layer_subr.c revision 1.22.6.2 1 /* $NetBSD: layer_subr.c,v 1.22.6.2 2007/12/06 21:03:38 ad 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 * Copyright (c) 1992, 1993
37 * The Regents of the University of California. All rights reserved.
38 *
39 * This code is derived from software donated to Berkeley by
40 * Jan-Simon Pendry.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * from: Id: lofs_subr.c,v 1.11 1992/05/30 10:05:43 jsp Exp
67 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95
68 */
69
70 #include <sys/cdefs.h>
71 __KERNEL_RCSID(0, "$NetBSD: layer_subr.c,v 1.22.6.2 2007/12/06 21:03:38 ad Exp $");
72
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/proc.h>
76 #include <sys/time.h>
77 #include <sys/vnode.h>
78 #include <sys/mount.h>
79 #include <sys/namei.h>
80 #include <sys/kmem.h>
81 #include <sys/malloc.h>
82
83 #include <miscfs/specfs/specdev.h>
84 #include <miscfs/genfs/layer.h>
85 #include <miscfs/genfs/layer_extern.h>
86
87 #define NLAYERNODECACHE 16
88
89 #ifdef LAYERFS_DIAGNOSTIC
90 int layerfs_debug = 1;
91 #endif
92
93 /*
94 * layer cache:
95 * Each cache entry holds a reference to the lower vnode
96 * along with a pointer to the alias vnode. When an
97 * entry is added the lower vnode is VREF'd. When the
98 * alias is removed the lower vnode is vrele'd.
99 */
100
101 /*
102 * Initialise cache headers
103 */
104 void
105 layerfs_init()
106 {
107 #ifdef LAYERFS_DIAGNOSTIC
108 if (layerfs_debug)
109 printf("layerfs_init\n"); /* printed during system boot */
110 #endif
111 }
112
113 /*
114 * Free global resources of layerfs.
115 */
116 void
117 layerfs_done()
118 {
119 #ifdef LAYERFS_DIAGNOSTIC
120 if (layerfs_debug)
121 printf("layerfs_done\n"); /* printed on layerfs detach */
122 #endif
123 }
124
125 /*
126 * Return a locked, VREF'ed alias for lower vnode if already exists, else NULL.
127 * The layermp's hashlock must be held on entry.
128 * It will be held upon return iff we return NULL.
129 */
130 struct vnode *
131 layer_node_find(mp, lowervp)
132 struct mount *mp;
133 struct vnode *lowervp;
134 {
135 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
136 struct layer_node_hashhead *hd;
137 struct layer_node *a;
138 struct vnode *vp;
139 int error;
140
141 /*
142 * Find hash base, and then search the (two-way) linked
143 * list looking for a layer_node structure which is referencing
144 * the lower vnode. If found, the increment the layer_node
145 * reference count (but NOT the lower vnode's VREF counter)
146 * and return the vnode locked.
147 */
148 hd = LAYER_NHASH(lmp, lowervp);
149 loop:
150 LIST_FOREACH(a, hd, layer_hash) {
151 if (a->layer_lowervp == lowervp && LAYERTOV(a)->v_mount == mp) {
152 vp = LAYERTOV(a);
153
154 /*
155 * We must not let vget() try to lock the layer vp,
156 * since the lower vp is already locked and locking the
157 * layer vp will involve locking the lower vp (whether
158 * or not they actually share a lock). Instead, take
159 * the layer vp's lock separately afterward, but only
160 * if it does not share the lower vp's lock.
161 */
162 mutex_enter(&vp->v_interlock);
163 mutex_exit(&lmp->layerm_hashlock);
164 error = vget(vp, LK_INTERLOCK);
165 if (error) {
166 mutex_enter(&lmp->layerm_hashlock);
167 goto loop;
168 }
169 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error);
170 return (vp);
171 }
172 }
173 return NULL;
174 }
175
176
177 /*
178 * Make a new layer_node node.
179 * Vp is the alias vnode, lowervp is the lower vnode.
180 * Maintain a reference to lowervp.
181 */
182 int
183 layer_node_alloc(mp, lowervp, vpp)
184 struct mount *mp;
185 struct vnode *lowervp;
186 struct vnode **vpp;
187 {
188 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
189 struct layer_node_hashhead *hd;
190 struct layer_node *xp;
191 struct vnode *vp, *nvp;
192 int error;
193 extern int (**dead_vnodeop_p)(void *);
194
195 error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, &vp);
196 if (error != 0)
197 return (error);
198 vp->v_type = lowervp->v_type;
199 mutex_enter(&vp->v_interlock);
200 vp->v_iflag |= VI_LAYER;
201 mutex_exit(&vp->v_interlock);
202
203 xp = kmem_alloc(lmp->layerm_size, KM_SLEEP);
204 if (xp == NULL) {
205 ungetnewvnode(vp);
206 return ENOMEM;
207 }
208 if (vp->v_type == VBLK || vp->v_type == VCHR) {
209 MALLOC(vp->v_specinfo, struct specinfo *,
210 sizeof(struct specinfo), M_VNODE, M_WAITOK);
211 vp->v_hashchain = NULL;
212 vp->v_rdev = lowervp->v_rdev;
213 }
214
215 vp->v_data = xp;
216 xp->layer_vnode = vp;
217 xp->layer_lowervp = lowervp;
218 xp->layer_flags = 0;
219
220 /*
221 * Before we insert our new node onto the hash chains,
222 * check to see if someone else has beaten us to it.
223 * (We could have slept in MALLOC.)
224 */
225 mutex_enter(&lmp->layerm_hashlock);
226 if ((nvp = layer_node_find(mp, lowervp)) != NULL) {
227 *vpp = nvp;
228
229 /* free the substructures we've allocated. */
230 kmem_free(xp, lmp->layerm_size);
231 if (vp->v_type == VBLK || vp->v_type == VCHR)
232 FREE(vp->v_specinfo, M_VNODE);
233
234 vp->v_type = VBAD; /* node is discarded */
235 vp->v_op = dead_vnodeop_p; /* so ops will still work */
236 vrele(vp); /* get rid of it. */
237 return (0);
238 }
239
240 /*
241 * Now lock the new node. We rely on the fact that we were passed
242 * a locked vnode. If the lower node is exporting a struct lock
243 * (v_vnlock != NULL) then we just set the upper v_vnlock to the
244 * lower one, and both are now locked. If the lower node is exporting
245 * NULL, then we copy that up and manually lock the upper node.
246 *
247 * LAYERFS_UPPERLOCK already has the test, so we use it after copying
248 * up the v_vnlock from below.
249 */
250
251 vp->v_vnlock = lowervp->v_vnlock;
252 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error);
253 KASSERT(error == 0);
254
255 /*
256 * Insert the new node into the hash.
257 * Add a reference to the lower node.
258 */
259
260 *vpp = vp;
261 VREF(lowervp);
262 hd = LAYER_NHASH(lmp, lowervp);
263 LIST_INSERT_HEAD(hd, xp, layer_hash);
264 uvm_vnp_setsize(vp, 0);
265 mutex_exit(&lmp->layerm_hashlock);
266 return (0);
267 }
268
269
270 /*
271 * Try to find an existing layer_node vnode refering
272 * to it, otherwise make a new layer_node vnode which
273 * contains a reference to the lower vnode.
274 *
275 * >>> we assume that the lower node is already locked upon entry, so we
276 * propagate the lock state to upper node <<
277 */
278 int
279 layer_node_create(mp, lowervp, newvpp)
280 struct mount *mp;
281 struct vnode *lowervp;
282 struct vnode **newvpp;
283 {
284 struct vnode *aliasvp;
285 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
286
287 mutex_enter(&lmp->layerm_hashlock);
288 aliasvp = layer_node_find(mp, lowervp);
289 if (aliasvp != NULL) {
290 /*
291 * layer_node_find has taken another reference
292 * to the alias vnode and moved the lock holding to
293 * aliasvp
294 */
295 #ifdef LAYERFS_DIAGNOSTIC
296 if (layerfs_debug)
297 vprint("layer_node_create: exists", aliasvp);
298 #endif
299 } else {
300 int error;
301
302 mutex_exit(&lmp->layerm_hashlock);
303
304 /*
305 * Get new vnode.
306 */
307 #ifdef LAYERFS_DIAGNOSTIC
308 if (layerfs_debug)
309 printf("layer_node_create: create new alias vnode\n");
310 #endif
311
312 /*
313 * Make new vnode reference the layer_node.
314 */
315 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0)
316 return error;
317
318 /*
319 * aliasvp is already VREF'd by getnewvnode()
320 */
321 }
322
323 /*
324 * Now that we have VREF'd the upper vnode, release the reference
325 * to the lower node. The existence of the layer_node retains one
326 * reference to the lower node.
327 */
328 vrele(lowervp);
329
330 #ifdef DIAGNOSTIC
331 if (lowervp->v_usecount < 1) {
332 /* Should never happen... */
333 vprint("layer_node_create: alias", aliasvp);
334 vprint("layer_node_create: lower", lowervp);
335 panic("layer_node_create: lower has 0 usecount.");
336 }
337 #endif
338
339 #ifdef LAYERFS_DIAGNOSTIC
340 if (layerfs_debug)
341 vprint("layer_node_create: alias", aliasvp);
342 #endif
343 *newvpp = aliasvp;
344 return (0);
345 }
346
347 #ifdef LAYERFS_DIAGNOSTIC
348 struct vnode *
349 layer_checkvp(vp, fil, lno)
350 struct vnode *vp;
351 const char *fil;
352 int lno;
353 {
354 struct layer_node *a = VTOLAYER(vp);
355 #ifdef notyet
356 /*
357 * Can't do this check because vop_reclaim runs
358 * with a funny vop vector.
359 *
360 * WRS - no it doesnt...
361 */
362 if (vp->v_op != layer_vnodeop_p) {
363 printf ("layer_checkvp: on non-layer-node\n");
364 #ifdef notyet
365 while (layer_checkvp_barrier) /*WAIT*/ ;
366 #endif
367 panic("layer_checkvp");
368 };
369 #endif
370 if (a->layer_lowervp == NULL) {
371 /* Should never happen */
372 int i; u_long *p;
373 printf("vp = %p, ZERO ptr\n", vp);
374 for (p = (u_long *) a, i = 0; i < 8; i++)
375 printf(" %lx", p[i]);
376 printf("\n");
377 /* wait for debugger */
378 panic("layer_checkvp");
379 }
380 if (a->layer_lowervp->v_usecount < 1) {
381 int i; u_long *p;
382 printf("vp = %p, unref'ed lowervp\n", vp);
383 for (p = (u_long *) a, i = 0; i < 8; i++)
384 printf(" %lx", p[i]);
385 printf("\n");
386 /* wait for debugger */
387 panic ("layer with unref'ed lowervp");
388 };
389 #ifdef notnow
390 printf("layer %p/%d -> %p/%d [%s, %d]\n",
391 LAYERTOV(a), LAYERTOV(a)->v_usecount,
392 a->layer_lowervp, a->layer_lowervp->v_usecount,
393 fil, lno);
394 #endif
395 return a->layer_lowervp;
396 }
397 #endif
398