xref: /netbsd-src/sys/fs/unionfs/unionfs_subr.c (revision 4817a0b0b8fe9612e8ebe21a9bf2d97b95038a97)
1 /*-
2  * Copyright (c) 1994 Jan-Simon Pendry
3  * Copyright (c) 1994
4  *	The Regents of the University of California.  All rights reserved.
5  * Copyright (c) 2005, 2006 Masanori Ozawa <ozawa@ongs.co.jp>, ONGS Inc.
6  * Copyright (c) 2006 Daichi Goto <daichi@freebsd.org>
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Jan-Simon Pendry.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)union_subr.c	8.20 (Berkeley) 5/20/95
36  * $FreeBSD: src/sys/fs/unionfs/union_subr.c,v 1.99 2008/01/24 12:34:27 attilio Exp $
37  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/mutex.h>
44 #include <sys/malloc.h>
45 #include <sys/mount.h>
46 #include <sys/namei.h>
47 #include <sys/proc.h>
48 #include <sys/vnode.h>
49 #include <sys/dirent.h>
50 #include <sys/fcntl.h>
51 #include <sys/filedesc.h>
52 #include <sys/stat.h>
53 #include <sys/kauth.h>
54 #include <sys/resourcevar.h>
55 
56 #include <fs/unionfs/unionfs.h>
57 
58 MALLOC_DEFINE(M_UNIONFSPATH, "UNIONFS path", "UNIONFS path private part");
59 
60 /*
61  * Make a new or get existing unionfs node.
62  *
63  * uppervp and lowervp should be unlocked. Because if new unionfs vnode is
64  * locked, uppervp or lowervp is locked too. In order to prevent dead lock,
65  * you should not lock plurality simultaneously.
66  */
67 int
68 unionfs_nodeget(struct mount *mp, struct vnode *uppervp,
69 		struct vnode *lowervp, struct vnode *dvp,
70 		struct vnode **vpp, struct componentname *cnp)
71 {
72 	struct unionfs_mount *ump;
73 	struct unionfs_node *unp;
74 	struct vnode   *vp;
75 	int		error;
76 	const char	*path;
77 
78 	ump = MOUNTTOUNIONFSMOUNT(mp);
79 	path = (cnp ? cnp->cn_nameptr : NULL);
80 
81 	if (uppervp == NULLVP && lowervp == NULLVP)
82 		panic("unionfs_nodeget: upper and lower is null");
83 
84 	/* If it has no ISLASTCN flag, path check is skipped. */
85 	if (cnp && !(cnp->cn_flags & ISLASTCN))
86 		path = NULL;
87 
88 	if ((uppervp == NULLVP || ump->um_uppervp != uppervp) ||
89 	    (lowervp == NULLVP || ump->um_lowervp != lowervp)) {
90 		if (dvp == NULLVP)
91 			return (EINVAL);
92 	}
93 
94 	unp = kmem_zalloc(sizeof(*unp), KM_SLEEP);
95 	if (unp == NULL)
96 		return (ENOMEM);
97 	error = getnewvnode(VT_UNION, mp, unionfs_vnodeop_p, &vp);
98 	if (error != 0) {
99 		kmem_free(unp, sizeof(*unp));
100 		return (error);
101 	}
102 	if (dvp != NULLVP)
103 		vref(dvp);
104 	if (uppervp != NULLVP)
105 		vref(uppervp);
106 	if (lowervp != NULLVP)
107 		vref(lowervp);
108 
109 	unp->un_vnode = vp;
110 	unp->un_uppervp = uppervp;
111 	unp->un_lowervp = lowervp;
112 	unp->un_dvp = dvp;
113 
114 	if (path != NULL) {
115 		unp->un_path = (char *)
116 		    malloc(cnp->cn_namelen +1, M_UNIONFSPATH, M_WAITOK|M_ZERO);
117 		memcpy(unp->un_path, cnp->cn_nameptr, cnp->cn_namelen);
118 		unp->un_path[cnp->cn_namelen] = '\0';
119 	}
120 	vp->v_type = (uppervp != NULLVP ? uppervp->v_type : lowervp->v_type);
121 	vp->v_data = unp;
122 	uvm_vnp_setsize(vp, 0);
123 
124 	if ((uppervp != NULLVP && ump->um_uppervp == uppervp) &&
125 	    (lowervp != NULLVP && ump->um_lowervp == lowervp))
126 		vp->v_vflag |= VV_ROOT;
127 
128 	*vpp = vp;
129 
130 	return (0);
131 }
132 
133 /*
134  * Clean up the unionfs node.
135  */
136 void
137 unionfs_noderem(struct vnode *vp)
138 {
139 	struct unionfs_mount *ump;
140 	struct unionfs_node *unp;
141 	struct unionfs_node_status *unsp;
142 	struct vnode   *lvp;
143 	struct vnode   *uvp;
144 
145 	ump = MOUNTTOUNIONFSMOUNT(vp->v_mount);
146 
147 	/*
148 	 * Use the interlock to protect the clearing of v_data to
149 	 * prevent faults in unionfs_lock().
150 	 */
151 	unp = VTOUNIONFS(vp);
152 	lvp = unp->un_lowervp;
153 	uvp = unp->un_uppervp;
154 	unp->un_lowervp = unp->un_uppervp = NULLVP;
155 	vp->v_data = NULL;
156 
157 	if (lvp != NULLVP)
158 		vrele(lvp);
159 	if (uvp != NULLVP)
160 		vrele(uvp);
161 	if (unp->un_dvp != NULLVP) {
162 		vrele(unp->un_dvp);
163 		unp->un_dvp = NULLVP;
164 	}
165 	if (unp->un_path) {
166 		free(unp->un_path, M_UNIONFSPATH);
167 		unp->un_path = NULL;
168 	}
169 
170 	while ((unsp = LIST_FIRST(&(unp->un_unshead))) != NULL) {
171 		LIST_REMOVE(unsp, uns_list);
172 		free(unsp, M_TEMP);
173 	}
174 	kmem_free(unp, sizeof(*unp));
175 }
176 
177 /*
178  * Get the unionfs node status.
179  * You need exclusive lock this vnode.
180  */
181 void
182 unionfs_get_node_status(struct unionfs_node *unp,
183 			struct unionfs_node_status **unspp)
184 {
185 	struct unionfs_node_status *unsp;
186 	pid_t pid;
187 	lwpid_t lid;
188 
189 	KASSERT(NULL != unspp);
190 	KASSERT(VOP_ISLOCKED(UNIONFSTOV(unp)) == LK_EXCLUSIVE);
191 
192 	pid = curproc->p_pid;
193 	lid = curlwp->l_lid;
194 
195 	LIST_FOREACH(unsp, &(unp->un_unshead), uns_list) {
196 		if (unsp->uns_pid == pid && unsp->uns_lid == lid) {
197 			*unspp = unsp;
198 			return;
199 		}
200 	}
201 
202 	/* create a new unionfs node status */
203 	unsp = kmem_zalloc(sizeof(*unsp), KM_SLEEP);
204 	unsp->uns_pid = pid;
205 	unsp->uns_lid = lid;
206 	LIST_INSERT_HEAD(&(unp->un_unshead), unsp, uns_list);
207 
208 	*unspp = unsp;
209 }
210 
211 /*
212  * Remove the unionfs node status, if you can.
213  * You need exclusive lock this vnode.
214  */
215 void
216 unionfs_tryrem_node_status(struct unionfs_node *unp,
217 			   struct unionfs_node_status *unsp)
218 {
219 	KASSERT(NULL != unsp);
220 	KASSERT(VOP_ISLOCKED(UNIONFSTOV(unp)) == LK_EXCLUSIVE);
221 
222 	if (0 < unsp->uns_lower_opencnt || 0 < unsp->uns_upper_opencnt)
223 		return;
224 
225 	LIST_REMOVE(unsp, uns_list);
226 	kmem_free(unsp, sizeof(*unsp));
227 }
228 
229 /*
230  * Create upper node attr.
231  */
232 void
233 unionfs_create_uppervattr_core(struct unionfs_mount *ump,
234 			       struct vattr *lva,
235 			       struct vattr *uva)
236 {
237 	vattr_null(uva);
238 	uva->va_type = lva->va_type;
239 	uva->va_atime = lva->va_atime;
240 	uva->va_mtime = lva->va_mtime;
241 	uva->va_ctime = lva->va_ctime;
242 
243 	switch (ump->um_copymode) {
244 	case UNIONFS_TRANSPARENT:
245 		uva->va_mode = lva->va_mode;
246 		uva->va_uid = lva->va_uid;
247 		uva->va_gid = lva->va_gid;
248 		break;
249 	case UNIONFS_MASQUERADE:
250 		if (ump->um_uid == lva->va_uid) {
251 			uva->va_mode = lva->va_mode & 077077;
252 			uva->va_mode |= (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile) & 0700;
253 			uva->va_uid = lva->va_uid;
254 			uva->va_gid = lva->va_gid;
255 		} else {
256 			uva->va_mode = (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile);
257 			uva->va_uid = ump->um_uid;
258 			uva->va_gid = ump->um_gid;
259 		}
260 		break;
261 	default:		/* UNIONFS_TRADITIONAL */
262 		uva->va_mode = 0777 & ~curproc->p_cwdi->cwdi_cmask;
263 		uva->va_uid = ump->um_uid;
264 		uva->va_gid = ump->um_gid;
265 		break;
266 	}
267 }
268 
269 /*
270  * Create upper node attr.
271  */
272 int
273 unionfs_create_uppervattr(struct unionfs_mount *ump,
274 			  struct vnode *lvp,
275 			  struct vattr *uva,
276 			  kauth_cred_t cred)
277 {
278 	int		error;
279 	struct vattr	lva;
280 
281 	if ((error = VOP_GETATTR(lvp, &lva, cred)))
282 		return (error);
283 
284 	unionfs_create_uppervattr_core(ump, &lva, uva);
285 
286 	return (error);
287 }
288 
289 /*
290  * relookup
291  *
292  * dvp should be locked on entry and will be locked on return.
293  *
294  * If an error is returned, *vpp will be invalid, otherwise it will hold a
295  * locked, referenced vnode. If *vpp == dvp then remember that only one
296  * LK_EXCLUSIVE lock is held.
297  */
298 static int
299 unionfs_relookup(struct vnode *dvp, struct vnode **vpp,
300 		 struct componentname *cnp, struct componentname *cn,
301 		 char **pnbuf_ret,
302 		 const char *path, int pathlen, u_long nameiop)
303 {
304 	int	error;
305 	char *pnbuf;
306 
307 	cn->cn_namelen = pathlen;
308 	pnbuf = PNBUF_GET();
309 	memcpy(pnbuf, path, pathlen);
310 	pnbuf[pathlen] = '\0';
311 
312 	cn->cn_nameiop = nameiop;
313 	cn->cn_flags = (LOCKPARENT | LOCKLEAF | ISLASTCN);
314 	cn->cn_cred = cnp->cn_cred;
315 
316 	cn->cn_nameptr = pnbuf;
317 	cn->cn_consume = cnp->cn_consume;
318 
319 	if (nameiop == DELETE)
320 		cn->cn_flags |= (cnp->cn_flags & (DOWHITEOUT | SAVESTART));
321 	else if (RENAME == nameiop)
322 		cn->cn_flags |= (cnp->cn_flags & SAVESTART);
323 
324 	vref(dvp);
325 	VOP_UNLOCK(dvp);
326 
327 	if ((error = relookup(dvp, vpp, cn))) {
328 		PNBUF_PUT(pnbuf);
329 		*pnbuf_ret = NULL;
330 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
331 	} else {
332 		*pnbuf_ret = pnbuf;
333 		vrele(dvp);
334 	}
335 
336 	return (error);
337 }
338 
339 /*
340  * relookup for CREATE namei operation.
341  *
342  * dvp is unionfs vnode. dvp should be locked.
343  *
344  * If it called 'unionfs_copyfile' function by unionfs_link etc,
345  * VOP_LOOKUP information is broken.
346  * So it need relookup in order to create link etc.
347  */
348 int
349 unionfs_relookup_for_create(struct vnode *dvp, struct componentname *cnp)
350 {
351 	int	error;
352 	struct vnode *udvp;
353 	struct vnode *vp;
354 	struct componentname cn;
355 	char *pnbuf;
356 
357 	udvp = UNIONFSVPTOUPPERVP(dvp);
358 	vp = NULLVP;
359 
360 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf,
361 	    cnp->cn_nameptr,
362 	    strlen(cnp->cn_nameptr), CREATE);
363 	if (error)
364 		return (error);
365 
366 	if (vp != NULLVP) {
367 		if (udvp == vp)
368 			vrele(vp);
369 		else
370 			vput(vp);
371 
372 		error = EEXIST;
373 	}
374 
375 	PNBUF_PUT(pnbuf);
376 
377 	if (!error) {
378 		cnp->cn_flags = cn.cn_flags;
379 	}
380 
381 	return (error);
382 }
383 
384 /*
385  * relookup for DELETE namei operation.
386  *
387  * dvp is unionfs vnode. dvp should be locked.
388  */
389 int
390 unionfs_relookup_for_delete(struct vnode *dvp, struct componentname *cnp)
391 {
392 	int	error;
393 	struct vnode *udvp;
394 	struct vnode *vp;
395 	struct componentname cn;
396 	char *pnbuf;
397 
398 	udvp = UNIONFSVPTOUPPERVP(dvp);
399 	vp = NULLVP;
400 
401 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
402 	    strlen(cnp->cn_nameptr), DELETE);
403 	if (error)
404 		return (error);
405 
406 	if (vp == NULLVP)
407 		error = ENOENT;
408 	else {
409 		if (udvp == vp)
410 			vrele(vp);
411 		else
412 			vput(vp);
413 	}
414 
415 	PNBUF_PUT(pnbuf);
416 
417 	if (!error) {
418 		cnp->cn_flags = cn.cn_flags;
419 	}
420 
421 	return (error);
422 }
423 
424 /*
425  * relookup for RENAME namei operation.
426  *
427  * dvp is unionfs vnode. dvp should be locked.
428  */
429 int
430 unionfs_relookup_for_rename(struct vnode *dvp, struct componentname *cnp)
431 {
432 	int error;
433 	struct vnode *udvp;
434 	struct vnode *vp;
435 	struct componentname cn;
436 	char *pnbuf;
437 
438 	udvp = UNIONFSVPTOUPPERVP(dvp);
439 	vp = NULLVP;
440 
441 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
442 	    strlen(cnp->cn_nameptr), RENAME);
443 	if (error)
444 		return (error);
445 
446 	if (vp != NULLVP) {
447 		if (udvp == vp)
448 			vrele(vp);
449 		else
450 			vput(vp);
451 	}
452 
453 	PNBUF_PUT(pnbuf);
454 
455 	if (!error) {
456 		cnp->cn_flags = cn.cn_flags;
457 	}
458 
459 	return (error);
460 
461 }
462 
463 /*
464  * Update the unionfs_node.
465  *
466  * uvp is new locked upper vnode. unionfs vnode's lock will be exchanged to the
467  * uvp's lock and lower's lock will be unlocked.
468  */
469 static void
470 unionfs_node_update(struct unionfs_node *unp, struct vnode *uvp)
471 {
472 	struct vnode   *vp;
473 	struct vnode   *lvp;
474 
475 	vp = UNIONFSTOV(unp);
476 	lvp = unp->un_lowervp;
477 
478 	/*
479 	 * lock update
480 	 */
481 	mutex_enter(&vp->v_interlock);
482 	unp->un_uppervp = uvp;
483 	KASSERT(VOP_ISLOCKED(lvp) == LK_EXCLUSIVE);
484 	mutex_exit(&vp->v_interlock);
485 }
486 
487 /*
488  * Create a new shadow dir.
489  *
490  * udvp should be locked on entry and will be locked on return.
491  *
492  * If no error returned, unp will be updated.
493  */
494 int
495 unionfs_mkshadowdir(struct unionfs_mount *ump, struct vnode *udvp,
496 		    struct unionfs_node *unp, struct componentname *cnp)
497 {
498 	int		error;
499 	struct vnode   *lvp;
500 	struct vnode   *uvp;
501 	struct vattr	va;
502 	struct vattr	lva;
503 	struct componentname cn;
504 	char *pnbuf;
505 
506 	if (unp->un_uppervp != NULLVP)
507 		return (EEXIST);
508 
509 	lvp = unp->un_lowervp;
510 	uvp = NULLVP;
511 
512 	memset(&cn, 0, sizeof(cn));
513 
514 	if ((error = VOP_GETATTR(lvp, &lva, cnp->cn_cred)))
515 		goto unionfs_mkshadowdir_abort;
516 
517 	if ((error = unionfs_relookup(udvp, &uvp, cnp, &cn, &pnbuf,
518 			cnp->cn_nameptr, cnp->cn_namelen, CREATE)))
519 		goto unionfs_mkshadowdir_abort;
520 	if (uvp != NULLVP) {
521 		if (udvp == uvp)
522 			vrele(uvp);
523 		else
524 			vput(uvp);
525 
526 		error = EEXIST;
527 		goto unionfs_mkshadowdir_free_out;
528 	}
529 
530 	unionfs_create_uppervattr_core(ump, &lva, &va);
531 
532 	error = VOP_MKDIR(udvp, &uvp, &cn, &va);
533 
534 	if (!error) {
535 		unionfs_node_update(unp, uvp);
536 
537 		/*
538 		 * XXX The bug which cannot set uid/gid was corrected.
539 		 * Ignore errors.   XXXNETBSD Why is this done as root?
540 		 */
541 		va.va_type = VNON;
542 		VOP_SETATTR(uvp, &va, lwp0.l_cred);
543 	}
544 
545 unionfs_mkshadowdir_free_out:
546 	PNBUF_PUT(pnbuf);
547 
548 unionfs_mkshadowdir_abort:
549 
550 	return (error);
551 }
552 
553 /*
554  * Create a new whiteout.
555  *
556  * dvp should be locked on entry and will be locked on return.
557  */
558 int
559 unionfs_mkwhiteout(struct vnode *dvp, struct componentname *cnp, const char *path)
560 {
561 	int		error;
562 	struct vnode   *wvp;
563 	struct componentname cn;
564 	char *pnbuf;
565 
566 	if (path == NULL)
567 		path = cnp->cn_nameptr;
568 
569 	wvp = NULLVP;
570 	if ((error = unionfs_relookup(dvp, &wvp, cnp, &cn, &pnbuf,
571 			path, strlen(path), CREATE)))
572 		return (error);
573 	if (wvp != NULLVP) {
574 		PNBUF_PUT(pnbuf);
575 		if (dvp == wvp)
576 			vrele(wvp);
577 		else
578 			vput(wvp);
579 
580 		return (EEXIST);
581 	}
582 
583 	PNBUF_PUT(pnbuf);
584 
585 	return (error);
586 }
587 
588 /*
589  * Create a new vnode for create a new shadow file.
590  *
591  * If an error is returned, *vpp will be invalid, otherwise it will hold a
592  * locked, referenced and opened vnode.
593  *
594  * unp is never updated.
595  */
596 static int
597 unionfs_vn_create_on_upper(struct vnode **vpp, struct vnode *udvp,
598 			   struct unionfs_node *unp, struct vattr *uvap)
599 {
600 	struct unionfs_mount *ump;
601 	struct vnode   *vp;
602 	struct vnode   *lvp;
603 	kauth_cred_t    cred;
604 	struct vattr	lva;
605 	int		fmode;
606 	int		error;
607 	struct componentname cn;
608 	char *pnbuf;
609 
610 	ump = MOUNTTOUNIONFSMOUNT(UNIONFSTOV(unp)->v_mount);
611 	vp = NULLVP;
612 	lvp = unp->un_lowervp;
613 	cred = kauth_cred_get();
614 	fmode = FFLAGS(O_WRONLY | O_CREAT | O_TRUNC | O_EXCL);
615 	error = 0;
616 
617 	if ((error = VOP_GETATTR(lvp, &lva, cred)) != 0)
618 		return (error);
619 	unionfs_create_uppervattr_core(ump, &lva, uvap);
620 
621 	if (unp->un_path == NULL)
622 		panic("unionfs: un_path is null");
623 
624 	cn.cn_namelen = strlen(unp->un_path);
625 	pnbuf = PNBUF_GET();
626 	memcpy(pnbuf, unp->un_path, cn.cn_namelen + 1);
627 	cn.cn_nameiop = CREATE;
628 	cn.cn_flags = (LOCKPARENT | LOCKLEAF | ISLASTCN);
629 	cn.cn_cred = cred;
630 	cn.cn_nameptr = pnbuf;
631 	cn.cn_consume = 0;
632 
633 	vref(udvp);
634 	if ((error = relookup(udvp, &vp, &cn)) != 0)
635 		goto unionfs_vn_create_on_upper_free_out2;
636 	vrele(udvp);
637 
638 	if (vp != NULLVP) {
639 		if (vp == udvp)
640 			vrele(vp);
641 		else
642 			vput(vp);
643 		error = EEXIST;
644 		goto unionfs_vn_create_on_upper_free_out1;
645 	}
646 
647 	if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
648 		goto unionfs_vn_create_on_upper_free_out1;
649 
650 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0) {
651 		vput(vp);
652 		goto unionfs_vn_create_on_upper_free_out1;
653 	}
654 	vp->v_writecount++;
655 	*vpp = vp;
656 
657 unionfs_vn_create_on_upper_free_out1:
658 	VOP_UNLOCK(udvp);
659 
660 unionfs_vn_create_on_upper_free_out2:
661 	PNBUF_PUT(pnbuf);
662 
663 	return (error);
664 }
665 
666 /*
667  * Copy from lvp to uvp.
668  *
669  * lvp and uvp should be locked and opened on entry and will be locked and
670  * opened on return.
671  */
672 static int
673 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
674 		      kauth_cred_t cred)
675 {
676 	int		error;
677 	off_t		offset;
678 	int		count;
679 	int		bufoffset;
680 	char           *buf;
681 	struct uio	uio;
682 	struct iovec	iov;
683 
684 	error = 0;
685 	memset(&uio, 0, sizeof(uio));
686 	UIO_SETUP_SYSSPACE(&uio);
687 	uio.uio_offset = 0;
688 
689 	buf = kmem_alloc(MAXBSIZE, KM_SLEEP);
690 	if (buf == NULL)
691 		return ENOMEM;
692 
693 	while (error == 0) {
694 		offset = uio.uio_offset;
695 
696 		uio.uio_iov = &iov;
697 		uio.uio_iovcnt = 1;
698 		iov.iov_base = buf;
699 		iov.iov_len = MAXBSIZE;
700 		uio.uio_resid = iov.iov_len;
701 		uio.uio_rw = UIO_READ;
702 
703 		if ((error = VOP_READ(lvp, &uio, 0, cred)) != 0)
704 			break;
705 		if ((count = MAXBSIZE - uio.uio_resid) == 0)
706 			break;
707 
708 		bufoffset = 0;
709 		while (bufoffset < count) {
710 			uio.uio_iov = &iov;
711 			uio.uio_iovcnt = 1;
712 			iov.iov_base = buf + bufoffset;
713 			iov.iov_len = count - bufoffset;
714 			uio.uio_offset = offset + bufoffset;
715 			uio.uio_resid = iov.iov_len;
716 			uio.uio_rw = UIO_WRITE;
717 
718 			if ((error = VOP_WRITE(uvp, &uio, 0, cred)) != 0)
719 				break;
720 
721 			bufoffset += (count - bufoffset) - uio.uio_resid;
722 		}
723 
724 		uio.uio_offset = offset + bufoffset;
725 	}
726 
727 	kmem_free(buf, MAXBSIZE);
728 
729 	return (error);
730 }
731 
732 /*
733  * Copy file from lower to upper.
734  *
735  * If you need copy of the contents, set 1 to docopy. Otherwise, set 0 to
736  * docopy.
737  *
738  * If no error returned, unp will be updated.
739  */
740 int
741 unionfs_copyfile(struct unionfs_node *unp, int docopy, kauth_cred_t cred)
742 {
743 	int		error;
744 	struct vnode   *udvp;
745 	struct vnode   *lvp;
746 	struct vnode   *uvp;
747 	struct vattr	uva;
748 
749 	lvp = unp->un_lowervp;
750 	uvp = NULLVP;
751 
752 	if ((UNIONFSTOV(unp)->v_mount->mnt_flag & MNT_RDONLY))
753 		return (EROFS);
754 	if (unp->un_dvp == NULLVP)
755 		return (EINVAL);
756 	if (unp->un_uppervp != NULLVP)
757 		return (EEXIST);
758 	udvp = VTOUNIONFS(unp->un_dvp)->un_uppervp;
759 	if (udvp == NULLVP)
760 		return (EROFS);
761 	if ((udvp->v_mount->mnt_flag & MNT_RDONLY))
762 		return (EROFS);
763 
764 	error = VOP_ACCESS(lvp, VREAD, cred);
765 	if (error != 0)
766 		return (error);
767 
768 	error = unionfs_vn_create_on_upper(&uvp, udvp, unp, &uva);
769 	if (error != 0)
770 		return (error);
771 
772 	if (docopy != 0) {
773 		error = VOP_OPEN(lvp, FREAD, cred);
774 		if (error == 0) {
775 			error = unionfs_copyfile_core(lvp, uvp, cred);
776 			VOP_CLOSE(lvp, FREAD, cred);
777 		}
778 	}
779 	VOP_CLOSE(uvp, FWRITE, cred);
780 	uvp->v_writecount--;
781 
782 	if (error == 0) {
783 		/* Reset the attributes. Ignore errors. */
784 		uva.va_type = VNON;
785 		VOP_SETATTR(uvp, &uva, cred);
786 	}
787 
788 	unionfs_node_update(unp, uvp);
789 
790 	return (error);
791 }
792 
793 /*
794  * It checks whether vp can rmdir. (check empty)
795  *
796  * vp is unionfs vnode.
797  * vp should be locked.
798  */
799 int
800 unionfs_check_rmdir(struct vnode *vp, kauth_cred_t cred)
801 {
802 	int		error;
803 	int		eofflag;
804 	int		lookuperr;
805 	struct vnode   *uvp;
806 	struct vnode   *lvp;
807 	struct vnode   *tvp;
808 	struct vattr	va;
809 	struct componentname cn;
810 	/*
811 	 * The size of buf needs to be larger than DIRBLKSIZ.
812 	 */
813 	char		buf[256 * 6];
814 	struct dirent  *dp;
815 	struct dirent  *edp;
816 	struct uio	uio;
817 	struct iovec	iov;
818 
819 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
820 
821 	eofflag = 0;
822 	uvp = UNIONFSVPTOUPPERVP(vp);
823 	lvp = UNIONFSVPTOLOWERVP(vp);
824 
825 	/* check opaque */
826 	if ((error = VOP_GETATTR(uvp, &va, cred)) != 0)
827 		return (error);
828 	if (va.va_flags & OPAQUE)
829 		return (0);
830 
831 	/* open vnode */
832 	if ((error = VOP_ACCESS(vp, VEXEC|VREAD, cred)) != 0)
833 		return (error);
834 	if ((error = VOP_OPEN(vp, FREAD, cred)) != 0)
835 		return (error);
836 
837 	UIO_SETUP_SYSSPACE(&uio);
838 	uio.uio_rw = UIO_READ;
839 	uio.uio_offset = 0;
840 
841 	while (!error && !eofflag) {
842 		iov.iov_base = buf;
843 		iov.iov_len = sizeof(buf);
844 		uio.uio_iov = &iov;
845 		uio.uio_iovcnt = 1;
846 		uio.uio_resid = iov.iov_len;
847 
848 		error = VOP_READDIR(lvp, &uio, cred, &eofflag, NULL, NULL);
849 		if (error)
850 			break;
851 
852 		edp = (struct dirent*)&buf[sizeof(buf) - uio.uio_resid];
853 		for (dp = (struct dirent*)buf; !error && dp < edp;
854 		     dp = (struct dirent*)((char *)dp + dp->d_reclen)) {
855 			if (dp->d_type == DT_WHT ||
856 			    (dp->d_namlen == 1 && dp->d_name[0] == '.') ||
857 			    (dp->d_namlen == 2 && !memcmp(dp->d_name, "..", 2)))
858 				continue;
859 
860 			cn.cn_namelen = dp->d_namlen;
861 			cn.cn_nameptr = dp->d_name;
862 			cn.cn_nameiop = LOOKUP;
863 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
864 			cn.cn_cred = cred;
865 			cn.cn_consume = 0;
866 
867 			/*
868 			 * check entry in lower.
869 			 * Sometimes, readdir function returns
870 			 * wrong entry.
871 			 */
872 			lookuperr = VOP_LOOKUP(lvp, &tvp, &cn);
873 
874 			if (!lookuperr)
875 				vput(tvp);
876 			else
877 				continue; /* skip entry */
878 
879 			/*
880 			 * check entry
881 			 * If it has no exist/whiteout entry in upper,
882 			 * directory is not empty.
883 			 */
884 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
885 			lookuperr = VOP_LOOKUP(uvp, &tvp, &cn);
886 
887 			if (!lookuperr)
888 				vput(tvp);
889 
890 			/* ignore exist or whiteout entry */
891 			if (!lookuperr ||
892 			    (lookuperr == ENOENT && (cn.cn_flags & ISWHITEOUT)))
893 				continue;
894 
895 			error = ENOTEMPTY;
896 		}
897 	}
898 
899 	/* close vnode */
900 	VOP_CLOSE(vp, FREAD, cred);
901 
902 	return (error);
903 }
904 
905 #ifdef DIAGNOSTIC
906 
907 struct vnode   *
908 unionfs_checkuppervp(struct vnode *vp, const char *fil, int lno)
909 {
910 	struct unionfs_node *unp;
911 
912 	unp = VTOUNIONFS(vp);
913 
914 #ifdef notyet
915 	if (vp->v_op != unionfs_vnodeop_p) {
916 		printf("unionfs_checkuppervp: on non-unionfs-node.\n");
917 #ifdef KDB
918 		kdb_enter(KDB_WHY_UNIONFS,
919 		    "unionfs_checkuppervp: on non-unionfs-node.\n");
920 #endif
921 		panic("unionfs_checkuppervp");
922 	};
923 #endif
924 	return (unp->un_uppervp);
925 }
926 
927 struct vnode   *
928 unionfs_checklowervp(struct vnode *vp, const char *fil, int lno)
929 {
930 	struct unionfs_node *unp;
931 
932 	unp = VTOUNIONFS(vp);
933 
934 #ifdef notyet
935 	if (vp->v_op != unionfs_vnodeop_p) {
936 		printf("unionfs_checklowervp: on non-unionfs-node.\n");
937 #ifdef KDB
938 		kdb_enter(KDB_WHY_UNIONFS,
939 		    "unionfs_checklowervp: on non-unionfs-node.\n");
940 #endif
941 		panic("unionfs_checklowervp");
942 	};
943 #endif
944 	return (unp->un_lowervp);
945 }
946 #endif
947