1 /*
2 * Copyright (c) 2016 Apple Inc. All rights reserved.
3 *
4 * @APPLE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. Please obtain a copy of the License at
10 * http://www.opensource.apple.com/apsl/ and read it before using this
11 * file.
12 *
13 * The Original Code and all software distributed under the License are
14 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
15 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
16 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
18 * Please see the License for the specific language governing rights and
19 * limitations under the License.
20 *
21 * @APPLE_LICENSE_HEADER_END@
22 */
23
24 /*-
25 * Portions Copyright (c) 1992, 1993
26 * The Regents of the University of California. All rights reserved.
27 *
28 * This code is derived from software donated to Berkeley by
29 * Jan-Simon Pendry.
30 *
31 * Redistribution and use in source and binary forms, with or without
32 * modification, are permitted provided that the following conditions
33 * are met:
34 * 1. Redistributions of source code must retain the above copyright
35 * notice, this list of conditions and the following disclaimer.
36 * 2. Redistributions in binary form must reproduce the above copyright
37 * notice, this list of conditions and the following disclaimer in the
38 * documentation and/or other materials provided with the distribution.
39 * 4. Neither the name of the University nor the names of its contributors
40 * may be used to endorse or promote products derived from this software
41 * without specific prior written permission.
42 *
43 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
44 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
45 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
46 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
47 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
48 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
49 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
50 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
51 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
52 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * SUCH DAMAGE.
54 *
55 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95
56 *
57 * $FreeBSD$
58 */
59 #include <sys/param.h>
60 #include <sys/systm.h>
61 #include <sys/kernel.h>
62 #include <sys/lock.h>
63 #include <sys/malloc.h>
64 #include <sys/mount.h>
65 #include <sys/proc.h>
66 #include <sys/vnode.h>
67
68 #include "nullfs.h"
69
70 /*
71 * Null layer cache:
72 * Each cache entry holds a reference to the lower vnode
73 * along with a pointer to the alias vnode. When an
74 * entry is added the lower vnode is VREF'd. When the
75 * alias is removed the lower vnode is vrele'd.
76 */
77
78 #define NULL_HASH_SIZE (desiredvnodes / 10)
79
80 /* osx doesn't really have the functionality freebsd uses here..gonna try this
81 * hacked hash...*/
82 #define NULL_NHASH(vp) (&null_node_hashtbl[((((uintptr_t)vp) >> vnsz2log) + (uintptr_t)vnode_mount(vp)) & null_hash_mask])
83
84 static LIST_HEAD(null_node_hashhead, null_node) * null_node_hashtbl;
85 static LCK_GRP_DECLARE(null_hashlck_grp, "com.apple.filesystems.nullfs");
86 static LCK_MTX_DECLARE(null_hashmtx, &null_hashlck_grp);
87 static u_long null_hash_mask;
88
89 /* os x doesn't have hashes built into vnode. gonna try doing what freebsd does
90 * anyway
91 * Don't want to create a dependency on vnode_internal.h and the real struct
92 * vnode.
93 * 9 is an eyeball of the log 2 size of vnode */
94 static int vnsz2log = 9;
95
96 static int null_hashins(struct mount *, struct null_node *, struct vnode **);
97
98 void
nullfs_init_lck(lck_mtx_t * lck)99 nullfs_init_lck(lck_mtx_t * lck)
100 {
101 lck_mtx_init(lck, &null_hashlck_grp, LCK_ATTR_NULL);
102 }
103
104 void
nullfs_destroy_lck(lck_mtx_t * lck)105 nullfs_destroy_lck(lck_mtx_t * lck)
106 {
107 lck_mtx_destroy(lck, &null_hashlck_grp);
108 }
109
110 /*
111 * Initialise cache headers
112 */
113 int
nullfs_init(__unused struct vfsconf * vfsp)114 nullfs_init(__unused struct vfsconf * vfsp)
115 {
116 NULLFSDEBUG("%s\n", __FUNCTION__);
117 null_node_hashtbl = hashinit(NULL_HASH_SIZE, M_TEMP, &null_hash_mask);
118 NULLFSDEBUG("%s finished\n", __FUNCTION__);
119 return 0;
120 }
121
122 int
nullfs_uninit(void)123 nullfs_uninit(void)
124 {
125 /* This gets called when the fs is uninstalled, there wasn't an exact
126 * equivalent in vfsops */
127 hashdestroy(null_node_hashtbl, M_TEMP, null_hash_mask);
128 return 0;
129 }
130
131 /*
132 * Find the nullfs vnode mapped to lowervp. Return it in *vpp with an iocount if found.
133 * Return 0 on success. On failure *vpp will be null and a non-zero error code will be returned.
134 */
135 int
null_hashget(struct mount * mp,struct vnode * lowervp,struct vnode ** vpp)136 null_hashget(struct mount * mp, struct vnode * lowervp, struct vnode ** vpp)
137 {
138 struct null_node_hashhead * hd = NULL;
139 struct null_node * a = NULL;
140 struct vnode * vp = NULL;
141 uint32_t vp_vid = 0;
142 int error = ENOENT;
143
144 /*
145 * Find hash base, and then search the (two-way) linked
146 * list looking for a null_node structure which is referencing
147 * the lower vnode. We only give up our reference at reclaim so
148 * just check whether the lowervp has gotten pulled from under us
149 */
150 hd = NULL_NHASH(lowervp);
151 // In the future we should consider using a per bucket lock
152 lck_mtx_lock(&null_hashmtx);
153 LIST_FOREACH(a, hd, null_hash)
154 {
155 if (a->null_lowervp == lowervp && vnode_mount(NULLTOV(a)) == mp) {
156 vp = NULLTOV(a);
157 if (a->null_lowervid != vnode_vid(lowervp)) {
158 /*lowervp has reved */
159 error = EIO;
160 vp = NULL;
161 } else {
162 vp_vid = a->null_myvid;
163 }
164 // In the case of a succesful look-up we should consider moving the object to the top of the head
165 break;
166 }
167 }
168 if (vp != NULL) {
169 vnode_hold(vp);
170 }
171 lck_mtx_unlock(&null_hashmtx);
172 if (vp != NULL) {
173 error = vnode_getwithvid(vp, vp_vid);
174 if (error == 0) {
175 *vpp = vp;
176 }
177 vnode_drop(vp);
178 }
179 return error;
180 }
181
182 /*
183 * Act like null_hashget, but add passed null_node to hash if no existing
184 * node found.
185 */
186 static int
null_hashins(struct mount * mp,struct null_node * xp,struct vnode ** vpp)187 null_hashins(struct mount * mp, struct null_node * xp, struct vnode ** vpp)
188 {
189 struct null_node_hashhead * hd = NULL;
190 struct null_node * oxp = NULL;
191 struct vnode * ovp = NULL;
192 uint32_t oxp_vid = 0;
193 int error = 0;
194
195 hd = NULL_NHASH(xp->null_lowervp);
196 lck_mtx_lock(&null_hashmtx);
197 LIST_FOREACH(oxp, hd, null_hash)
198 {
199 if (oxp->null_lowervp == xp->null_lowervp && vnode_mount(NULLTOV(oxp)) == mp) {
200 /*
201 * See null_hashget for a description of this
202 * operation.
203 */
204 ovp = NULLTOV(oxp);
205 if (oxp->null_lowervid != vnode_vid(oxp->null_lowervp)) {
206 /*vp doesn't exist so return null (not sure we are actually gonna catch
207 * recycle right now
208 * This is an exceptional case right now, it suggests the vnode we are
209 * trying to add has been recycled
210 * don't add it.*/
211 error = EIO;
212 ovp = NULL;
213 } else {
214 oxp_vid = oxp->null_myvid;
215 }
216 goto end;
217 }
218 }
219 /* if it wasn't in the hash map then the vnode pointed to by xp already has a
220 * iocount so don't bother */
221 LIST_INSERT_HEAD(hd, xp, null_hash);
222 xp->null_flags |= NULL_FLAG_HASHED;
223 end:
224 if (ovp != NULL) {
225 vnode_hold(ovp);
226 }
227 lck_mtx_unlock(&null_hashmtx);
228 if (ovp != NULL) {
229 /* if we found something in the hash map then grab an iocount */
230 error = vnode_getwithvid(ovp, oxp_vid);
231 if (error == 0) {
232 *vpp = ovp;
233 }
234 vnode_drop(ovp);
235 }
236 return error;
237 }
238
239 /*
240 * Remove node from hash.
241 */
242 void
null_hashrem(struct null_node * xp)243 null_hashrem(struct null_node * xp)
244 {
245 lck_mtx_lock(&null_hashmtx);
246 LIST_REMOVE(xp, null_hash);
247 lck_mtx_unlock(&null_hashmtx);
248 }
249
250 static struct null_node *
null_nodecreate(struct vnode * lowervp)251 null_nodecreate(struct vnode * lowervp)
252 {
253 struct null_node * xp;
254
255 xp = kalloc_type(struct null_node, Z_WAITOK | Z_ZERO | Z_NOFAIL);
256 if (lowervp) {
257 xp->null_lowervp = lowervp;
258 xp->null_lowervid = vnode_vid(lowervp);
259 }
260 return xp;
261 }
262
263 /* assumption is that vnode has iocount on it after vnode create */
264 int
null_getnewvnode(struct mount * mp,struct vnode * lowervp,struct vnode * dvp,struct vnode ** vpp,struct componentname * cnp,int root)265 null_getnewvnode(
266 struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
267 {
268 struct vnode_fsparam vnfs_param;
269 int error = 0;
270 enum vtype type = VDIR;
271 struct null_node * xp = null_nodecreate(lowervp);
272
273 if (xp == NULL) {
274 return ENOMEM;
275 }
276
277 if (lowervp) {
278 type = vnode_vtype(lowervp);
279 }
280
281 vnfs_param.vnfs_mp = mp;
282 vnfs_param.vnfs_vtype = type;
283 vnfs_param.vnfs_str = "nullfs";
284 vnfs_param.vnfs_dvp = dvp;
285 vnfs_param.vnfs_fsnode = (void *)xp;
286 vnfs_param.vnfs_vops = nullfs_vnodeop_p;
287 vnfs_param.vnfs_markroot = root;
288 vnfs_param.vnfs_marksystem = 0;
289 vnfs_param.vnfs_rdev = 0;
290 vnfs_param.vnfs_filesize = 0; // set this to 0 since we should only be shadowing non-regular files
291 vnfs_param.vnfs_cnp = cnp;
292 vnfs_param.vnfs_flags = VNFS_ADDFSREF;
293
294 error = vnode_create_ext(VNCREATE_FLAVOR, VCREATESIZE, &vnfs_param, vpp, VNODE_CREATE_DEFAULT);
295 if (error == 0) {
296 xp->null_vnode = *vpp;
297 xp->null_myvid = vnode_vid(*vpp);
298 vnode_settag(*vpp, VT_NULL);
299 } else {
300 kfree_type(struct null_node, xp);
301 }
302 return error;
303 }
304
305 /*
306 * Make a new or get existing nullfs node.
307 * Vp is the alias vnode, lowervp is the lower vnode.
308 *
309 * lowervp is assumed to have an iocount on it from the caller
310 */
311 int
null_nodeget(struct mount * mp,struct vnode * lowervp,struct vnode * dvp,struct vnode ** vpp,struct componentname * cnp,int root)312 null_nodeget(
313 struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
314 {
315 struct vnode * vp;
316 int error;
317
318 /* Lookup the hash firstly. */
319 error = null_hashget(mp, lowervp, vpp);
320 /* ENOENT means it wasn't found, EIO is a failure we should bail from, 0 is it
321 * was found */
322 if (error != ENOENT) {
323 /* null_hashget checked the vid, so if we got something here its legit to
324 * the best of our knowledge*/
325 /* if we found something then there is an iocount on vpp,
326 * if we didn't find something then vpp shouldn't be used by the caller */
327 return error;
328 }
329
330 /*
331 * We do not serialize vnode creation, instead we will check for
332 * duplicates later, when adding new vnode to hash.
333 */
334 error = vnode_ref(lowervp); // take a ref on lowervp so we let the system know we care about it
335 if (error) {
336 // Failed to get a reference on the lower vp so bail. Lowervp may be gone already.
337 return error;
338 }
339
340 error = null_getnewvnode(mp, lowervp, dvp, &vp, cnp, root);
341
342 if (error) {
343 vnode_rele(lowervp);
344 return error;
345 }
346
347 /*
348 * Atomically insert our new node into the hash or vget existing
349 * if someone else has beaten us to it.
350 */
351 error = null_hashins(mp, VTONULL(vp), vpp);
352 if (error || *vpp != NULL) {
353 /* recycle will call reclaim which will get rid of the internals */
354 vnode_recycle(vp);
355 vnode_put(vp);
356 /* if we found vpp, then null_hashins put an iocount on it */
357 return error;
358 }
359
360 /* vp has an iocount from null_getnewvnode */
361 *vpp = vp;
362
363 return 0;
364 }
365