xref: /freebsd/sys/fs/nullfs/null_vfsops.c (revision 792bbaba989533a1fc93823df1720c8c4aaf0442)
1 /*-
2  * Copyright (c) 1992, 1993, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software donated to Berkeley by
6  * Jan-Simon Pendry.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)null_vfsops.c	8.2 (Berkeley) 1/21/94
33  *
34  * @(#)lofs_vfsops.c	1.2 (Berkeley) 6/18/92
35  * $FreeBSD$
36  */
37 
38 /*
39  * Null Layer
40  * (See null_vnops.c for a description of what this does.)
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/fcntl.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/namei.h>
51 #include <sys/proc.h>
52 #include <sys/vnode.h>
53 #include <sys/jail.h>
54 
55 #include <fs/nullfs/null.h>
56 
57 static MALLOC_DEFINE(M_NULLFSMNT, "nullfs_mount", "NULLFS mount structure");
58 
59 static vfs_fhtovp_t	nullfs_fhtovp;
60 static vfs_mount_t	nullfs_mount;
61 static vfs_quotactl_t	nullfs_quotactl;
62 static vfs_root_t	nullfs_root;
63 static vfs_sync_t	nullfs_sync;
64 static vfs_statfs_t	nullfs_statfs;
65 static vfs_unmount_t	nullfs_unmount;
66 static vfs_vget_t	nullfs_vget;
67 static vfs_extattrctl_t	nullfs_extattrctl;
68 
69 /*
70  * Mount null layer
71  */
72 static int
73 nullfs_mount(struct mount *mp)
74 {
75 	int error = 0;
76 	struct vnode *lowerrootvp, *vp;
77 	struct vnode *nullm_rootvp;
78 	struct null_mount *xmp;
79 	struct thread *td = curthread;
80 	char *target;
81 	int isvnunlocked = 0, len;
82 	struct nameidata nd, *ndp = &nd;
83 
84 	NULLFSDEBUG("nullfs_mount(mp = %p)\n", (void *)mp);
85 
86 	if (!prison_allow(td->td_ucred, PR_ALLOW_MOUNT_NULLFS))
87 		return (EPERM);
88 	if (mp->mnt_flag & MNT_ROOTFS)
89 		return (EOPNOTSUPP);
90 
91 	/*
92 	 * Update is a no-op
93 	 */
94 	if (mp->mnt_flag & MNT_UPDATE) {
95 		/*
96 		 * Only support update mounts for NFS export.
97 		 */
98 		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0))
99 			return (0);
100 		else
101 			return (EOPNOTSUPP);
102 	}
103 
104 	/*
105 	 * Get argument
106 	 */
107 	error = vfs_getopt(mp->mnt_optnew, "target", (void **)&target, &len);
108 	if (error || target[len - 1] != '\0')
109 		return (EINVAL);
110 
111 	/*
112 	 * Unlock lower node to avoid possible deadlock.
113 	 */
114 	if ((mp->mnt_vnodecovered->v_op == &null_vnodeops) &&
115 	    VOP_ISLOCKED(mp->mnt_vnodecovered) == LK_EXCLUSIVE) {
116 		VOP_UNLOCK(mp->mnt_vnodecovered, 0);
117 		isvnunlocked = 1;
118 	}
119 	/*
120 	 * Find lower node
121 	 */
122 	NDINIT(ndp, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, target, curthread);
123 	error = namei(ndp);
124 
125 	/*
126 	 * Re-lock vnode.
127 	 * XXXKIB This is deadlock-prone as well.
128 	 */
129 	if (isvnunlocked)
130 		vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY);
131 
132 	if (error)
133 		return (error);
134 	NDFREE(ndp, NDF_ONLY_PNBUF);
135 
136 	/*
137 	 * Sanity check on lower vnode
138 	 */
139 	lowerrootvp = ndp->ni_vp;
140 
141 	/*
142 	 * Check multi null mount to avoid `lock against myself' panic.
143 	 */
144 	if (lowerrootvp == VTONULL(mp->mnt_vnodecovered)->null_lowervp) {
145 		NULLFSDEBUG("nullfs_mount: multi null mount?\n");
146 		vput(lowerrootvp);
147 		return (EDEADLK);
148 	}
149 
150 	xmp = (struct null_mount *) malloc(sizeof(struct null_mount),
151 	    M_NULLFSMNT, M_WAITOK | M_ZERO);
152 
153 	/*
154 	 * Save reference to underlying FS
155 	 */
156 	xmp->nullm_vfs = lowerrootvp->v_mount;
157 
158 	/*
159 	 * Save reference.  Each mount also holds
160 	 * a reference on the root vnode.
161 	 */
162 	error = null_nodeget(mp, lowerrootvp, &vp);
163 	/*
164 	 * Make sure the node alias worked
165 	 */
166 	if (error) {
167 		free(xmp, M_NULLFSMNT);
168 		return (error);
169 	}
170 
171 	/*
172 	 * Keep a held reference to the root vnode.
173 	 * It is vrele'd in nullfs_unmount.
174 	 */
175 	nullm_rootvp = vp;
176 	nullm_rootvp->v_vflag |= VV_ROOT;
177 	xmp->nullm_rootvp = nullm_rootvp;
178 
179 	/*
180 	 * Unlock the node (either the lower or the alias)
181 	 */
182 	VOP_UNLOCK(vp, 0);
183 
184 	if (NULLVPTOLOWERVP(nullm_rootvp)->v_mount->mnt_flag & MNT_LOCAL) {
185 		MNT_ILOCK(mp);
186 		mp->mnt_flag |= MNT_LOCAL;
187 		MNT_IUNLOCK(mp);
188 	}
189 
190 	xmp->nullm_flags |= NULLM_CACHE;
191 	if (vfs_getopt(mp->mnt_optnew, "nocache", NULL, NULL) == 0 ||
192 	    (xmp->nullm_vfs->mnt_kern_flag & MNTK_NULL_NOCACHE) != 0)
193 		xmp->nullm_flags &= ~NULLM_CACHE;
194 
195 	MNT_ILOCK(mp);
196 	if ((xmp->nullm_flags & NULLM_CACHE) != 0) {
197 		mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag &
198 		    (MNTK_SHARED_WRITES | MNTK_LOOKUP_SHARED |
199 		    MNTK_EXTENDED_SHARED);
200 	}
201 	mp->mnt_kern_flag |= MNTK_LOOKUP_EXCL_DOTDOT;
202 	mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag &
203 	    (MNTK_USES_BCACHE | MNTK_NO_IOPF | MNTK_UNMAPPED_BUFS);
204 	MNT_IUNLOCK(mp);
205 	mp->mnt_data = xmp;
206 	vfs_getnewfsid(mp);
207 	if ((xmp->nullm_flags & NULLM_CACHE) != 0) {
208 		MNT_ILOCK(xmp->nullm_vfs);
209 		TAILQ_INSERT_TAIL(&xmp->nullm_vfs->mnt_uppers, mp,
210 		    mnt_upper_link);
211 		MNT_IUNLOCK(xmp->nullm_vfs);
212 	}
213 
214 	vfs_mountedfrom(mp, target);
215 
216 	NULLFSDEBUG("nullfs_mount: lower %s, alias at %s\n",
217 		mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
218 	return (0);
219 }
220 
221 /*
222  * Free reference to null layer
223  */
224 static int
225 nullfs_unmount(mp, mntflags)
226 	struct mount *mp;
227 	int mntflags;
228 {
229 	struct null_mount *mntdata;
230 	struct mount *ump;
231 	int error, flags;
232 
233 	NULLFSDEBUG("nullfs_unmount: mp = %p\n", (void *)mp);
234 
235 	if (mntflags & MNT_FORCE)
236 		flags = FORCECLOSE;
237 	else
238 		flags = 0;
239 
240 	/* There is 1 extra root vnode reference (nullm_rootvp). */
241 	error = vflush(mp, 1, flags, curthread);
242 	if (error)
243 		return (error);
244 
245 	/*
246 	 * Finally, throw away the null_mount structure
247 	 */
248 	mntdata = mp->mnt_data;
249 	ump = mntdata->nullm_vfs;
250 	if ((mntdata->nullm_flags & NULLM_CACHE) != 0) {
251 		MNT_ILOCK(ump);
252 		while ((ump->mnt_kern_flag & MNTK_VGONE_UPPER) != 0) {
253 			ump->mnt_kern_flag |= MNTK_VGONE_WAITER;
254 			msleep(&ump->mnt_uppers, &ump->mnt_mtx, 0, "vgnupw", 0);
255 		}
256 		TAILQ_REMOVE(&ump->mnt_uppers, mp, mnt_upper_link);
257 		MNT_IUNLOCK(ump);
258 	}
259 	mp->mnt_data = NULL;
260 	free(mntdata, M_NULLFSMNT);
261 	return (0);
262 }
263 
264 static int
265 nullfs_root(mp, flags, vpp)
266 	struct mount *mp;
267 	int flags;
268 	struct vnode **vpp;
269 {
270 	struct vnode *vp;
271 
272 	NULLFSDEBUG("nullfs_root(mp = %p, vp = %p->%p)\n", (void *)mp,
273 	    (void *)MOUNTTONULLMOUNT(mp)->nullm_rootvp,
274 	    (void *)NULLVPTOLOWERVP(MOUNTTONULLMOUNT(mp)->nullm_rootvp));
275 
276 	/*
277 	 * Return locked reference to root.
278 	 */
279 	vp = MOUNTTONULLMOUNT(mp)->nullm_rootvp;
280 	VREF(vp);
281 
282 	ASSERT_VOP_UNLOCKED(vp, "root vnode is locked");
283 	vn_lock(vp, flags | LK_RETRY);
284 	*vpp = vp;
285 	return 0;
286 }
287 
288 static int
289 nullfs_quotactl(mp, cmd, uid, arg)
290 	struct mount *mp;
291 	int cmd;
292 	uid_t uid;
293 	void *arg;
294 {
295 	return VFS_QUOTACTL(MOUNTTONULLMOUNT(mp)->nullm_vfs, cmd, uid, arg);
296 }
297 
298 static int
299 nullfs_statfs(mp, sbp)
300 	struct mount *mp;
301 	struct statfs *sbp;
302 {
303 	int error;
304 	struct statfs *mstat;
305 
306 	NULLFSDEBUG("nullfs_statfs(mp = %p, vp = %p->%p)\n", (void *)mp,
307 	    (void *)MOUNTTONULLMOUNT(mp)->nullm_rootvp,
308 	    (void *)NULLVPTOLOWERVP(MOUNTTONULLMOUNT(mp)->nullm_rootvp));
309 
310 	mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO);
311 
312 	error = VFS_STATFS(MOUNTTONULLMOUNT(mp)->nullm_vfs, mstat);
313 	if (error) {
314 		free(mstat, M_STATFS);
315 		return (error);
316 	}
317 
318 	/* now copy across the "interesting" information and fake the rest */
319 	sbp->f_type = mstat->f_type;
320 	sbp->f_flags = (sbp->f_flags & (MNT_RDONLY | MNT_NOEXEC | MNT_NOSUID |
321 	    MNT_UNION | MNT_NOSYMFOLLOW)) | (mstat->f_flags & ~MNT_ROOTFS);
322 	sbp->f_bsize = mstat->f_bsize;
323 	sbp->f_iosize = mstat->f_iosize;
324 	sbp->f_blocks = mstat->f_blocks;
325 	sbp->f_bfree = mstat->f_bfree;
326 	sbp->f_bavail = mstat->f_bavail;
327 	sbp->f_files = mstat->f_files;
328 	sbp->f_ffree = mstat->f_ffree;
329 
330 	free(mstat, M_STATFS);
331 	return (0);
332 }
333 
334 static int
335 nullfs_sync(mp, waitfor)
336 	struct mount *mp;
337 	int waitfor;
338 {
339 	/*
340 	 * XXX - Assumes no data cached at null layer.
341 	 */
342 	return (0);
343 }
344 
345 static int
346 nullfs_vget(mp, ino, flags, vpp)
347 	struct mount *mp;
348 	ino_t ino;
349 	int flags;
350 	struct vnode **vpp;
351 {
352 	int error;
353 
354 	KASSERT((flags & LK_TYPE_MASK) != 0,
355 	    ("nullfs_vget: no lock requested"));
356 
357 	error = VFS_VGET(MOUNTTONULLMOUNT(mp)->nullm_vfs, ino, flags, vpp);
358 	if (error != 0)
359 		return (error);
360 	return (null_nodeget(mp, *vpp, vpp));
361 }
362 
363 static int
364 nullfs_fhtovp(mp, fidp, flags, vpp)
365 	struct mount *mp;
366 	struct fid *fidp;
367 	int flags;
368 	struct vnode **vpp;
369 {
370 	int error;
371 
372 	error = VFS_FHTOVP(MOUNTTONULLMOUNT(mp)->nullm_vfs, fidp, flags,
373 	    vpp);
374 	if (error != 0)
375 		return (error);
376 	return (null_nodeget(mp, *vpp, vpp));
377 }
378 
379 static int
380 nullfs_extattrctl(mp, cmd, filename_vp, namespace, attrname)
381 	struct mount *mp;
382 	int cmd;
383 	struct vnode *filename_vp;
384 	int namespace;
385 	const char *attrname;
386 {
387 
388 	return (VFS_EXTATTRCTL(MOUNTTONULLMOUNT(mp)->nullm_vfs, cmd,
389 	    filename_vp, namespace, attrname));
390 }
391 
392 static void
393 nullfs_reclaim_lowervp(struct mount *mp, struct vnode *lowervp)
394 {
395 	struct vnode *vp;
396 
397 	vp = null_hashget(mp, lowervp);
398 	if (vp == NULL)
399 		return;
400 	VTONULL(vp)->null_flags |= NULLV_NOUNLOCK;
401 	vgone(vp);
402 	vput(vp);
403 }
404 
405 static void
406 nullfs_unlink_lowervp(struct mount *mp, struct vnode *lowervp)
407 {
408 	struct vnode *vp;
409 	struct null_node *xp;
410 
411 	vp = null_hashget(mp, lowervp);
412 	if (vp == NULL)
413 		return;
414 	xp = VTONULL(vp);
415 	xp->null_flags |= NULLV_DROP | NULLV_NOUNLOCK;
416 	vhold(vp);
417 	vunref(vp);
418 
419 	if (vp->v_usecount == 0) {
420 		/*
421 		 * If vunref() dropped the last use reference on the
422 		 * nullfs vnode, it must be reclaimed, and its lock
423 		 * was split from the lower vnode lock.  Need to do
424 		 * extra unlock before allowing the final vdrop() to
425 		 * free the vnode.
426 		 */
427 		KASSERT((vp->v_iflag & VI_DOOMED) != 0,
428 		    ("not reclaimed nullfs vnode %p", vp));
429 		VOP_UNLOCK(vp, 0);
430 	} else {
431 		/*
432 		 * Otherwise, the nullfs vnode still shares the lock
433 		 * with the lower vnode, and must not be unlocked.
434 		 * Also clear the NULLV_NOUNLOCK, the flag is not
435 		 * relevant for future reclamations.
436 		 */
437 		ASSERT_VOP_ELOCKED(vp, "unlink_lowervp");
438 		KASSERT((vp->v_iflag & VI_DOOMED) == 0,
439 		    ("reclaimed nullfs vnode %p", vp));
440 		xp->null_flags &= ~NULLV_NOUNLOCK;
441 	}
442 	vdrop(vp);
443 }
444 
445 static struct vfsops null_vfsops = {
446 	.vfs_extattrctl =	nullfs_extattrctl,
447 	.vfs_fhtovp =		nullfs_fhtovp,
448 	.vfs_init =		nullfs_init,
449 	.vfs_mount =		nullfs_mount,
450 	.vfs_quotactl =		nullfs_quotactl,
451 	.vfs_root =		nullfs_root,
452 	.vfs_statfs =		nullfs_statfs,
453 	.vfs_sync =		nullfs_sync,
454 	.vfs_uninit =		nullfs_uninit,
455 	.vfs_unmount =		nullfs_unmount,
456 	.vfs_vget =		nullfs_vget,
457 	.vfs_reclaim_lowervp =	nullfs_reclaim_lowervp,
458 	.vfs_unlink_lowervp =	nullfs_unlink_lowervp,
459 };
460 
461 VFS_SET(null_vfsops, nullfs, VFCF_LOOPBACK | VFCF_JAIL);
462