xref: /netbsd-src/sys/fs/unionfs/unionfs_subr.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
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 	/*
95 	 * Get a new vnode and share the lock with upper layer vnode,
96 	 * unless layers are inverted.
97 	 */
98 	vnode_t *svp = (uppervp != NULLVP) ? uppervp : lowervp;
99 	error = getnewvnode(VT_UNION, mp, unionfs_vnodeop_p,
100 	    svp->v_interlock, &vp);
101 	if (error != 0) {
102 		return (error);
103 	}
104 	if (dvp != NULLVP)
105 		vref(dvp);
106 	if (uppervp != NULLVP)
107 		vref(uppervp);
108 	if (lowervp != NULLVP)
109 		vref(lowervp);
110 
111 	unp = kmem_zalloc(sizeof(*unp), KM_SLEEP);
112 	unp->un_vnode = vp;
113 	unp->un_uppervp = uppervp;
114 	unp->un_lowervp = lowervp;
115 	unp->un_dvp = dvp;
116 
117 	if (path != NULL) {
118 		unp->un_path = (char *)
119 		    malloc(cnp->cn_namelen +1, M_UNIONFSPATH, M_WAITOK|M_ZERO);
120 		memcpy(unp->un_path, cnp->cn_nameptr, cnp->cn_namelen);
121 		unp->un_path[cnp->cn_namelen] = '\0';
122 	}
123 	vp->v_type = (uppervp != NULLVP ? uppervp->v_type : lowervp->v_type);
124 	vp->v_data = unp;
125 	uvm_vnp_setsize(vp, 0);
126 
127 	if ((uppervp != NULLVP && ump->um_uppervp == uppervp) &&
128 	    (lowervp != NULLVP && ump->um_lowervp == lowervp))
129 		vp->v_vflag |= VV_ROOT;
130 
131 	*vpp = vp;
132 
133 	return (0);
134 }
135 
136 /*
137  * Clean up the unionfs node.
138  */
139 void
140 unionfs_noderem(struct vnode *vp)
141 {
142 	struct unionfs_mount *ump;
143 	struct unionfs_node *unp;
144 	struct unionfs_node_status *unsp;
145 	struct vnode   *lvp;
146 	struct vnode   *uvp;
147 
148 	ump = MOUNTTOUNIONFSMOUNT(vp->v_mount);
149 
150 	/*
151 	 * Use the interlock to protect the clearing of v_data to
152 	 * prevent faults in unionfs_lock().
153 	 */
154 	unp = VTOUNIONFS(vp);
155 	lvp = unp->un_lowervp;
156 	uvp = unp->un_uppervp;
157 	unp->un_lowervp = unp->un_uppervp = NULLVP;
158 	vp->v_data = NULL;
159 
160 	if (lvp != NULLVP)
161 		vrele(lvp);
162 	if (uvp != NULLVP)
163 		vrele(uvp);
164 	if (unp->un_dvp != NULLVP) {
165 		vrele(unp->un_dvp);
166 		unp->un_dvp = NULLVP;
167 	}
168 	if (unp->un_path) {
169 		free(unp->un_path, M_UNIONFSPATH);
170 		unp->un_path = NULL;
171 	}
172 
173 	while ((unsp = LIST_FIRST(&(unp->un_unshead))) != NULL) {
174 		LIST_REMOVE(unsp, uns_list);
175 		free(unsp, M_TEMP);
176 	}
177 	kmem_free(unp, sizeof(*unp));
178 }
179 
180 /*
181  * Get the unionfs node status.
182  * You need exclusive lock this vnode.
183  */
184 void
185 unionfs_get_node_status(struct unionfs_node *unp,
186 			struct unionfs_node_status **unspp)
187 {
188 	struct unionfs_node_status *unsp;
189 	pid_t pid;
190 	lwpid_t lid;
191 
192 	KASSERT(NULL != unspp);
193 	KASSERT(VOP_ISLOCKED(UNIONFSTOV(unp)) == LK_EXCLUSIVE);
194 
195 	pid = curproc->p_pid;
196 	lid = curlwp->l_lid;
197 
198 	LIST_FOREACH(unsp, &(unp->un_unshead), uns_list) {
199 		if (unsp->uns_pid == pid && unsp->uns_lid == lid) {
200 			*unspp = unsp;
201 			return;
202 		}
203 	}
204 
205 	/* create a new unionfs node status */
206 	unsp = kmem_zalloc(sizeof(*unsp), KM_SLEEP);
207 	unsp->uns_pid = pid;
208 	unsp->uns_lid = lid;
209 	LIST_INSERT_HEAD(&(unp->un_unshead), unsp, uns_list);
210 
211 	*unspp = unsp;
212 }
213 
214 /*
215  * Remove the unionfs node status, if you can.
216  * You need exclusive lock this vnode.
217  */
218 void
219 unionfs_tryrem_node_status(struct unionfs_node *unp,
220 			   struct unionfs_node_status *unsp)
221 {
222 	KASSERT(NULL != unsp);
223 	KASSERT(VOP_ISLOCKED(UNIONFSTOV(unp)) == LK_EXCLUSIVE);
224 
225 	if (0 < unsp->uns_lower_opencnt || 0 < unsp->uns_upper_opencnt)
226 		return;
227 
228 	LIST_REMOVE(unsp, uns_list);
229 	kmem_free(unsp, sizeof(*unsp));
230 }
231 
232 /*
233  * Create upper node attr.
234  */
235 void
236 unionfs_create_uppervattr_core(struct unionfs_mount *ump,
237 			       struct vattr *lva,
238 			       struct vattr *uva)
239 {
240 	vattr_null(uva);
241 	uva->va_type = lva->va_type;
242 	uva->va_atime = lva->va_atime;
243 	uva->va_mtime = lva->va_mtime;
244 	uva->va_ctime = lva->va_ctime;
245 
246 	switch (ump->um_copymode) {
247 	case UNIONFS_TRANSPARENT:
248 		uva->va_mode = lva->va_mode;
249 		uva->va_uid = lva->va_uid;
250 		uva->va_gid = lva->va_gid;
251 		break;
252 	case UNIONFS_MASQUERADE:
253 		if (ump->um_uid == lva->va_uid) {
254 			uva->va_mode = lva->va_mode & 077077;
255 			uva->va_mode |= (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile) & 0700;
256 			uva->va_uid = lva->va_uid;
257 			uva->va_gid = lva->va_gid;
258 		} else {
259 			uva->va_mode = (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile);
260 			uva->va_uid = ump->um_uid;
261 			uva->va_gid = ump->um_gid;
262 		}
263 		break;
264 	default:		/* UNIONFS_TRADITIONAL */
265 		uva->va_mode = 0777 & ~curproc->p_cwdi->cwdi_cmask;
266 		uva->va_uid = ump->um_uid;
267 		uva->va_gid = ump->um_gid;
268 		break;
269 	}
270 }
271 
272 /*
273  * Create upper node attr.
274  */
275 int
276 unionfs_create_uppervattr(struct unionfs_mount *ump,
277 			  struct vnode *lvp,
278 			  struct vattr *uva,
279 			  kauth_cred_t cred)
280 {
281 	int		error;
282 	struct vattr	lva;
283 
284 	if ((error = VOP_GETATTR(lvp, &lva, cred)))
285 		return (error);
286 
287 	unionfs_create_uppervattr_core(ump, &lva, uva);
288 
289 	return (error);
290 }
291 
292 /*
293  * relookup
294  *
295  * dvp should be locked on entry and will be locked on return.
296  *
297  * If an error is returned, *vpp will be invalid, otherwise it will hold a
298  * locked, referenced vnode. If *vpp == dvp then remember that only one
299  * LK_EXCLUSIVE lock is held.
300  */
301 static int
302 unionfs_relookup(struct vnode *dvp, struct vnode **vpp,
303 		 struct componentname *cnp, struct componentname *cn,
304 		 char **pnbuf_ret,
305 		 const char *path, int pathlen, u_long nameiop)
306 {
307 	int	error;
308 	char *pnbuf;
309 
310 	cn->cn_namelen = pathlen;
311 	pnbuf = PNBUF_GET();
312 	memcpy(pnbuf, path, pathlen);
313 	pnbuf[pathlen] = '\0';
314 
315 	cn->cn_nameiop = nameiop;
316 	cn->cn_flags = (LOCKPARENT | LOCKLEAF | ISLASTCN);
317 	cn->cn_cred = cnp->cn_cred;
318 
319 	cn->cn_nameptr = pnbuf;
320 	cn->cn_consume = cnp->cn_consume;
321 
322 	if (nameiop == DELETE)
323 		cn->cn_flags |= (cnp->cn_flags & DOWHITEOUT);
324 
325 	vref(dvp);
326 	VOP_UNLOCK(dvp);
327 
328 	if ((error = relookup(dvp, vpp, cn, 0))) {
329 		PNBUF_PUT(pnbuf);
330 		*pnbuf_ret = NULL;
331 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
332 	} else {
333 		*pnbuf_ret = pnbuf;
334 		vrele(dvp);
335 	}
336 
337 	return (error);
338 }
339 
340 /*
341  * relookup for CREATE namei operation.
342  *
343  * dvp is unionfs vnode. dvp should be locked.
344  *
345  * If it called 'unionfs_copyfile' function by unionfs_link etc,
346  * VOP_LOOKUP information is broken.
347  * So it need relookup in order to create link etc.
348  */
349 int
350 unionfs_relookup_for_create(struct vnode *dvp, struct componentname *cnp)
351 {
352 	int	error;
353 	struct vnode *udvp;
354 	struct vnode *vp;
355 	struct componentname cn;
356 	char *pnbuf;
357 
358 	udvp = UNIONFSVPTOUPPERVP(dvp);
359 	vp = NULLVP;
360 
361 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf,
362 	    cnp->cn_nameptr,
363 	    strlen(cnp->cn_nameptr), CREATE);
364 	if (error)
365 		return (error);
366 
367 	if (vp != NULLVP) {
368 		if (udvp == vp)
369 			vrele(vp);
370 		else
371 			vput(vp);
372 
373 		error = EEXIST;
374 	}
375 
376 	PNBUF_PUT(pnbuf);
377 
378 	if (!error) {
379 		cnp->cn_flags = cn.cn_flags;
380 	}
381 
382 	return (error);
383 }
384 
385 /*
386  * relookup for DELETE namei operation.
387  *
388  * dvp is unionfs vnode. dvp should be locked.
389  */
390 int
391 unionfs_relookup_for_delete(struct vnode *dvp, struct componentname *cnp)
392 {
393 	int	error;
394 	struct vnode *udvp;
395 	struct vnode *vp;
396 	struct componentname cn;
397 	char *pnbuf;
398 
399 	udvp = UNIONFSVPTOUPPERVP(dvp);
400 	vp = NULLVP;
401 
402 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
403 	    strlen(cnp->cn_nameptr), DELETE);
404 	if (error)
405 		return (error);
406 
407 	if (vp == NULLVP)
408 		error = ENOENT;
409 	else {
410 		if (udvp == vp)
411 			vrele(vp);
412 		else
413 			vput(vp);
414 	}
415 
416 	PNBUF_PUT(pnbuf);
417 
418 	if (!error) {
419 		cnp->cn_flags = cn.cn_flags;
420 	}
421 
422 	return (error);
423 }
424 
425 /*
426  * relookup for RENAME namei operation.
427  *
428  * dvp is unionfs vnode. dvp should be locked.
429  */
430 int
431 unionfs_relookup_for_rename(struct vnode *dvp, struct componentname *cnp)
432 {
433 	int error;
434 	struct vnode *udvp;
435 	struct vnode *vp;
436 	struct componentname cn;
437 	char *pnbuf;
438 
439 	udvp = UNIONFSVPTOUPPERVP(dvp);
440 	vp = NULLVP;
441 
442 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
443 	    strlen(cnp->cn_nameptr), RENAME);
444 	if (error)
445 		return (error);
446 
447 	if (vp != NULLVP) {
448 		if (udvp == vp)
449 			vrele(vp);
450 		else
451 			vput(vp);
452 	}
453 
454 	PNBUF_PUT(pnbuf);
455 
456 	if (!error) {
457 		cnp->cn_flags = cn.cn_flags;
458 	}
459 
460 	return (error);
461 
462 }
463 
464 /*
465  * Update the unionfs_node.
466  *
467  * uvp is new locked upper vnode. unionfs vnode's lock will be exchanged to the
468  * uvp's lock and lower's lock will be unlocked.
469  */
470 static void
471 unionfs_node_update(struct unionfs_node *unp, struct vnode *uvp)
472 {
473 	struct vnode   *vp;
474 	struct vnode   *lvp;
475 
476 	vp = UNIONFSTOV(unp);
477 	lvp = unp->un_lowervp;
478 
479 	/*
480 	 * lock update
481 	 */
482 	mutex_enter(vp->v_interlock);
483 	unp->un_uppervp = uvp;
484 	KASSERT(VOP_ISLOCKED(lvp) == LK_EXCLUSIVE);
485 	mutex_exit(vp->v_interlock);
486 }
487 
488 /*
489  * Create a new shadow dir.
490  *
491  * udvp should be locked on entry and will be locked on return.
492  *
493  * If no error returned, unp will be updated.
494  */
495 int
496 unionfs_mkshadowdir(struct unionfs_mount *ump, struct vnode *udvp,
497 		    struct unionfs_node *unp, struct componentname *cnp)
498 {
499 	int		error;
500 	struct vnode   *lvp;
501 	struct vnode   *uvp;
502 	struct vattr	va;
503 	struct vattr	lva;
504 	struct componentname cn;
505 	char *pnbuf;
506 
507 	if (unp->un_uppervp != NULLVP)
508 		return (EEXIST);
509 
510 	lvp = unp->un_lowervp;
511 	uvp = NULLVP;
512 
513 	memset(&cn, 0, sizeof(cn));
514 
515 	if ((error = VOP_GETATTR(lvp, &lva, cnp->cn_cred)))
516 		goto unionfs_mkshadowdir_abort;
517 
518 	if ((error = unionfs_relookup(udvp, &uvp, cnp, &cn, &pnbuf,
519 			cnp->cn_nameptr, cnp->cn_namelen, CREATE)))
520 		goto unionfs_mkshadowdir_abort;
521 	if (uvp != NULLVP) {
522 		if (udvp == uvp)
523 			vrele(uvp);
524 		else
525 			vput(uvp);
526 
527 		error = EEXIST;
528 		goto unionfs_mkshadowdir_free_out;
529 	}
530 
531 	unionfs_create_uppervattr_core(ump, &lva, &va);
532 
533 	error = VOP_MKDIR(udvp, &uvp, &cn, &va);
534 
535 	if (!error) {
536 		unionfs_node_update(unp, uvp);
537 
538 		/*
539 		 * XXX The bug which cannot set uid/gid was corrected.
540 		 * Ignore errors.   XXXNETBSD Why is this done as root?
541 		 */
542 		va.va_type = VNON;
543 		VOP_SETATTR(uvp, &va, lwp0.l_cred);
544 	}
545 
546 unionfs_mkshadowdir_free_out:
547 	PNBUF_PUT(pnbuf);
548 
549 unionfs_mkshadowdir_abort:
550 
551 	return (error);
552 }
553 
554 /*
555  * Create a new whiteout.
556  *
557  * dvp should be locked on entry and will be locked on return.
558  */
559 int
560 unionfs_mkwhiteout(struct vnode *dvp, struct componentname *cnp, const char *path)
561 {
562 	int		error;
563 	struct vnode   *wvp;
564 	struct componentname cn;
565 	char *pnbuf;
566 
567 	if (path == NULL)
568 		path = cnp->cn_nameptr;
569 
570 	wvp = NULLVP;
571 	if ((error = unionfs_relookup(dvp, &wvp, cnp, &cn, &pnbuf,
572 			path, strlen(path), CREATE)))
573 		return (error);
574 	if (wvp != NULLVP) {
575 		PNBUF_PUT(pnbuf);
576 		if (dvp == wvp)
577 			vrele(wvp);
578 		else
579 			vput(wvp);
580 
581 		return (EEXIST);
582 	}
583 
584 	PNBUF_PUT(pnbuf);
585 
586 	return (error);
587 }
588 
589 /*
590  * Create a new vnode for create a new shadow file.
591  *
592  * If an error is returned, *vpp will be invalid, otherwise it will hold a
593  * locked, referenced and opened vnode.
594  *
595  * unp is never updated.
596  */
597 static int
598 unionfs_vn_create_on_upper(struct vnode **vpp, struct vnode *udvp,
599 			   struct unionfs_node *unp, struct vattr *uvap)
600 {
601 	struct unionfs_mount *ump;
602 	struct vnode   *vp;
603 	struct vnode   *lvp;
604 	kauth_cred_t    cred;
605 	struct vattr	lva;
606 	int		fmode;
607 	int		error;
608 	struct componentname cn;
609 	char *pnbuf;
610 
611 	ump = MOUNTTOUNIONFSMOUNT(UNIONFSTOV(unp)->v_mount);
612 	vp = NULLVP;
613 	lvp = unp->un_lowervp;
614 	cred = kauth_cred_get();
615 	fmode = FFLAGS(O_WRONLY | O_CREAT | O_TRUNC | O_EXCL);
616 	error = 0;
617 
618 	if ((error = VOP_GETATTR(lvp, &lva, cred)) != 0)
619 		return (error);
620 	unionfs_create_uppervattr_core(ump, &lva, uvap);
621 
622 	if (unp->un_path == NULL)
623 		panic("unionfs: un_path is null");
624 
625 	cn.cn_namelen = strlen(unp->un_path);
626 	pnbuf = PNBUF_GET();
627 	memcpy(pnbuf, unp->un_path, cn.cn_namelen + 1);
628 	cn.cn_nameiop = CREATE;
629 	cn.cn_flags = (LOCKPARENT | LOCKLEAF | ISLASTCN);
630 	cn.cn_cred = cred;
631 	cn.cn_nameptr = pnbuf;
632 	cn.cn_consume = 0;
633 
634 	vref(udvp);
635 	if ((error = relookup(udvp, &vp, &cn, 0)) != 0)
636 		goto unionfs_vn_create_on_upper_free_out2;
637 	vrele(udvp);
638 
639 	if (vp != NULLVP) {
640 		if (vp == udvp)
641 			vrele(vp);
642 		else
643 			vput(vp);
644 		error = EEXIST;
645 		goto unionfs_vn_create_on_upper_free_out1;
646 	}
647 
648 	if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
649 		goto unionfs_vn_create_on_upper_free_out1;
650 
651 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0) {
652 		vput(vp);
653 		goto unionfs_vn_create_on_upper_free_out1;
654 	}
655 	vp->v_writecount++;
656 	*vpp = vp;
657 
658 unionfs_vn_create_on_upper_free_out1:
659 	VOP_UNLOCK(udvp);
660 
661 unionfs_vn_create_on_upper_free_out2:
662 	PNBUF_PUT(pnbuf);
663 
664 	return (error);
665 }
666 
667 /*
668  * Copy from lvp to uvp.
669  *
670  * lvp and uvp should be locked and opened on entry and will be locked and
671  * opened on return.
672  */
673 static int
674 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
675 		      kauth_cred_t cred)
676 {
677 	int		error;
678 	off_t		offset;
679 	int		count;
680 	int		bufoffset;
681 	char           *buf;
682 	struct uio	uio;
683 	struct iovec	iov;
684 
685 	error = 0;
686 	memset(&uio, 0, sizeof(uio));
687 	UIO_SETUP_SYSSPACE(&uio);
688 	uio.uio_offset = 0;
689 
690 	buf = kmem_alloc(MAXBSIZE, KM_SLEEP);
691 	while (error == 0) {
692 		offset = uio.uio_offset;
693 
694 		uio.uio_iov = &iov;
695 		uio.uio_iovcnt = 1;
696 		iov.iov_base = buf;
697 		iov.iov_len = MAXBSIZE;
698 		uio.uio_resid = iov.iov_len;
699 		uio.uio_rw = UIO_READ;
700 
701 		if ((error = VOP_READ(lvp, &uio, 0, cred)) != 0)
702 			break;
703 		if ((count = MAXBSIZE - uio.uio_resid) == 0)
704 			break;
705 
706 		bufoffset = 0;
707 		while (bufoffset < count) {
708 			uio.uio_iov = &iov;
709 			uio.uio_iovcnt = 1;
710 			iov.iov_base = buf + bufoffset;
711 			iov.iov_len = count - bufoffset;
712 			uio.uio_offset = offset + bufoffset;
713 			uio.uio_resid = iov.iov_len;
714 			uio.uio_rw = UIO_WRITE;
715 
716 			if ((error = VOP_WRITE(uvp, &uio, 0, cred)) != 0)
717 				break;
718 
719 			bufoffset += (count - bufoffset) - uio.uio_resid;
720 		}
721 
722 		uio.uio_offset = offset + bufoffset;
723 	}
724 
725 	kmem_free(buf, MAXBSIZE);
726 
727 	return (error);
728 }
729 
730 /*
731  * Copy file from lower to upper.
732  *
733  * If you need copy of the contents, set 1 to docopy. Otherwise, set 0 to
734  * docopy.
735  *
736  * If no error returned, unp will be updated.
737  */
738 int
739 unionfs_copyfile(struct unionfs_node *unp, int docopy, kauth_cred_t cred)
740 {
741 	int		error;
742 	struct vnode   *udvp;
743 	struct vnode   *lvp;
744 	struct vnode   *uvp;
745 	struct vattr	uva;
746 
747 	lvp = unp->un_lowervp;
748 	uvp = NULLVP;
749 
750 	if ((UNIONFSTOV(unp)->v_mount->mnt_flag & MNT_RDONLY))
751 		return (EROFS);
752 	if (unp->un_dvp == NULLVP)
753 		return (EINVAL);
754 	if (unp->un_uppervp != NULLVP)
755 		return (EEXIST);
756 	udvp = VTOUNIONFS(unp->un_dvp)->un_uppervp;
757 	if (udvp == NULLVP)
758 		return (EROFS);
759 	if ((udvp->v_mount->mnt_flag & MNT_RDONLY))
760 		return (EROFS);
761 
762 	error = VOP_ACCESS(lvp, VREAD, cred);
763 	if (error != 0)
764 		return (error);
765 
766 	error = unionfs_vn_create_on_upper(&uvp, udvp, unp, &uva);
767 	if (error != 0)
768 		return (error);
769 
770 	if (docopy != 0) {
771 		error = VOP_OPEN(lvp, FREAD, cred);
772 		if (error == 0) {
773 			error = unionfs_copyfile_core(lvp, uvp, cred);
774 			VOP_CLOSE(lvp, FREAD, cred);
775 		}
776 	}
777 	VOP_CLOSE(uvp, FWRITE, cred);
778 	uvp->v_writecount--;
779 
780 	if (error == 0) {
781 		/* Reset the attributes. Ignore errors. */
782 		uva.va_type = VNON;
783 		VOP_SETATTR(uvp, &uva, cred);
784 	}
785 
786 	unionfs_node_update(unp, uvp);
787 
788 	return (error);
789 }
790 
791 /*
792  * It checks whether vp can rmdir. (check empty)
793  *
794  * vp is unionfs vnode.
795  * vp should be locked.
796  */
797 int
798 unionfs_check_rmdir(struct vnode *vp, kauth_cred_t cred)
799 {
800 	int		error;
801 	int		eofflag;
802 	int		lookuperr;
803 	struct vnode   *uvp;
804 	struct vnode   *lvp;
805 	struct vnode   *tvp;
806 	struct vattr	va;
807 	struct componentname cn;
808 	/*
809 	 * The size of buf needs to be larger than DIRBLKSIZ.
810 	 */
811 	char		buf[256 * 6];
812 	struct dirent  *dp;
813 	struct dirent  *edp;
814 	struct uio	uio;
815 	struct iovec	iov;
816 
817 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
818 
819 	eofflag = 0;
820 	uvp = UNIONFSVPTOUPPERVP(vp);
821 	lvp = UNIONFSVPTOLOWERVP(vp);
822 
823 	/* check opaque */
824 	if ((error = VOP_GETATTR(uvp, &va, cred)) != 0)
825 		return (error);
826 	if (va.va_flags & OPAQUE)
827 		return (0);
828 
829 	/* open vnode */
830 	if ((error = VOP_ACCESS(vp, VEXEC|VREAD, cred)) != 0)
831 		return (error);
832 	if ((error = VOP_OPEN(vp, FREAD, cred)) != 0)
833 		return (error);
834 
835 	UIO_SETUP_SYSSPACE(&uio);
836 	uio.uio_rw = UIO_READ;
837 	uio.uio_offset = 0;
838 
839 	while (!error && !eofflag) {
840 		iov.iov_base = buf;
841 		iov.iov_len = sizeof(buf);
842 		uio.uio_iov = &iov;
843 		uio.uio_iovcnt = 1;
844 		uio.uio_resid = iov.iov_len;
845 
846 		error = VOP_READDIR(lvp, &uio, cred, &eofflag, NULL, NULL);
847 		if (error)
848 			break;
849 
850 		edp = (struct dirent*)&buf[sizeof(buf) - uio.uio_resid];
851 		for (dp = (struct dirent*)buf; !error && dp < edp;
852 		     dp = (struct dirent*)((char *)dp + dp->d_reclen)) {
853 			if (dp->d_type == DT_WHT ||
854 			    (dp->d_namlen == 1 && dp->d_name[0] == '.') ||
855 			    (dp->d_namlen == 2 && !memcmp(dp->d_name, "..", 2)))
856 				continue;
857 
858 			cn.cn_namelen = dp->d_namlen;
859 			cn.cn_nameptr = dp->d_name;
860 			cn.cn_nameiop = LOOKUP;
861 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
862 			cn.cn_cred = cred;
863 			cn.cn_consume = 0;
864 
865 			/*
866 			 * check entry in lower.
867 			 * Sometimes, readdir function returns
868 			 * wrong entry.
869 			 */
870 			lookuperr = VOP_LOOKUP(lvp, &tvp, &cn);
871 
872 			if (!lookuperr)
873 				vput(tvp);
874 			else
875 				continue; /* skip entry */
876 
877 			/*
878 			 * check entry
879 			 * If it has no exist/whiteout entry in upper,
880 			 * directory is not empty.
881 			 */
882 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
883 			lookuperr = VOP_LOOKUP(uvp, &tvp, &cn);
884 
885 			if (!lookuperr)
886 				vput(tvp);
887 
888 			/* ignore exist or whiteout entry */
889 			if (!lookuperr ||
890 			    (lookuperr == ENOENT && (cn.cn_flags & ISWHITEOUT)))
891 				continue;
892 
893 			error = ENOTEMPTY;
894 		}
895 	}
896 
897 	/* close vnode */
898 	VOP_CLOSE(vp, FREAD, cred);
899 
900 	return (error);
901 }
902 
903 #ifdef DIAGNOSTIC
904 
905 struct vnode   *
906 unionfs_checkuppervp(struct vnode *vp, const char *fil, int lno)
907 {
908 	struct unionfs_node *unp;
909 
910 	unp = VTOUNIONFS(vp);
911 
912 #ifdef notyet
913 	if (vp->v_op != unionfs_vnodeop_p) {
914 		printf("unionfs_checkuppervp: on non-unionfs-node.\n");
915 #ifdef KDB
916 		kdb_enter(KDB_WHY_UNIONFS,
917 		    "unionfs_checkuppervp: on non-unionfs-node.\n");
918 #endif
919 		panic("unionfs_checkuppervp");
920 	};
921 #endif
922 	return (unp->un_uppervp);
923 }
924 
925 struct vnode   *
926 unionfs_checklowervp(struct vnode *vp, const char *fil, int lno)
927 {
928 	struct unionfs_node *unp;
929 
930 	unp = VTOUNIONFS(vp);
931 
932 #ifdef notyet
933 	if (vp->v_op != unionfs_vnodeop_p) {
934 		printf("unionfs_checklowervp: on non-unionfs-node.\n");
935 #ifdef KDB
936 		kdb_enter(KDB_WHY_UNIONFS,
937 		    "unionfs_checklowervp: on non-unionfs-node.\n");
938 #endif
939 		panic("unionfs_checklowervp");
940 	};
941 #endif
942 	return (unp->un_lowervp);
943 }
944 #endif
945