1 /*        $NetBSD: layer_vfsops.c,v 1.56 2022/12/09 10:33:18 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, 1995
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_vfsops.c,v 1.9 1992/05/30 10:26:24 jsp Exp
68  *        from: @(#)lofs_vfsops.c       1.2 (Berkeley) 6/18/92
69  *        @(#)null_vfsops.c   8.7 (Berkeley) 5/14/95
70  */
71 
72 /*
73  * Generic layer VFS operations.
74  */
75 
76 #include <sys/cdefs.h>
77 __KERNEL_RCSID(0, "$NetBSD: layer_vfsops.c,v 1.56 2022/12/09 10:33:18 hannken Exp $");
78 
79 #include <sys/param.h>
80 #include <sys/sysctl.h>
81 #include <sys/systm.h>
82 #include <sys/vnode.h>
83 #include <sys/mount.h>
84 #include <sys/namei.h>
85 #include <sys/kauth.h>
86 #include <sys/module.h>
87 
88 #include <miscfs/specfs/specdev.h>
89 #include <miscfs/genfs/genfs.h>
90 #include <miscfs/genfs/layer.h>
91 #include <miscfs/genfs/layer_extern.h>
92 
93 SYSCTL_SETUP_PROTO(sysctl_vfs_layerfs_setup);
94 
95 MODULE(MODULE_CLASS_MISC, layerfs, NULL);
96 
97 static int
layerfs_modcmd(modcmd_t cmd,void * arg)98 layerfs_modcmd(modcmd_t cmd, void *arg)
99 {
100 
101           switch (cmd) {
102           case MODULE_CMD_INIT:
103                     return 0;
104           case MODULE_CMD_FINI:
105                     return 0;
106           default:
107                     return ENOTTY;
108           }
109           return 0;
110 }
111 
112 /*
113  * VFS start.  Nothing needed here - the start routine on the underlying
114  * filesystem will have been called when that filesystem was mounted.
115  */
116 int
layerfs_start(struct mount * mp,int flags)117 layerfs_start(struct mount *mp, int flags)
118 {
119 
120 #ifdef notyet
121           return VFS_START(mp->mnt_lower, flags);
122 #else
123           return 0;
124 #endif
125 }
126 
127 int
layerfs_root(struct mount * mp,int lktype,struct vnode ** vpp)128 layerfs_root(struct mount *mp, int lktype, struct vnode **vpp)
129 {
130           struct vnode *vp;
131 
132           vp = MOUNTTOLAYERMOUNT(mp)->layerm_rootvp;
133           if (vp == NULL) {
134                     *vpp = NULL;
135                     return EINVAL;
136           }
137           /*
138            * Return root vnode with locked and with a reference held.
139            */
140           vref(vp);
141           vn_lock(vp, lktype | LK_RETRY);
142           *vpp = vp;
143           return 0;
144 }
145 
146 int
layerfs_quotactl(struct mount * mp,struct quotactl_args * args)147 layerfs_quotactl(struct mount *mp, struct quotactl_args *args)
148 {
149           int error;
150 
151           error = vfs_busy(mp);
152           if (error == 0) {
153                     error = VFS_QUOTACTL(mp->mnt_lower, args);
154                     vfs_unbusy(mp);
155           }
156 
157           return error;
158 }
159 
160 int
layerfs_statvfs(struct mount * mp,struct statvfs * sbp)161 layerfs_statvfs(struct mount *mp, struct statvfs *sbp)
162 {
163           struct statvfs *sbuf;
164           int error;
165 
166           sbuf = kmem_zalloc(sizeof(*sbuf), KM_SLEEP);
167           error = vfs_busy(mp);
168           if (error == 0) {
169                     error = VFS_STATVFS(mp->mnt_lower, sbuf);
170                     vfs_unbusy(mp);
171           }
172           if (error) {
173                     goto done;
174           }
175           /* Copy across the relevant data and fake the rest. */
176           sbp->f_flag = sbuf->f_flag;
177           sbp->f_bsize = sbuf->f_bsize;
178           sbp->f_frsize = sbuf->f_frsize;
179           sbp->f_iosize = sbuf->f_iosize;
180           sbp->f_blocks = sbuf->f_blocks;
181           sbp->f_bfree = sbuf->f_bfree;
182           sbp->f_bavail = sbuf->f_bavail;
183           sbp->f_bresvd = sbuf->f_bresvd;
184           sbp->f_files = sbuf->f_files;
185           sbp->f_ffree = sbuf->f_ffree;
186           sbp->f_favail = sbuf->f_favail;
187           sbp->f_fresvd = sbuf->f_fresvd;
188           sbp->f_namemax = sbuf->f_namemax;
189           copy_statvfs_info(sbp, mp);
190 done:
191           kmem_free(sbuf, sizeof(*sbuf));
192           return error;
193 }
194 
195 int
layerfs_sync(struct mount * mp,int waitfor,kauth_cred_t cred)196 layerfs_sync(struct mount *mp, int waitfor,
197     kauth_cred_t cred)
198 {
199 
200           /*
201            * XXX - Assumes no data cached at layer.
202            */
203           return 0;
204 }
205 
206 int
layerfs_loadvnode(struct mount * mp,struct vnode * vp,const void * key,size_t key_len,const void ** new_key)207 layerfs_loadvnode(struct mount *mp, struct vnode *vp,
208     const void *key, size_t key_len, const void **new_key)
209 {
210           struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
211           struct vnode *lowervp;
212           struct layer_node *xp;
213 
214           KASSERT(key_len == sizeof(struct vnode *));
215           memcpy(&lowervp, key, key_len);
216 
217           xp = kmem_alloc(lmp->layerm_size, KM_SLEEP);
218 
219           /* Share the interlock, vmobjlock, and klist with the lower node. */
220           vshareilock(vp, lowervp);
221           rw_obj_hold(lowervp->v_uobj.vmobjlock);
222           uvm_obj_setlock(&vp->v_uobj, lowervp->v_uobj.vmobjlock);
223           vshareklist(vp, lowervp);
224 
225           vp->v_tag = lmp->layerm_tag;
226           vp->v_type = lowervp->v_type;
227           vp->v_op = lmp->layerm_vnodeop_p;
228           if (vp->v_type == VBLK || vp->v_type == VCHR)
229                     spec_node_init(vp, lowervp->v_rdev);
230           vp->v_data = xp;
231           xp->layer_vnode = vp;
232           xp->layer_lowervp = lowervp;
233           xp->layer_flags = 0;
234           uvm_vnp_setsize(vp, 0);
235 
236           /*  Add a reference to the lower node. */
237           vref(lowervp);
238           *new_key = &xp->layer_lowervp;
239           return 0;
240 }
241 
242 int
layerfs_vget(struct mount * mp,ino_t ino,int lktype,struct vnode ** vpp)243 layerfs_vget(struct mount *mp, ino_t ino, int lktype, struct vnode **vpp)
244 {
245           struct vnode *vp;
246           int error;
247 
248           error = vfs_busy(mp);
249           if (error == 0) {
250                     error = VFS_VGET(mp->mnt_lower, ino, lktype, &vp);
251                     vfs_unbusy(mp);
252           }
253           if (error) {
254                     *vpp = NULL;
255                     return error;
256           }
257           VOP_UNLOCK(vp);
258           error = layer_node_create(mp, vp, vpp);
259           if (error) {
260                     vrele(vp);
261                     *vpp = NULL;
262                     return error;
263           }
264           error = vn_lock(*vpp, lktype);
265           if (error) {
266                     vrele(*vpp);
267                     *vpp = NULL;
268                     return error;
269           }
270           return 0;
271 }
272 
273 int
layerfs_fhtovp(struct mount * mp,struct fid * fidp,int lktype,struct vnode ** vpp)274 layerfs_fhtovp(struct mount *mp, struct fid *fidp, int lktype,
275     struct vnode **vpp)
276 {
277           struct vnode *vp;
278           int error;
279 
280           error = vfs_busy(mp);
281           if (error == 0) {
282                     error = VFS_FHTOVP(mp->mnt_lower, fidp, lktype, &vp);
283                     vfs_unbusy(mp);
284           }
285           if (error) {
286                     *vpp = NULL;
287                     return error;
288           }
289           VOP_UNLOCK(vp);
290           error = layer_node_create(mp, vp, vpp);
291           if (error) {
292                     vput(vp);
293                     *vpp = NULL;
294                     return (error);
295           }
296           error = vn_lock(*vpp, lktype);
297           if (error) {
298                     vrele(*vpp);
299                     *vpp = NULL;
300                     return error;
301           }
302           return 0;
303 }
304 
305 int
layerfs_vptofh(struct vnode * vp,struct fid * fhp,size_t * fh_size)306 layerfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
307 {
308 
309           return VFS_VPTOFH(LAYERVPTOLOWERVP(vp), fhp, fh_size);
310 }
311 
312 /*
313  * layerfs_snapshot - handle a snapshot through a layered file system
314  *
315  * At present, we do NOT support snapshotting through a layered file
316  * system as the ffs implementation changes v_vnlock of the snapshot
317  * vnodes to point to one common lock. As there is no way for us to
318  * absolutely pass this change up the stack, a layered file system
319  * would end up referencing the wrong lock.
320  *
321  * This routine serves as a central resource for this behavior; all
322  * layered file systems don't need to worry about the above. Also, if
323  * things get fixed, all layers get the benefit.
324  */
325 int
layerfs_snapshot(struct mount * mp,struct vnode * vp,struct timespec * ts)326 layerfs_snapshot(struct mount *mp, struct vnode *vp,
327     struct timespec *ts)
328 {
329 
330           return EOPNOTSUPP;
331 }
332 
333 /*
334  * layerfs_suspendctl - suspend a layered file system
335  *
336  * Here we should suspend the lower file system(s) too.  At present
337  * this will deadlock as we don't know which to suspend first.
338  *
339  * This routine serves as a central resource for this behavior; all
340  * layered file systems don't need to worry about the above. Also, if
341  * things get fixed, all layers get the benefit.
342  */
343 int
layerfs_suspendctl(struct mount * mp,int cmd)344 layerfs_suspendctl(struct mount *mp, int cmd)
345 {
346 
347           return genfs_suspendctl(mp, cmd);
348 }
349 
350 SYSCTL_SETUP(sysctl_vfs_layerfs_setup, "sysctl vfs.layerfs subtree setup")
351 {
352           const struct sysctlnode *layerfs_node = NULL;
353 
354           sysctl_createv(clog, 0, NULL, &layerfs_node,
355 #ifdef _MODULE
356                            0,
357 #else
358                            CTLFLAG_PERMANENT,
359 #endif
360                            CTLTYPE_NODE, "layerfs",
361                            SYSCTL_DESCR("Generic layered file system"),
362                            NULL, 0, NULL, 0,
363                            CTL_VFS, CTL_CREATE, CTL_EOL);
364 
365 #ifdef LAYERFS_DIAGNOSTIC
366           sysctl_createv(clog, 0, &layerfs_node, NULL,
367 #ifndef _MODULE
368                          CTLFLAG_PERMANENT |
369 #endif
370                            CTLFLAG_READWRITE,
371                          CTLTYPE_INT,
372                          "debug",
373                          SYSCTL_DESCR("Verbose debugging messages"),
374                          NULL, 0, &layerfs_debug, 0,
375                          CTL_CREATE, CTL_EOL);
376 #endif
377 
378           /*
379            * other subtrees should really be aliases to this, but since
380            * they can't tell if layerfs has been instantiated yet, they
381            * can't do that...not easily.  not yet.  :-)
382            */
383 }
384 
385 int
layerfs_renamelock_enter(struct mount * mp)386 layerfs_renamelock_enter(struct mount *mp)
387 {
388 
389           return VFS_RENAMELOCK_ENTER(mp->mnt_lower);
390 }
391 
392 void
layerfs_renamelock_exit(struct mount * mp)393 layerfs_renamelock_exit(struct mount *mp)
394 {
395 
396           VFS_RENAMELOCK_EXIT(mp->mnt_lower);
397 }
398