xref: /netbsd-src/sys/miscfs/genfs/layer_subr.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: layer_subr.c,v 1.31 2010/07/21 17:52:12 hannken Exp $	*/
2 
3 /*
4  * Copyright (c) 1999 National Aeronautics & Space Administration
5  * All rights reserved.
6  *
7  * This software was written by William Studenmund of the
8  * Numerical Aerospace Simulation Facility, NASA Ames Research Center.
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 National Aeronautics & Space Administration
19  *    nor the names of its contributors may be used to endorse or promote
20  *    products derived from this software without specific prior written
21  *    permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE NATIONAL AERONAUTICS & SPACE ADMINISTRATION
24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE ADMINISTRATION OR CONTRIB-
27  * UTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
28  * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /*
37  * Copyright (c) 1992, 1993
38  *	The Regents of the University of California.  All rights reserved.
39  *
40  * This code is derived from software donated to Berkeley by
41  * Jan-Simon Pendry.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. Neither the name of the University nor the names of its contributors
52  *    may be used to endorse or promote products derived from this software
53  *    without specific prior written permission.
54  *
55  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65  * SUCH DAMAGE.
66  *
67  *	from: Id: lofs_subr.c,v 1.11 1992/05/30 10:05:43 jsp Exp
68  *	@(#)null_subr.c	8.7 (Berkeley) 5/14/95
69  */
70 
71 #include <sys/cdefs.h>
72 __KERNEL_RCSID(0, "$NetBSD: layer_subr.c,v 1.31 2010/07/21 17:52:12 hannken Exp $");
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/proc.h>
77 #include <sys/time.h>
78 #include <sys/vnode.h>
79 #include <sys/mount.h>
80 #include <sys/namei.h>
81 #include <sys/kmem.h>
82 #include <sys/malloc.h>
83 
84 #include <miscfs/specfs/specdev.h>
85 #include <miscfs/genfs/layer.h>
86 #include <miscfs/genfs/layer_extern.h>
87 
88 #ifdef LAYERFS_DIAGNOSTIC
89 int layerfs_debug = 1;
90 #endif
91 
92 /*
93  * layer cache:
94  * Each cache entry holds a reference to the lower vnode
95  * along with a pointer to the alias vnode.  When an
96  * entry is added the lower vnode is VREF'd.  When the
97  * alias is removed the lower vnode is vrele'd.
98  */
99 
100 void
101 layerfs_init(void)
102 {
103 	/* Nothing. */
104 }
105 
106 void
107 layerfs_done(void)
108 {
109 	/* Nothing. */
110 }
111 
112 /*
113  * layer_node_find: find and return alias for lower vnode or NULL.
114  *
115  * => Return alias vnode locked and referenced. if already exists.
116  * => The layermp's hashlock must be held on entry, we will unlock on success.
117  */
118 struct vnode *
119 layer_node_find(struct mount *mp, struct vnode *lowervp)
120 {
121 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
122 	struct layer_node_hashhead *hd;
123 	struct layer_node *a;
124 	struct vnode *vp;
125 	int error;
126 
127 	/*
128 	 * Find hash bucket and search the (two-way) linked list looking
129 	 * for a layerfs node structure which is referencing the lower vnode.
130 	 * If found, the increment the layer_node reference count, but NOT
131 	 * the lower vnode's reference counter.  Return vnode locked.
132 	 */
133 	KASSERT(mutex_owned(&lmp->layerm_hashlock));
134 	hd = LAYER_NHASH(lmp, lowervp);
135 loop:
136 	LIST_FOREACH(a, hd, layer_hash) {
137 		if (a->layer_lowervp != lowervp) {
138 			continue;
139 		}
140 		vp = LAYERTOV(a);
141 		if (vp->v_mount != mp) {
142 			continue;
143 		}
144 		mutex_enter(&vp->v_interlock);
145 		/*
146 		 * If we find a node being cleaned out, then ignore it and
147 		 * continue.  A thread trying to clean out the extant layer
148 		 * vnode needs to acquire the shared lock (i.e. the lower
149 		 * vnode's lock), which our caller already holds.  To allow
150 		 * the cleaning to succeed the current thread must make
151 		 * progress.  So, for a brief time more than one vnode in a
152 		 * layered file system may refer to a single vnode in the
153 		 * lower file system.
154 		 */
155 		if ((vp->v_iflag & VI_XLOCK) != 0) {
156 			mutex_exit(&vp->v_interlock);
157 			continue;
158 		}
159 		mutex_exit(&lmp->layerm_hashlock);
160 		/*
161 		 * We must not let vget() try to lock the layer vp, since
162 		 * the lower vp is already locked and locking the layer vp
163 		 * will involve locking the lower vp.
164 		 */
165 		error = vget(vp, LK_NOWAIT);
166 		if (error) {
167 			kpause("layerfs", false, 1, NULL);
168 			mutex_enter(&lmp->layerm_hashlock);
169 			goto loop;
170 		}
171 		return vp;
172 	}
173 	return NULL;
174 }
175 
176 /*
177  * layer_node_alloc: make a new layerfs vnode.
178  *
179  * => vp is the alias vnode, lowervp is the lower vnode.
180  * => We will hold a reference to lowervp.
181  */
182 int
183 layer_node_alloc(struct mount *mp, struct vnode *lowervp, struct vnode **vpp)
184 {
185 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
186 	struct layer_node_hashhead *hd;
187 	struct layer_node *xp;
188 	struct vnode *vp, *nvp;
189 	int error;
190 	extern int (**dead_vnodeop_p)(void *);
191 
192 	error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, &vp);
193 	if (error) {
194 		return error;
195 	}
196 	vp->v_type = lowervp->v_type;
197 	mutex_enter(&vp->v_interlock);
198 	vp->v_iflag |= VI_LAYER;
199 	mutex_exit(&vp->v_interlock);
200 
201 	xp = kmem_alloc(lmp->layerm_size, KM_SLEEP);
202 	if (xp == NULL) {
203 		ungetnewvnode(vp);
204 		return ENOMEM;
205 	}
206 	if (vp->v_type == VBLK || vp->v_type == VCHR) {
207 		spec_node_init(vp, lowervp->v_rdev);
208 	}
209 
210 	vp->v_data = xp;
211 	vp->v_vflag = (vp->v_vflag & ~VV_MPSAFE) |
212 	    (lowervp->v_vflag & VV_MPSAFE);
213 	xp->layer_vnode = vp;
214 	xp->layer_lowervp = lowervp;
215 	xp->layer_flags = 0;
216 
217 	/*
218 	 * Before inserting the node into the hash, check if other thread
219 	 * did not race with us.  If so - return that node, destroy ours.
220 	 */
221 	mutex_enter(&lmp->layerm_hashlock);
222 	if ((nvp = layer_node_find(mp, lowervp)) != NULL) {
223 		/* Free the structures we have created. */
224 		if (vp->v_type == VBLK || vp->v_type == VCHR)
225 			spec_node_destroy(vp);
226 
227 		vp->v_type = VBAD;		/* node is discarded */
228 		vp->v_op = dead_vnodeop_p;	/* so ops will still work */
229 		vrele(vp);			/* get rid of it. */
230 		kmem_free(xp, lmp->layerm_size);
231 		*vpp = nvp;
232 		return 0;
233 	}
234 
235 	/*
236 	 * Insert the new node into the hash.
237 	 * Add a reference to the lower node.
238 	 */
239 	vref(lowervp);
240 	hd = LAYER_NHASH(lmp, lowervp);
241 	LIST_INSERT_HEAD(hd, xp, layer_hash);
242 	uvm_vnp_setsize(vp, 0);
243 	mutex_exit(&lmp->layerm_hashlock);
244 
245 	*vpp = vp;
246 	return 0;
247 }
248 
249 /*
250  * layer_node_create: try to find an existing layerfs vnode refering to it,
251  * otherwise make a new vnode which contains a reference to the lower vnode.
252  *
253  * => Caller should lock the lower node.
254  */
255 int
256 layer_node_create(struct mount *mp, struct vnode *lowervp, struct vnode **nvpp)
257 {
258 	struct vnode *aliasvp;
259 	struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
260 
261 	KASSERT(VOP_ISLOCKED(lowervp));
262 
263 	mutex_enter(&lmp->layerm_hashlock);
264 	aliasvp = layer_node_find(mp, lowervp);
265 	if (aliasvp != NULL) {
266 		/*
267 		 * Note: layer_node_find() has taken another reference to
268 		 * the alias vnode and moved the lock holding to aliasvp.
269 		 */
270 #ifdef LAYERFS_DIAGNOSTIC
271 		if (layerfs_debug)
272 			vprint("layer_node_create: exists", aliasvp);
273 #endif
274 	} else {
275 		int error;
276 
277 		mutex_exit(&lmp->layerm_hashlock);
278 		/*
279 		 * Get a new vnode.  Make it to reference the layer_node.
280 		 * Note: aliasvp will be return with the reference held.
281 		 */
282 		error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp);
283 		if (error)
284 			return error;
285 #ifdef LAYERFS_DIAGNOSTIC
286 		if (layerfs_debug)
287 			printf("layer_node_create: create new alias vnode\n");
288 #endif
289 	}
290 
291 	/*
292 	 * Now that we acquired a reference on the upper vnode, release one
293 	 * on the lower node.  The existence of the layer_node retains one
294 	 * reference to the lower node.
295 	 */
296 	vrele(lowervp);
297 	KASSERT(lowervp->v_usecount > 0);
298 
299 #ifdef LAYERFS_DIAGNOSTIC
300 	if (layerfs_debug)
301 		vprint("layer_node_create: alias", aliasvp);
302 #endif
303 	*nvpp = aliasvp;
304 	return 0;
305 }
306 
307 #ifdef LAYERFS_DIAGNOSTIC
308 struct vnode *
309 layer_checkvp(struct vnode *vp, const char *fil, int lno)
310 {
311 	struct layer_node *a = VTOLAYER(vp);
312 #ifdef notyet
313 	/*
314 	 * Can't do this check because vop_reclaim runs
315 	 * with a funny vop vector.
316 	 *
317 	 * WRS - no it doesnt...
318 	 */
319 	if (vp->v_op != layer_vnodeop_p) {
320 		printf ("layer_checkvp: on non-layer-node\n");
321 #ifdef notyet
322 		while (layer_checkvp_barrier) /*WAIT*/ ;
323 #endif
324 		panic("layer_checkvp");
325 	};
326 #endif
327 	if (a->layer_lowervp == NULL) {
328 		/* Should never happen */
329 		int i; u_long *p;
330 		printf("vp = %p, ZERO ptr\n", vp);
331 		for (p = (u_long *) a, i = 0; i < 8; i++)
332 			printf(" %lx", p[i]);
333 		printf("\n");
334 		/* wait for debugger */
335 		panic("layer_checkvp");
336 	}
337 	if (a->layer_lowervp->v_usecount < 1) {
338 		int i; u_long *p;
339 		printf("vp = %p, unref'ed lowervp\n", vp);
340 		for (p = (u_long *) a, i = 0; i < 8; i++)
341 			printf(" %lx", p[i]);
342 		printf("\n");
343 		/* wait for debugger */
344 		panic ("layer with unref'ed lowervp");
345 	};
346 #ifdef notnow
347 	printf("layer %p/%d -> %p/%d [%s, %d]\n",
348 	        LAYERTOV(a), LAYERTOV(a)->v_usecount,
349 		a->layer_lowervp, a->layer_lowervp->v_usecount,
350 		fil, lno);
351 #endif
352 	return a->layer_lowervp;
353 }
354 #endif
355