free(xp, M_NULLFSNODE);
178}
179
180/*
181 * Make a new or get existing nullfs node.
182 * Vp is the alias vnode, lowervp is the lower vnode.
183 *
184 * The lowervp assumed to be locked and having "spare" reference. This routine
185 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers"
186 * the caller's "spare" reference to created nullfs vnode.
187 */
188int
189null_nodeget(struct mount *mp, struct vnode *lowervp, struct vnode **vpp)
190{
191	struct null_node *xp;
192	struct vnode *vp;
193	int error;
194
195	ASSERT_VOP_LOCKED(lowervp, "lowervp");
196	VNPASS(lowervp->v_usecount > 0, lowervp);
197
198	/* Lookup the hash firstly. */
199	*vpp = null_hashget(mp, lowervp);
200	if (*vpp != NULL) {
201		vrele(lowervp);
202		return (0);
203	}
204
205	/*
206	 * We do not serialize vnode creation, instead we will check for
207	 * duplicates later, when adding new vnode to hash.
208	 * Note that duplicate can only appear in hash if the lowervp is
209	 * locked LK_SHARED.
210	 */
211	xp = malloc(sizeof(struct null_node), M_NULLFSNODE, M_WAITOK);
212
213	error = getnewvnode("nullfs", mp, &null_vnodeops, &vp);
214	if (error) {
215		vput(lowervp);
216		free(xp, M_NULLFSNODE);
217		return (error);
218	}
219
220	VNPASS(vp->v_object == NULL, vp);
221	VNPASS((vn_irflag_read(vp) & VIRF_PGREAD) == 0, vp);
222
223	rw_wlock(&null_hash_lock);
224	xp->null_vnode = vp;
225	xp->null_lowervp = lowervp;
226	xp->null_flags = 0;
227	vp->v_type = lowervp->v_type;
228	vp->v_data = xp;
229	vp->v_vnlock = lowervp->v_vnlock;
230	*vpp = null_hashget_locked(mp, lowervp);
231	if (*vpp != NULL) {
232		rw_wunlock(&null_hash_lock);
233		vrele(lowervp);
234		null_destroy_proto(vp, xp);
235		return (0);
236	}
237
238	/*
239	 * We might miss the case where lower vnode sets VIRF_PGREAD
240	 * some time after construction, which is typical case.
241	 * null_open rechecks.
242	 */
243	if ((vn_irflag_read(lowervp) & VIRF_PGREAD) != 0) {
244		MPASS(lowervp->v_object != NULL);
245		vp->v_object = lowervp->v_object;
246		vn_irflag_set(vp, VIRF_PGREAD);
247	}
248	if (lowervp == MOUNTTONULLMOUNT(mp)->nullm_lowerrootvp)
249		vp->v_vflag |= VV_ROOT;
250
251	error = insmntque1(vp, mp);
252	if (error != 0) {
253		rw_wunlock(&null_hash_lock);
254		vput(lowervp);
255		vp->v_object = NULL;
256		null_destroy_proto(vp, xp);
257		return (error);
258	}
259
260	null_hashins(mp, xp);
261	vn_set_state(vp, VSTATE_CONSTRUCTED);
262	rw_wunlock(&null_hash_lock);
263	*vpp = vp;
264
265	return (0);
266}
267
268/*
269 * Remove node from hash.
270 */
271void
272null_hashrem(struct null_node *xp)
273{
274
275	rw_wlock(&null_hash_lock);
276	LIST_REMOVE(xp, null_hash);
277	rw_wunlock(&null_hash_lock);
278}
279
280#ifdef DIAGNOSTIC
281
282struct vnode *
283null_checkvp(struct vnode *vp, char *fil, int lno)
284{
285	struct null_node *a = VTONULL(vp);
286
287#ifdef notyet
288	/*
289	 * Can't do this check because vop_reclaim runs
290	 * with a funny vop vector.
291	 */
292	if (vp->v_op != null_vnodeop_p) {
293		printf ("null_checkvp: on non-null-node\n");
294		panic("null_checkvp");
295	}
296#endif
297	if (a->null_lowervp == NULLVP) {
298		/* Should never happen */
299		panic("null_checkvp %p", vp);
300	}
301	VI_LOCK_FLAGS(a->null_lowervp, MTX_DUPOK);
302	if (a->null_lowervp->v_usecount < 1)
303		panic ("null with unref'ed lowervp, vp %p lvp %p",
304		    vp, a->null_lowervp);
305	VI_UNLOCK(a->null_lowervp);
306#ifdef notyet
307	printf("null %x/%d -> %x/%d [%s, %d]\n",
308	        NULLTOV(a), vrefcnt(NULLTOV(a)),
309		a->null_lowervp, vrefcnt(a->null_lowervp),
310		fil, lno);
311#endif
312	return (a->null_lowervp);
313}
314#endif
315