xref: /freebsd/sys/fs/nullfs/null_vfsops.c (revision a7dea1671b87c07d2d266f836bfa8b58efc7c134)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1992, 1993, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software donated to Berkeley by
8  * Jan-Simon Pendry.
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 University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)null_vfsops.c	8.2 (Berkeley) 1/21/94
35  *
36  * @(#)lofs_vfsops.c	1.2 (Berkeley) 6/18/92
37  * $FreeBSD$
38  */
39 
40 /*
41  * Null Layer
42  * (See null_vnops.c for a description of what this does.)
43  */
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/fcntl.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/namei.h>
53 #include <sys/proc.h>
54 #include <sys/vnode.h>
55 #include <sys/jail.h>
56 
57 #include <fs/nullfs/null.h>
58 
59 static MALLOC_DEFINE(M_NULLFSMNT, "nullfs_mount", "NULLFS mount structure");
60 
61 static vfs_fhtovp_t	nullfs_fhtovp;
62 static vfs_mount_t	nullfs_mount;
63 static vfs_quotactl_t	nullfs_quotactl;
64 static vfs_root_t	nullfs_root;
65 static vfs_sync_t	nullfs_sync;
66 static vfs_statfs_t	nullfs_statfs;
67 static vfs_unmount_t	nullfs_unmount;
68 static vfs_vget_t	nullfs_vget;
69 static vfs_extattrctl_t	nullfs_extattrctl;
70 
71 /*
72  * Mount null layer
73  */
74 static int
75 nullfs_mount(struct mount *mp)
76 {
77 	struct vnode *lowerrootvp, *vp;
78 	struct vnode *nullm_rootvp;
79 	struct null_mount *xmp;
80 	struct null_node *nn;
81 	struct nameidata nd, *ndp;
82 	char *target;
83 	int error, len;
84 	bool isvnunlocked;
85 
86 	NULLFSDEBUG("nullfs_mount(mp = %p)\n", (void *)mp);
87 
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);
117 		isvnunlocked = true;
118 	} else {
119 		isvnunlocked = false;
120 	}
121 
122 	/*
123 	 * Find lower node
124 	 */
125 	ndp = &nd;
126 	NDINIT(ndp, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, target, curthread);
127 	error = namei(ndp);
128 
129 	/*
130 	 * Re-lock vnode.
131 	 * XXXKIB This is deadlock-prone as well.
132 	 */
133 	if (isvnunlocked)
134 		vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY);
135 
136 	if (error)
137 		return (error);
138 	NDFREE(ndp, NDF_ONLY_PNBUF);
139 
140 	/*
141 	 * Sanity check on lower vnode
142 	 */
143 	lowerrootvp = ndp->ni_vp;
144 
145 	/*
146 	 * Check multi null mount to avoid `lock against myself' panic.
147 	 */
148 	if (mp->mnt_vnodecovered->v_op == &null_vnodeops) {
149 		nn = VTONULL(mp->mnt_vnodecovered);
150 		if (nn == NULL || lowerrootvp == nn->null_lowervp) {
151 			NULLFSDEBUG("nullfs_mount: multi null mount?\n");
152 			vput(lowerrootvp);
153 			return (EDEADLK);
154 		}
155 	}
156 
157 	xmp = (struct null_mount *) malloc(sizeof(struct null_mount),
158 	    M_NULLFSMNT, M_WAITOK | M_ZERO);
159 
160 	/*
161 	 * Save reference to underlying FS
162 	 */
163 	xmp->nullm_vfs = lowerrootvp->v_mount;
164 
165 	/*
166 	 * Save reference.  Each mount also holds
167 	 * a reference on the root vnode.
168 	 */
169 	error = null_nodeget(mp, lowerrootvp, &vp);
170 	/*
171 	 * Make sure the node alias worked
172 	 */
173 	if (error) {
174 		free(xmp, M_NULLFSMNT);
175 		return (error);
176 	}
177 
178 	/*
179 	 * Keep a held reference to the root vnode.
180 	 * It is vrele'd in nullfs_unmount.
181 	 */
182 	nullm_rootvp = vp;
183 	nullm_rootvp->v_vflag |= VV_ROOT;
184 	xmp->nullm_rootvp = nullm_rootvp;
185 
186 	/*
187 	 * Unlock the node (either the lower or the alias)
188 	 */
189 	VOP_UNLOCK(vp);
190 
191 	if (NULLVPTOLOWERVP(nullm_rootvp)->v_mount->mnt_flag & MNT_LOCAL) {
192 		MNT_ILOCK(mp);
193 		mp->mnt_flag |= MNT_LOCAL;
194 		MNT_IUNLOCK(mp);
195 	}
196 
197 	xmp->nullm_flags |= NULLM_CACHE;
198 	if (vfs_getopt(mp->mnt_optnew, "nocache", NULL, NULL) == 0 ||
199 	    (xmp->nullm_vfs->mnt_kern_flag & MNTK_NULL_NOCACHE) != 0)
200 		xmp->nullm_flags &= ~NULLM_CACHE;
201 
202 	MNT_ILOCK(mp);
203 	if ((xmp->nullm_flags & NULLM_CACHE) != 0) {
204 		mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag &
205 		    (MNTK_SHARED_WRITES | MNTK_LOOKUP_SHARED |
206 		    MNTK_EXTENDED_SHARED);
207 	}
208 	mp->mnt_kern_flag |= MNTK_LOOKUP_EXCL_DOTDOT | MNTK_NOMSYNC;
209 	mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag &
210 	    (MNTK_USES_BCACHE | MNTK_NO_IOPF | MNTK_UNMAPPED_BUFS);
211 	MNT_IUNLOCK(mp);
212 	mp->mnt_data = xmp;
213 	vfs_getnewfsid(mp);
214 	if ((xmp->nullm_flags & NULLM_CACHE) != 0) {
215 		MNT_ILOCK(xmp->nullm_vfs);
216 		TAILQ_INSERT_TAIL(&xmp->nullm_vfs->mnt_uppers, mp,
217 		    mnt_upper_link);
218 		MNT_IUNLOCK(xmp->nullm_vfs);
219 	}
220 
221 	vfs_mountedfrom(mp, target);
222 
223 	NULLFSDEBUG("nullfs_mount: lower %s, alias at %s\n",
224 		mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
225 	return (0);
226 }
227 
228 /*
229  * Free reference to null layer
230  */
231 static int
232 nullfs_unmount(mp, mntflags)
233 	struct mount *mp;
234 	int mntflags;
235 {
236 	struct null_mount *mntdata;
237 	struct mount *ump;
238 	int error, flags, rootrefs;
239 
240 	NULLFSDEBUG("nullfs_unmount: mp = %p\n", (void *)mp);
241 
242 	if (mntflags & MNT_FORCE)
243 		flags = FORCECLOSE;
244 	else
245 		flags = 0;
246 
247 	for (rootrefs = 1;; rootrefs = 0) {
248 		/* There is 1 extra root vnode reference (nullm_rootvp). */
249 		error = vflush(mp, rootrefs, flags, curthread);
250 		if (error)
251 			return (error);
252 		MNT_ILOCK(mp);
253 		if (mp->mnt_nvnodelistsize == 0) {
254 			MNT_IUNLOCK(mp);
255 			break;
256 		}
257 		MNT_IUNLOCK(mp);
258 		if ((mntflags & MNT_FORCE) == 0)
259 			return (EBUSY);
260 	}
261 
262 	/*
263 	 * Finally, throw away the null_mount structure
264 	 */
265 	mntdata = mp->mnt_data;
266 	ump = mntdata->nullm_vfs;
267 	if ((mntdata->nullm_flags & NULLM_CACHE) != 0) {
268 		MNT_ILOCK(ump);
269 		while ((ump->mnt_kern_flag & MNTK_VGONE_UPPER) != 0) {
270 			ump->mnt_kern_flag |= MNTK_VGONE_WAITER;
271 			msleep(&ump->mnt_uppers, &ump->mnt_mtx, 0, "vgnupw", 0);
272 		}
273 		TAILQ_REMOVE(&ump->mnt_uppers, mp, mnt_upper_link);
274 		MNT_IUNLOCK(ump);
275 	}
276 	mp->mnt_data = NULL;
277 	free(mntdata, M_NULLFSMNT);
278 	return (0);
279 }
280 
281 static int
282 nullfs_root(mp, flags, vpp)
283 	struct mount *mp;
284 	int flags;
285 	struct vnode **vpp;
286 {
287 	struct vnode *vp;
288 
289 	NULLFSDEBUG("nullfs_root(mp = %p, vp = %p->%p)\n", (void *)mp,
290 	    (void *)MOUNTTONULLMOUNT(mp)->nullm_rootvp,
291 	    (void *)NULLVPTOLOWERVP(MOUNTTONULLMOUNT(mp)->nullm_rootvp));
292 
293 	/*
294 	 * Return locked reference to root.
295 	 */
296 	vp = MOUNTTONULLMOUNT(mp)->nullm_rootvp;
297 	VREF(vp);
298 
299 	ASSERT_VOP_UNLOCKED(vp, "root vnode is locked");
300 	vn_lock(vp, flags | LK_RETRY);
301 	*vpp = vp;
302 	return 0;
303 }
304 
305 static int
306 nullfs_quotactl(mp, cmd, uid, arg)
307 	struct mount *mp;
308 	int cmd;
309 	uid_t uid;
310 	void *arg;
311 {
312 	return VFS_QUOTACTL(MOUNTTONULLMOUNT(mp)->nullm_vfs, cmd, uid, arg);
313 }
314 
315 static int
316 nullfs_statfs(mp, sbp)
317 	struct mount *mp;
318 	struct statfs *sbp;
319 {
320 	int error;
321 	struct statfs *mstat;
322 
323 	NULLFSDEBUG("nullfs_statfs(mp = %p, vp = %p->%p)\n", (void *)mp,
324 	    (void *)MOUNTTONULLMOUNT(mp)->nullm_rootvp,
325 	    (void *)NULLVPTOLOWERVP(MOUNTTONULLMOUNT(mp)->nullm_rootvp));
326 
327 	mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO);
328 
329 	error = VFS_STATFS(MOUNTTONULLMOUNT(mp)->nullm_vfs, mstat);
330 	if (error) {
331 		free(mstat, M_STATFS);
332 		return (error);
333 	}
334 
335 	/* now copy across the "interesting" information and fake the rest */
336 	sbp->f_type = mstat->f_type;
337 	sbp->f_flags = (sbp->f_flags & (MNT_RDONLY | MNT_NOEXEC | MNT_NOSUID |
338 	    MNT_UNION | MNT_NOSYMFOLLOW | MNT_AUTOMOUNTED)) |
339 	    (mstat->f_flags & ~(MNT_ROOTFS | MNT_AUTOMOUNTED));
340 	sbp->f_bsize = mstat->f_bsize;
341 	sbp->f_iosize = mstat->f_iosize;
342 	sbp->f_blocks = mstat->f_blocks;
343 	sbp->f_bfree = mstat->f_bfree;
344 	sbp->f_bavail = mstat->f_bavail;
345 	sbp->f_files = mstat->f_files;
346 	sbp->f_ffree = mstat->f_ffree;
347 
348 	free(mstat, M_STATFS);
349 	return (0);
350 }
351 
352 static int
353 nullfs_sync(mp, waitfor)
354 	struct mount *mp;
355 	int waitfor;
356 {
357 	/*
358 	 * XXX - Assumes no data cached at null layer.
359 	 */
360 	return (0);
361 }
362 
363 static int
364 nullfs_vget(mp, ino, flags, vpp)
365 	struct mount *mp;
366 	ino_t ino;
367 	int flags;
368 	struct vnode **vpp;
369 {
370 	int error;
371 
372 	KASSERT((flags & LK_TYPE_MASK) != 0,
373 	    ("nullfs_vget: no lock requested"));
374 
375 	error = VFS_VGET(MOUNTTONULLMOUNT(mp)->nullm_vfs, ino, flags, vpp);
376 	if (error != 0)
377 		return (error);
378 	return (null_nodeget(mp, *vpp, vpp));
379 }
380 
381 static int
382 nullfs_fhtovp(mp, fidp, flags, vpp)
383 	struct mount *mp;
384 	struct fid *fidp;
385 	int flags;
386 	struct vnode **vpp;
387 {
388 	int error;
389 
390 	error = VFS_FHTOVP(MOUNTTONULLMOUNT(mp)->nullm_vfs, fidp, flags,
391 	    vpp);
392 	if (error != 0)
393 		return (error);
394 	return (null_nodeget(mp, *vpp, vpp));
395 }
396 
397 static int
398 nullfs_extattrctl(mp, cmd, filename_vp, namespace, attrname)
399 	struct mount *mp;
400 	int cmd;
401 	struct vnode *filename_vp;
402 	int namespace;
403 	const char *attrname;
404 {
405 
406 	return (VFS_EXTATTRCTL(MOUNTTONULLMOUNT(mp)->nullm_vfs, cmd,
407 	    filename_vp, namespace, attrname));
408 }
409 
410 static void
411 nullfs_reclaim_lowervp(struct mount *mp, struct vnode *lowervp)
412 {
413 	struct vnode *vp;
414 
415 	vp = null_hashget(mp, lowervp);
416 	if (vp == NULL)
417 		return;
418 	VTONULL(vp)->null_flags |= NULLV_NOUNLOCK;
419 	vgone(vp);
420 	vput(vp);
421 }
422 
423 static void
424 nullfs_unlink_lowervp(struct mount *mp, struct vnode *lowervp)
425 {
426 	struct vnode *vp;
427 	struct null_node *xp;
428 
429 	vp = null_hashget(mp, lowervp);
430 	if (vp == NULL)
431 		return;
432 	xp = VTONULL(vp);
433 	xp->null_flags |= NULLV_DROP | NULLV_NOUNLOCK;
434 	vhold(vp);
435 	vunref(vp);
436 
437 	if (vp->v_usecount == 0) {
438 		/*
439 		 * If vunref() dropped the last use reference on the
440 		 * nullfs vnode, it must be reclaimed, and its lock
441 		 * was split from the lower vnode lock.  Need to do
442 		 * extra unlock before allowing the final vdrop() to
443 		 * free the vnode.
444 		 */
445 		KASSERT(VN_IS_DOOMED(vp),
446 		    ("not reclaimed nullfs vnode %p", vp));
447 		VOP_UNLOCK(vp);
448 	} else {
449 		/*
450 		 * Otherwise, the nullfs vnode still shares the lock
451 		 * with the lower vnode, and must not be unlocked.
452 		 * Also clear the NULLV_NOUNLOCK, the flag is not
453 		 * relevant for future reclamations.
454 		 */
455 		ASSERT_VOP_ELOCKED(vp, "unlink_lowervp");
456 		KASSERT(!VN_IS_DOOMED(vp),
457 		    ("reclaimed nullfs vnode %p", vp));
458 		xp->null_flags &= ~NULLV_NOUNLOCK;
459 	}
460 	vdrop(vp);
461 }
462 
463 static struct vfsops null_vfsops = {
464 	.vfs_extattrctl =	nullfs_extattrctl,
465 	.vfs_fhtovp =		nullfs_fhtovp,
466 	.vfs_init =		nullfs_init,
467 	.vfs_mount =		nullfs_mount,
468 	.vfs_quotactl =		nullfs_quotactl,
469 	.vfs_root =		nullfs_root,
470 	.vfs_statfs =		nullfs_statfs,
471 	.vfs_sync =		nullfs_sync,
472 	.vfs_uninit =		nullfs_uninit,
473 	.vfs_unmount =		nullfs_unmount,
474 	.vfs_vget =		nullfs_vget,
475 	.vfs_reclaim_lowervp =	nullfs_reclaim_lowervp,
476 	.vfs_unlink_lowervp =	nullfs_unlink_lowervp,
477 };
478 
479 VFS_SET(null_vfsops, nullfs, VFCF_LOOPBACK | VFCF_JAIL);
480