xref: /netbsd-src/sys/fs/unionfs/unionfs_subr.c (revision 46f5119e40af2e51998f686b2fdcc76b5488f7f3)
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);
321 
322 	vref(dvp);
323 	VOP_UNLOCK(dvp);
324 
325 	if ((error = relookup(dvp, vpp, cn, 0))) {
326 		PNBUF_PUT(pnbuf);
327 		*pnbuf_ret = NULL;
328 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
329 	} else {
330 		*pnbuf_ret = pnbuf;
331 		vrele(dvp);
332 	}
333 
334 	return (error);
335 }
336 
337 /*
338  * relookup for CREATE namei operation.
339  *
340  * dvp is unionfs vnode. dvp should be locked.
341  *
342  * If it called 'unionfs_copyfile' function by unionfs_link etc,
343  * VOP_LOOKUP information is broken.
344  * So it need relookup in order to create link etc.
345  */
346 int
347 unionfs_relookup_for_create(struct vnode *dvp, struct componentname *cnp)
348 {
349 	int	error;
350 	struct vnode *udvp;
351 	struct vnode *vp;
352 	struct componentname cn;
353 	char *pnbuf;
354 
355 	udvp = UNIONFSVPTOUPPERVP(dvp);
356 	vp = NULLVP;
357 
358 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf,
359 	    cnp->cn_nameptr,
360 	    strlen(cnp->cn_nameptr), CREATE);
361 	if (error)
362 		return (error);
363 
364 	if (vp != NULLVP) {
365 		if (udvp == vp)
366 			vrele(vp);
367 		else
368 			vput(vp);
369 
370 		error = EEXIST;
371 	}
372 
373 	PNBUF_PUT(pnbuf);
374 
375 	if (!error) {
376 		cnp->cn_flags = cn.cn_flags;
377 	}
378 
379 	return (error);
380 }
381 
382 /*
383  * relookup for DELETE namei operation.
384  *
385  * dvp is unionfs vnode. dvp should be locked.
386  */
387 int
388 unionfs_relookup_for_delete(struct vnode *dvp, struct componentname *cnp)
389 {
390 	int	error;
391 	struct vnode *udvp;
392 	struct vnode *vp;
393 	struct componentname cn;
394 	char *pnbuf;
395 
396 	udvp = UNIONFSVPTOUPPERVP(dvp);
397 	vp = NULLVP;
398 
399 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
400 	    strlen(cnp->cn_nameptr), DELETE);
401 	if (error)
402 		return (error);
403 
404 	if (vp == NULLVP)
405 		error = ENOENT;
406 	else {
407 		if (udvp == vp)
408 			vrele(vp);
409 		else
410 			vput(vp);
411 	}
412 
413 	PNBUF_PUT(pnbuf);
414 
415 	if (!error) {
416 		cnp->cn_flags = cn.cn_flags;
417 	}
418 
419 	return (error);
420 }
421 
422 /*
423  * relookup for RENAME namei operation.
424  *
425  * dvp is unionfs vnode. dvp should be locked.
426  */
427 int
428 unionfs_relookup_for_rename(struct vnode *dvp, struct componentname *cnp)
429 {
430 	int error;
431 	struct vnode *udvp;
432 	struct vnode *vp;
433 	struct componentname cn;
434 	char *pnbuf;
435 
436 	udvp = UNIONFSVPTOUPPERVP(dvp);
437 	vp = NULLVP;
438 
439 	error = unionfs_relookup(udvp, &vp, cnp, &cn, &pnbuf, cnp->cn_nameptr,
440 	    strlen(cnp->cn_nameptr), RENAME);
441 	if (error)
442 		return (error);
443 
444 	if (vp != NULLVP) {
445 		if (udvp == vp)
446 			vrele(vp);
447 		else
448 			vput(vp);
449 	}
450 
451 	PNBUF_PUT(pnbuf);
452 
453 	if (!error) {
454 		cnp->cn_flags = cn.cn_flags;
455 	}
456 
457 	return (error);
458 
459 }
460 
461 /*
462  * Update the unionfs_node.
463  *
464  * uvp is new locked upper vnode. unionfs vnode's lock will be exchanged to the
465  * uvp's lock and lower's lock will be unlocked.
466  */
467 static void
468 unionfs_node_update(struct unionfs_node *unp, struct vnode *uvp)
469 {
470 	struct vnode   *vp;
471 	struct vnode   *lvp;
472 
473 	vp = UNIONFSTOV(unp);
474 	lvp = unp->un_lowervp;
475 
476 	/*
477 	 * lock update
478 	 */
479 	mutex_enter(&vp->v_interlock);
480 	unp->un_uppervp = uvp;
481 	KASSERT(VOP_ISLOCKED(lvp) == LK_EXCLUSIVE);
482 	mutex_exit(&vp->v_interlock);
483 }
484 
485 /*
486  * Create a new shadow dir.
487  *
488  * udvp should be locked on entry and will be locked on return.
489  *
490  * If no error returned, unp will be updated.
491  */
492 int
493 unionfs_mkshadowdir(struct unionfs_mount *ump, struct vnode *udvp,
494 		    struct unionfs_node *unp, struct componentname *cnp)
495 {
496 	int		error;
497 	struct vnode   *lvp;
498 	struct vnode   *uvp;
499 	struct vattr	va;
500 	struct vattr	lva;
501 	struct componentname cn;
502 	char *pnbuf;
503 
504 	if (unp->un_uppervp != NULLVP)
505 		return (EEXIST);
506 
507 	lvp = unp->un_lowervp;
508 	uvp = NULLVP;
509 
510 	memset(&cn, 0, sizeof(cn));
511 
512 	if ((error = VOP_GETATTR(lvp, &lva, cnp->cn_cred)))
513 		goto unionfs_mkshadowdir_abort;
514 
515 	if ((error = unionfs_relookup(udvp, &uvp, cnp, &cn, &pnbuf,
516 			cnp->cn_nameptr, cnp->cn_namelen, CREATE)))
517 		goto unionfs_mkshadowdir_abort;
518 	if (uvp != NULLVP) {
519 		if (udvp == uvp)
520 			vrele(uvp);
521 		else
522 			vput(uvp);
523 
524 		error = EEXIST;
525 		goto unionfs_mkshadowdir_free_out;
526 	}
527 
528 	unionfs_create_uppervattr_core(ump, &lva, &va);
529 
530 	error = VOP_MKDIR(udvp, &uvp, &cn, &va);
531 
532 	if (!error) {
533 		unionfs_node_update(unp, uvp);
534 
535 		/*
536 		 * XXX The bug which cannot set uid/gid was corrected.
537 		 * Ignore errors.   XXXNETBSD Why is this done as root?
538 		 */
539 		va.va_type = VNON;
540 		VOP_SETATTR(uvp, &va, lwp0.l_cred);
541 	}
542 
543 unionfs_mkshadowdir_free_out:
544 	PNBUF_PUT(pnbuf);
545 
546 unionfs_mkshadowdir_abort:
547 
548 	return (error);
549 }
550 
551 /*
552  * Create a new whiteout.
553  *
554  * dvp should be locked on entry and will be locked on return.
555  */
556 int
557 unionfs_mkwhiteout(struct vnode *dvp, struct componentname *cnp, const char *path)
558 {
559 	int		error;
560 	struct vnode   *wvp;
561 	struct componentname cn;
562 	char *pnbuf;
563 
564 	if (path == NULL)
565 		path = cnp->cn_nameptr;
566 
567 	wvp = NULLVP;
568 	if ((error = unionfs_relookup(dvp, &wvp, cnp, &cn, &pnbuf,
569 			path, strlen(path), CREATE)))
570 		return (error);
571 	if (wvp != NULLVP) {
572 		PNBUF_PUT(pnbuf);
573 		if (dvp == wvp)
574 			vrele(wvp);
575 		else
576 			vput(wvp);
577 
578 		return (EEXIST);
579 	}
580 
581 	PNBUF_PUT(pnbuf);
582 
583 	return (error);
584 }
585 
586 /*
587  * Create a new vnode for create a new shadow file.
588  *
589  * If an error is returned, *vpp will be invalid, otherwise it will hold a
590  * locked, referenced and opened vnode.
591  *
592  * unp is never updated.
593  */
594 static int
595 unionfs_vn_create_on_upper(struct vnode **vpp, struct vnode *udvp,
596 			   struct unionfs_node *unp, struct vattr *uvap)
597 {
598 	struct unionfs_mount *ump;
599 	struct vnode   *vp;
600 	struct vnode   *lvp;
601 	kauth_cred_t    cred;
602 	struct vattr	lva;
603 	int		fmode;
604 	int		error;
605 	struct componentname cn;
606 	char *pnbuf;
607 
608 	ump = MOUNTTOUNIONFSMOUNT(UNIONFSTOV(unp)->v_mount);
609 	vp = NULLVP;
610 	lvp = unp->un_lowervp;
611 	cred = kauth_cred_get();
612 	fmode = FFLAGS(O_WRONLY | O_CREAT | O_TRUNC | O_EXCL);
613 	error = 0;
614 
615 	if ((error = VOP_GETATTR(lvp, &lva, cred)) != 0)
616 		return (error);
617 	unionfs_create_uppervattr_core(ump, &lva, uvap);
618 
619 	if (unp->un_path == NULL)
620 		panic("unionfs: un_path is null");
621 
622 	cn.cn_namelen = strlen(unp->un_path);
623 	pnbuf = PNBUF_GET();
624 	memcpy(pnbuf, unp->un_path, cn.cn_namelen + 1);
625 	cn.cn_nameiop = CREATE;
626 	cn.cn_flags = (LOCKPARENT | LOCKLEAF | ISLASTCN);
627 	cn.cn_cred = cred;
628 	cn.cn_nameptr = pnbuf;
629 	cn.cn_consume = 0;
630 
631 	vref(udvp);
632 	if ((error = relookup(udvp, &vp, &cn, 0)) != 0)
633 		goto unionfs_vn_create_on_upper_free_out2;
634 	vrele(udvp);
635 
636 	if (vp != NULLVP) {
637 		if (vp == udvp)
638 			vrele(vp);
639 		else
640 			vput(vp);
641 		error = EEXIST;
642 		goto unionfs_vn_create_on_upper_free_out1;
643 	}
644 
645 	if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
646 		goto unionfs_vn_create_on_upper_free_out1;
647 
648 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0) {
649 		vput(vp);
650 		goto unionfs_vn_create_on_upper_free_out1;
651 	}
652 	vp->v_writecount++;
653 	*vpp = vp;
654 
655 unionfs_vn_create_on_upper_free_out1:
656 	VOP_UNLOCK(udvp);
657 
658 unionfs_vn_create_on_upper_free_out2:
659 	PNBUF_PUT(pnbuf);
660 
661 	return (error);
662 }
663 
664 /*
665  * Copy from lvp to uvp.
666  *
667  * lvp and uvp should be locked and opened on entry and will be locked and
668  * opened on return.
669  */
670 static int
671 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
672 		      kauth_cred_t cred)
673 {
674 	int		error;
675 	off_t		offset;
676 	int		count;
677 	int		bufoffset;
678 	char           *buf;
679 	struct uio	uio;
680 	struct iovec	iov;
681 
682 	error = 0;
683 	memset(&uio, 0, sizeof(uio));
684 	UIO_SETUP_SYSSPACE(&uio);
685 	uio.uio_offset = 0;
686 
687 	buf = kmem_alloc(MAXBSIZE, KM_SLEEP);
688 	if (buf == NULL)
689 		return ENOMEM;
690 
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