xref: /netbsd-src/sys/fs/unionfs/unionfs_subr.c (revision 53b02e147d4ed531c0d2a5ca9b3e8026ba3e99b5)
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 
321 	if (nameiop == DELETE)
322 		cn->cn_flags |= (cnp->cn_flags & DOWHITEOUT);
323 
324 	vref(dvp);
325 	VOP_UNLOCK(dvp);
326 
327 	if ((error = relookup(dvp, vpp, cn, 0))) {
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 
632 	vref(udvp);
633 	if ((error = relookup(udvp, &vp, &cn, 0)) != 0)
634 		goto unionfs_vn_create_on_upper_free_out2;
635 	vrele(udvp);
636 
637 	if (vp != NULLVP) {
638 		if (vp == udvp)
639 			vrele(vp);
640 		else
641 			vput(vp);
642 		error = EEXIST;
643 		goto unionfs_vn_create_on_upper_free_out1;
644 	}
645 
646 	if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
647 		goto unionfs_vn_create_on_upper_free_out1;
648 
649 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0) {
650 		vput(vp);
651 		goto unionfs_vn_create_on_upper_free_out1;
652 	}
653 	vp->v_writecount++;
654 	*vpp = vp;
655 
656 unionfs_vn_create_on_upper_free_out1:
657 	VOP_UNLOCK(udvp);
658 
659 unionfs_vn_create_on_upper_free_out2:
660 	PNBUF_PUT(pnbuf);
661 
662 	return (error);
663 }
664 
665 /*
666  * Copy from lvp to uvp.
667  *
668  * lvp and uvp should be locked and opened on entry and will be locked and
669  * opened on return.
670  */
671 static int
672 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
673 		      kauth_cred_t cred)
674 {
675 	int		error;
676 	off_t		offset;
677 	int		count;
678 	int		bufoffset;
679 	char           *buf;
680 	struct uio	uio;
681 	struct iovec	iov;
682 
683 	error = 0;
684 	memset(&uio, 0, sizeof(uio));
685 	UIO_SETUP_SYSSPACE(&uio);
686 	uio.uio_offset = 0;
687 
688 	buf = kmem_alloc(MAXBSIZE, KM_SLEEP);
689 	while (error == 0) {
690 		offset = uio.uio_offset;
691 
692 		uio.uio_iov = &iov;
693 		uio.uio_iovcnt = 1;
694 		iov.iov_base = buf;
695 		iov.iov_len = MAXBSIZE;
696 		uio.uio_resid = iov.iov_len;
697 		uio.uio_rw = UIO_READ;
698 
699 		if ((error = VOP_READ(lvp, &uio, 0, cred)) != 0)
700 			break;
701 		if ((count = MAXBSIZE - uio.uio_resid) == 0)
702 			break;
703 
704 		bufoffset = 0;
705 		while (bufoffset < count) {
706 			uio.uio_iov = &iov;
707 			uio.uio_iovcnt = 1;
708 			iov.iov_base = buf + bufoffset;
709 			iov.iov_len = count - bufoffset;
710 			uio.uio_offset = offset + bufoffset;
711 			uio.uio_resid = iov.iov_len;
712 			uio.uio_rw = UIO_WRITE;
713 
714 			if ((error = VOP_WRITE(uvp, &uio, 0, cred)) != 0)
715 				break;
716 
717 			bufoffset += (count - bufoffset) - uio.uio_resid;
718 		}
719 
720 		uio.uio_offset = offset + bufoffset;
721 	}
722 
723 	kmem_free(buf, MAXBSIZE);
724 
725 	return (error);
726 }
727 
728 /*
729  * Copy file from lower to upper.
730  *
731  * If you need copy of the contents, set 1 to docopy. Otherwise, set 0 to
732  * docopy.
733  *
734  * If no error returned, unp will be updated.
735  */
736 int
737 unionfs_copyfile(struct unionfs_node *unp, int docopy, kauth_cred_t cred)
738 {
739 	int		error;
740 	struct vnode   *udvp;
741 	struct vnode   *lvp;
742 	struct vnode   *uvp;
743 	struct vattr	uva;
744 
745 	lvp = unp->un_lowervp;
746 	uvp = NULLVP;
747 
748 	if ((UNIONFSTOV(unp)->v_mount->mnt_flag & MNT_RDONLY))
749 		return (EROFS);
750 	if (unp->un_dvp == NULLVP)
751 		return (EINVAL);
752 	if (unp->un_uppervp != NULLVP)
753 		return (EEXIST);
754 	udvp = VTOUNIONFS(unp->un_dvp)->un_uppervp;
755 	if (udvp == NULLVP)
756 		return (EROFS);
757 	if ((udvp->v_mount->mnt_flag & MNT_RDONLY))
758 		return (EROFS);
759 
760 	error = VOP_ACCESS(lvp, VREAD, cred);
761 	if (error != 0)
762 		return (error);
763 
764 	error = unionfs_vn_create_on_upper(&uvp, udvp, unp, &uva);
765 	if (error != 0)
766 		return (error);
767 
768 	if (docopy != 0) {
769 		error = VOP_OPEN(lvp, FREAD, cred);
770 		if (error == 0) {
771 			error = unionfs_copyfile_core(lvp, uvp, cred);
772 			VOP_CLOSE(lvp, FREAD, cred);
773 		}
774 	}
775 	VOP_CLOSE(uvp, FWRITE, cred);
776 	uvp->v_writecount--;
777 
778 	if (error == 0) {
779 		/* Reset the attributes. Ignore errors. */
780 		uva.va_type = VNON;
781 		VOP_SETATTR(uvp, &uva, cred);
782 	}
783 
784 	unionfs_node_update(unp, uvp);
785 
786 	return (error);
787 }
788 
789 /*
790  * It checks whether vp can rmdir. (check empty)
791  *
792  * vp is unionfs vnode.
793  * vp should be locked.
794  */
795 int
796 unionfs_check_rmdir(struct vnode *vp, kauth_cred_t cred)
797 {
798 	int		error;
799 	int		eofflag;
800 	int		lookuperr;
801 	struct vnode   *uvp;
802 	struct vnode   *lvp;
803 	struct vnode   *tvp;
804 	struct vattr	va;
805 	struct componentname cn;
806 	/*
807 	 * The size of buf needs to be larger than DIRBLKSIZ.
808 	 */
809 	char		buf[256 * 6];
810 	struct dirent  *dp;
811 	struct dirent  *edp;
812 	struct uio	uio;
813 	struct iovec	iov;
814 
815 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
816 
817 	eofflag = 0;
818 	uvp = UNIONFSVPTOUPPERVP(vp);
819 	lvp = UNIONFSVPTOLOWERVP(vp);
820 
821 	/* check opaque */
822 	if ((error = VOP_GETATTR(uvp, &va, cred)) != 0)
823 		return (error);
824 	if (va.va_flags & OPAQUE)
825 		return (0);
826 
827 	/* open vnode */
828 	if ((error = VOP_ACCESS(vp, VEXEC|VREAD, cred)) != 0)
829 		return (error);
830 	if ((error = VOP_OPEN(vp, FREAD, cred)) != 0)
831 		return (error);
832 
833 	UIO_SETUP_SYSSPACE(&uio);
834 	uio.uio_rw = UIO_READ;
835 	uio.uio_offset = 0;
836 
837 	while (!error && !eofflag) {
838 		iov.iov_base = buf;
839 		iov.iov_len = sizeof(buf);
840 		uio.uio_iov = &iov;
841 		uio.uio_iovcnt = 1;
842 		uio.uio_resid = iov.iov_len;
843 
844 		error = VOP_READDIR(lvp, &uio, cred, &eofflag, NULL, NULL);
845 		if (error)
846 			break;
847 
848 		edp = (struct dirent*)&buf[sizeof(buf) - uio.uio_resid];
849 		for (dp = (struct dirent*)buf; !error && dp < edp;
850 		     dp = (struct dirent*)((char *)dp + dp->d_reclen)) {
851 			if (dp->d_type == DT_WHT ||
852 			    (dp->d_namlen == 1 && dp->d_name[0] == '.') ||
853 			    (dp->d_namlen == 2 && !memcmp(dp->d_name, "..", 2)))
854 				continue;
855 
856 			cn.cn_namelen = dp->d_namlen;
857 			cn.cn_nameptr = dp->d_name;
858 			cn.cn_nameiop = LOOKUP;
859 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
860 			cn.cn_cred = cred;
861 
862 			/*
863 			 * check entry in lower.
864 			 * Sometimes, readdir function returns
865 			 * wrong entry.
866 			 */
867 			lookuperr = VOP_LOOKUP(lvp, &tvp, &cn);
868 
869 			if (!lookuperr)
870 				vput(tvp);
871 			else
872 				continue; /* skip entry */
873 
874 			/*
875 			 * check entry
876 			 * If it has no exist/whiteout entry in upper,
877 			 * directory is not empty.
878 			 */
879 			cn.cn_flags = (LOCKPARENT | LOCKLEAF | RDONLY | ISLASTCN);
880 			lookuperr = VOP_LOOKUP(uvp, &tvp, &cn);
881 
882 			if (!lookuperr)
883 				vput(tvp);
884 
885 			/* ignore exist or whiteout entry */
886 			if (!lookuperr ||
887 			    (lookuperr == ENOENT && (cn.cn_flags & ISWHITEOUT)))
888 				continue;
889 
890 			error = ENOTEMPTY;
891 		}
892 	}
893 
894 	/* close vnode */
895 	VOP_CLOSE(vp, FREAD, cred);
896 
897 	return (error);
898 }
899 
900 #ifdef DIAGNOSTIC
901 
902 struct vnode   *
903 unionfs_checkuppervp(struct vnode *vp, const char *fil, int lno)
904 {
905 	struct unionfs_node *unp;
906 
907 	unp = VTOUNIONFS(vp);
908 
909 #ifdef notyet
910 	if (vp->v_op != unionfs_vnodeop_p) {
911 		printf("unionfs_checkuppervp: on non-unionfs-node.\n");
912 #ifdef KDB
913 		kdb_enter(KDB_WHY_UNIONFS,
914 		    "unionfs_checkuppervp: on non-unionfs-node.\n");
915 #endif
916 		panic("unionfs_checkuppervp");
917 	};
918 #endif
919 	return (unp->un_uppervp);
920 }
921 
922 struct vnode   *
923 unionfs_checklowervp(struct vnode *vp, const char *fil, int lno)
924 {
925 	struct unionfs_node *unp;
926 
927 	unp = VTOUNIONFS(vp);
928 
929 #ifdef notyet
930 	if (vp->v_op != unionfs_vnodeop_p) {
931 		printf("unionfs_checklowervp: on non-unionfs-node.\n");
932 #ifdef KDB
933 		kdb_enter(KDB_WHY_UNIONFS,
934 		    "unionfs_checklowervp: on non-unionfs-node.\n");
935 #endif
936 		panic("unionfs_checklowervp");
937 	};
938 #endif
939 	return (unp->un_lowervp);
940 }
941 #endif
942