1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1992, 1993, 1994, 1995 Jan-Simon Pendry.
5 * Copyright (c) 1992, 1993, 1994, 1995
6 *      The Regents of the University of California.
7 * Copyright (c) 2005, 2006, 2012 Masanori Ozawa <ozawa@ongs.co.jp>, ONGS Inc.
8 * Copyright (c) 2006, 2012 Daichi Goto <daichi@freebsd.org>
9 * All rights reserved.
10 *
11 * This code is derived from software contributed to Berkeley by
12 * Jan-Simon Pendry.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 *    notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 *    notice, this list of conditions and the following disclaimer in the
21 *    documentation and/or other materials provided with the distribution.
22 * 3. Neither the name of the University nor the names of its contributors
23 *    may be used to endorse or promote products derived from this software
24 *    without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 *	@(#)union_vnops.c	8.32 (Berkeley) 6/23/95
39 * $FreeBSD$
40 *
41 */
42
43#include <sys/param.h>
44#include <sys/systm.h>
45#include <sys/conf.h>
46#include <sys/kernel.h>
47#include <sys/lock.h>
48#include <sys/malloc.h>
49#include <sys/mount.h>
50#include <sys/mutex.h>
51#include <sys/namei.h>
52#include <sys/sysctl.h>
53#include <sys/vnode.h>
54#include <sys/kdb.h>
55#include <sys/fcntl.h>
56#include <sys/stat.h>
57#include <sys/dirent.h>
58#include <sys/proc.h>
59#include <sys/bio.h>
60#include <sys/buf.h>
61
62#include <fs/unionfs/union.h>
63
64#include <vm/vm.h>
65#include <vm/vm_extern.h>
66#include <vm/vm_object.h>
67#include <vm/vnode_pager.h>
68
69#if 0
70#define UNIONFS_INTERNAL_DEBUG(msg, args...)    printf(msg, ## args)
71#define UNIONFS_IDBG_RENAME
72#else
73#define UNIONFS_INTERNAL_DEBUG(msg, args...)
74#endif
75
76#define KASSERT_UNIONFS_VNODE(vp) \
77	KASSERT(((vp)->v_op == &unionfs_vnodeops), \
78	    ("unionfs: it is not unionfs-vnode"))
79
80static int
81unionfs_lookup(struct vop_cachedlookup_args *ap)
82{
83	int		iswhiteout;
84	int		lockflag;
85	int		error , uerror, lerror;
86	u_long		nameiop;
87	u_long		cnflags, cnflagsbk;
88	struct unionfs_node *dunp;
89	struct vnode   *dvp, *udvp, *ldvp, *vp, *uvp, *lvp, *dtmpvp;
90	struct vattr	va;
91	struct componentname *cnp;
92	struct thread  *td;
93
94	iswhiteout = 0;
95	lockflag = 0;
96	error = uerror = lerror = ENOENT;
97	cnp = ap->a_cnp;
98	nameiop = cnp->cn_nameiop;
99	cnflags = cnp->cn_flags;
100	dvp = ap->a_dvp;
101	dunp = VTOUNIONFS(dvp);
102	udvp = dunp->un_uppervp;
103	ldvp = dunp->un_lowervp;
104	vp = uvp = lvp = NULLVP;
105	td = curthread;
106	*(ap->a_vpp) = NULLVP;
107
108	UNIONFS_INTERNAL_DEBUG("unionfs_lookup: enter: nameiop=%ld, flags=%lx, path=%s\n", nameiop, cnflags, cnp->cn_nameptr);
109
110	if (dvp->v_type != VDIR)
111		return (ENOTDIR);
112
113	/*
114	 * If read-only and op is not LOOKUP, will return EROFS.
115	 */
116	if ((cnflags & ISLASTCN) &&
117	    (dvp->v_mount->mnt_flag & MNT_RDONLY) &&
118	    LOOKUP != nameiop)
119		return (EROFS);
120
121	/*
122	 * lookup dotdot
123	 */
124	if (cnflags & ISDOTDOT) {
125		if (LOOKUP != nameiop && udvp == NULLVP)
126			return (EROFS);
127
128		if (udvp != NULLVP) {
129			dtmpvp = udvp;
130			if (ldvp != NULLVP)
131				VOP_UNLOCK(ldvp, LK_RELEASE);
132		}
133		else
134			dtmpvp = ldvp;
135
136		error = VOP_LOOKUP(dtmpvp, &vp, cnp);
137
138		if (dtmpvp == udvp && ldvp != NULLVP) {
139			VOP_UNLOCK(udvp, LK_RELEASE);
140			vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
141		}
142
143		if (error == 0) {
144			/*
145			 * Exchange lock and reference from vp to
146			 * dunp->un_dvp. vp is upper/lower vnode, but it
147			 * will need to return the unionfs vnode.
148			 */
149			if (nameiop == DELETE  || nameiop == RENAME ||
150			    (cnp->cn_lkflags & LK_TYPE_MASK))
151				VOP_UNLOCK(vp, LK_RELEASE);
152			vrele(vp);
153
154			VOP_UNLOCK(dvp, LK_RELEASE);
155			*(ap->a_vpp) = dunp->un_dvp;
156			vref(dunp->un_dvp);
157
158			if (nameiop == DELETE || nameiop == RENAME)
159				vn_lock(dunp->un_dvp, LK_EXCLUSIVE | LK_RETRY);
160			else if (cnp->cn_lkflags & LK_TYPE_MASK)
161				vn_lock(dunp->un_dvp, cnp->cn_lkflags |
162				    LK_RETRY);
163
164			vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
165		} else if (error == ENOENT && (cnflags & MAKEENTRY) != 0)
166			cache_enter(dvp, NULLVP, cnp);
167
168		UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", error);
169
170		return (error);
171	}
172
173	/*
174	 * lookup upper layer
175	 */
176	if (udvp != NULLVP) {
177		uerror = VOP_LOOKUP(udvp, &uvp, cnp);
178
179		if (uerror == 0) {
180			if (udvp == uvp) {	/* is dot */
181				vrele(uvp);
182				*(ap->a_vpp) = dvp;
183				vref(dvp);
184
185				UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", uerror);
186
187				return (uerror);
188			}
189			if (nameiop == DELETE || nameiop == RENAME ||
190			    (cnp->cn_lkflags & LK_TYPE_MASK))
191				VOP_UNLOCK(uvp, LK_RELEASE);
192		}
193
194		/* check whiteout */
195		if (uerror == ENOENT || uerror == EJUSTRETURN)
196			if (cnp->cn_flags & ISWHITEOUT)
197				iswhiteout = 1;	/* don't lookup lower */
198		if (iswhiteout == 0 && ldvp != NULLVP)
199			if (!VOP_GETATTR(udvp, &va, cnp->cn_cred) &&
200			    (va.va_flags & OPAQUE))
201				iswhiteout = 1;	/* don't lookup lower */
202#if 0
203		UNIONFS_INTERNAL_DEBUG("unionfs_lookup: debug: whiteout=%d, path=%s\n", iswhiteout, cnp->cn_nameptr);
204#endif
205	}
206
207	/*
208	 * lookup lower layer
209	 */
210	if (ldvp != NULLVP && !(cnflags & DOWHITEOUT) && iswhiteout == 0) {
211		/* always op is LOOKUP */
212		cnp->cn_nameiop = LOOKUP;
213		cnflagsbk = cnp->cn_flags;
214		cnp->cn_flags = cnflags;
215
216		lerror = VOP_LOOKUP(ldvp, &lvp, cnp);
217
218		cnp->cn_nameiop = nameiop;
219		if (udvp != NULLVP && (uerror == 0 || uerror == EJUSTRETURN))
220			cnp->cn_flags = cnflagsbk;
221
222		if (lerror == 0) {
223			if (ldvp == lvp) {	/* is dot */
224				if (uvp != NULLVP)
225					vrele(uvp);	/* no need? */
226				vrele(lvp);
227				*(ap->a_vpp) = dvp;
228				vref(dvp);
229
230				UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", lerror);
231
232				return (lerror);
233			}
234			if (cnp->cn_lkflags & LK_TYPE_MASK)
235				VOP_UNLOCK(lvp, LK_RELEASE);
236		}
237	}
238
239	/*
240	 * check lookup result
241	 */
242	if (uvp == NULLVP && lvp == NULLVP) {
243		UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n",
244		    (udvp != NULLVP ? uerror : lerror));
245		return (udvp != NULLVP ? uerror : lerror);
246	}
247
248	/*
249	 * check vnode type
250	 */
251	if (uvp != NULLVP && lvp != NULLVP && uvp->v_type != lvp->v_type) {
252		vrele(lvp);
253		lvp = NULLVP;
254	}
255
256	/*
257	 * check shadow dir
258	 */
259	if (uerror != 0 && uerror != EJUSTRETURN && udvp != NULLVP &&
260	    lerror == 0 && lvp != NULLVP && lvp->v_type == VDIR &&
261	    !(dvp->v_mount->mnt_flag & MNT_RDONLY) &&
262	    (1 < cnp->cn_namelen || '.' != *(cnp->cn_nameptr))) {
263		/* get unionfs vnode in order to create a new shadow dir. */
264		error = unionfs_nodeget(dvp->v_mount, NULLVP, lvp, dvp, &vp,
265		    cnp, td);
266		if (error != 0)
267			goto unionfs_lookup_out;
268
269		if (LK_SHARED == (cnp->cn_lkflags & LK_TYPE_MASK))
270			VOP_UNLOCK(vp, LK_RELEASE);
271		if (LK_EXCLUSIVE != VOP_ISLOCKED(vp)) {
272			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
273			lockflag = 1;
274		}
275		error = unionfs_mkshadowdir(MOUNTTOUNIONFSMOUNT(dvp->v_mount),
276		    udvp, VTOUNIONFS(vp), cnp, td);
277		if (lockflag != 0)
278			VOP_UNLOCK(vp, LK_RELEASE);
279		if (error != 0) {
280			UNIONFSDEBUG("unionfs_lookup: Unable to create shadow dir.");
281			if ((cnp->cn_lkflags & LK_TYPE_MASK) == LK_EXCLUSIVE)
282				vput(vp);
283			else
284				vrele(vp);
285			goto unionfs_lookup_out;
286		}
287		if ((cnp->cn_lkflags & LK_TYPE_MASK) == LK_SHARED)
288			vn_lock(vp, LK_SHARED | LK_RETRY);
289	}
290	/*
291	 * get unionfs vnode.
292	 */
293	else {
294		if (uvp != NULLVP)
295			error = uerror;
296		else
297			error = lerror;
298		if (error != 0)
299			goto unionfs_lookup_out;
300		/*
301		 * get socket vnode.
302		 */
303		if (uvp != NULLVP && uvp->v_type == VSOCK) {
304			vp = uvp;
305			vref(vp);
306			if (cnp->cn_lkflags & LK_TYPE_MASK)
307				vn_lock(vp, cnp->cn_lkflags | LK_RETRY);
308		}
309		else if (lvp != NULLVP && lvp->v_type == VSOCK) {
310			vp = lvp;
311			vref(vp);
312			if (cnp->cn_lkflags & LK_TYPE_MASK)
313				vn_lock(vp, cnp->cn_lkflags | LK_RETRY);
314		}
315		/*
316		 * get unionfs vnode.
317		 */
318		else
319			error = unionfs_nodeget(dvp->v_mount, uvp, lvp,
320			    dvp, &vp, cnp, td);
321		if (error != 0) {
322			UNIONFSDEBUG("unionfs_lookup: Unable to create unionfs vnode.");
323			goto unionfs_lookup_out;
324		}
325		if ((nameiop == DELETE || nameiop == RENAME) &&
326		    (cnp->cn_lkflags & LK_TYPE_MASK) == 0)
327			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
328	}
329
330	*(ap->a_vpp) = vp;
331
332	if ((cnflags & MAKEENTRY) && vp->v_type != VSOCK)
333		cache_enter(dvp, vp, cnp);
334
335unionfs_lookup_out:
336	if (uvp != NULLVP)
337		vrele(uvp);
338	if (lvp != NULLVP)
339		vrele(lvp);
340
341	if (error == ENOENT && (cnflags & MAKEENTRY) != 0)
342		cache_enter(dvp, NULLVP, cnp);
343
344	UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", error);
345
346	return (error);
347}
348
349static int
350unionfs_create(struct vop_create_args *ap)
351{
352	struct unionfs_node *dunp;
353	struct componentname *cnp;
354	struct vnode   *udvp;
355	struct vnode   *vp;
356	int		error;
357
358	UNIONFS_INTERNAL_DEBUG("unionfs_create: enter\n");
359
360	KASSERT_UNIONFS_VNODE(ap->a_dvp);
361
362	dunp = VTOUNIONFS(ap->a_dvp);
363	cnp = ap->a_cnp;
364	udvp = dunp->un_uppervp;
365	error = EROFS;
366
367	if (udvp != NULLVP) {
368		error = VOP_CREATE(udvp, &vp, cnp, ap->a_vap);
369		if (error != 0)
370			goto unionfs_create_abort;
371
372		if (vp->v_type == VSOCK)
373			*(ap->a_vpp) = vp;
374		else {
375			VOP_UNLOCK(vp, LK_RELEASE);
376			error = unionfs_nodeget(ap->a_dvp->v_mount, vp, NULLVP,
377			    ap->a_dvp, ap->a_vpp, cnp, curthread);
378			vrele(vp);
379		}
380	}
381
382unionfs_create_abort:
383	UNIONFS_INTERNAL_DEBUG("unionfs_create: leave (%d)\n", error);
384
385	return (error);
386}
387
388static int
389unionfs_whiteout(struct vop_whiteout_args *ap)
390{
391	struct unionfs_node *dunp;
392	struct componentname *cnp;
393	struct vnode   *udvp;
394	int		error;
395
396	UNIONFS_INTERNAL_DEBUG("unionfs_whiteout: enter\n");
397
398	KASSERT_UNIONFS_VNODE(ap->a_dvp);
399
400	dunp = VTOUNIONFS(ap->a_dvp);
401	cnp = ap->a_cnp;
402	udvp = dunp->un_uppervp;
403	error = EOPNOTSUPP;
404
405	if (udvp != NULLVP) {
406		switch (ap->a_flags) {
407		case CREATE:
408		case DELETE:
409		case LOOKUP:
410			error = VOP_WHITEOUT(udvp, cnp, ap->a_flags);
411			break;
412		default:
413			error = EINVAL;
414			break;
415		}
416	}
417
418	UNIONFS_INTERNAL_DEBUG("unionfs_whiteout: leave (%d)\n", error);
419
420	return (error);
421}
422
423static int
424unionfs_mknod(struct vop_mknod_args *ap)
425{
426	struct unionfs_node *dunp;
427	struct componentname *cnp;
428	struct vnode   *udvp;
429	struct vnode   *vp;
430	int		error;
431
432	UNIONFS_INTERNAL_DEBUG("unionfs_mknod: enter\n");
433
434	KASSERT_UNIONFS_VNODE(ap->a_dvp);
435
436	dunp = VTOUNIONFS(ap->a_dvp);
437	cnp = ap->a_cnp;
438	udvp = dunp->un_uppervp;
439	error = EROFS;
440
441	if (udvp != NULLVP) {
442		error = VOP_MKNOD(udvp, &vp, cnp, ap->a_vap);
443		if (error != 0)
444			goto unionfs_mknod_abort;
445
446		if (vp->v_type == VSOCK)
447			*(ap->a_vpp) = vp;
448		else {
449			VOP_UNLOCK(vp, LK_RELEASE);
450			error = unionfs_nodeget(ap->a_dvp->v_mount, vp, NULLVP,
451			    ap->a_dvp, ap->a_vpp, cnp, curthread);
452			vrele(vp);
453		}
454	}
455
456unionfs_mknod_abort:
457	UNIONFS_INTERNAL_DEBUG("unionfs_mknod: leave (%d)\n", error);
458
459	return (error);
460}
461
462static int
463unionfs_open(struct vop_open_args *ap)
464{
465	int		error;
466	struct unionfs_node *unp;
467	struct unionfs_node_status *unsp;
468	struct vnode   *uvp;
469	struct vnode   *lvp;
470	struct vnode   *targetvp;
471	struct ucred   *cred;
472	struct thread  *td;
473
474	UNIONFS_INTERNAL_DEBUG("unionfs_open: enter\n");
475
476	KASSERT_UNIONFS_VNODE(ap->a_vp);
477
478	error = 0;
479	unp = VTOUNIONFS(ap->a_vp);
480	uvp = unp->un_uppervp;
481	lvp = unp->un_lowervp;
482	targetvp = NULLVP;
483	cred = ap->a_cred;
484	td = ap->a_td;
485
486	unionfs_get_node_status(unp, td, &unsp);
487
488	if (unsp->uns_lower_opencnt > 0 || unsp->uns_upper_opencnt > 0) {
489		/* vnode is already opend. */
490		if (unsp->uns_upper_opencnt > 0)
491			targetvp = uvp;
492		else
493			targetvp = lvp;
494
495		if (targetvp == lvp &&
496		    (ap->a_mode & FWRITE) && lvp->v_type == VREG)
497			targetvp = NULLVP;
498	}
499	if (targetvp == NULLVP) {
500		if (uvp == NULLVP) {
501			if ((ap->a_mode & FWRITE) && lvp->v_type == VREG) {
502				error = unionfs_copyfile(unp,
503				    !(ap->a_mode & O_TRUNC), cred, td);
504				if (error != 0)
505					goto unionfs_open_abort;
506				targetvp = uvp = unp->un_uppervp;
507			} else
508				targetvp = lvp;
509		} else
510			targetvp = uvp;
511	}
512
513	error = VOP_OPEN(targetvp, ap->a_mode, cred, td, ap->a_fp);
514	if (error == 0) {
515		if (targetvp == uvp) {
516			if (uvp->v_type == VDIR && lvp != NULLVP &&
517			    unsp->uns_lower_opencnt <= 0) {
518				/* open lower for readdir */
519				error = VOP_OPEN(lvp, FREAD, cred, td, NULL);
520				if (error != 0) {
521					VOP_CLOSE(uvp, ap->a_mode, cred, td);
522					goto unionfs_open_abort;
523				}
524				unsp->uns_node_flag |= UNS_OPENL_4_READDIR;
525				unsp->uns_lower_opencnt++;
526			}
527			unsp->uns_upper_opencnt++;
528		} else {
529			unsp->uns_lower_opencnt++;
530			unsp->uns_lower_openmode = ap->a_mode;
531		}
532		ap->a_vp->v_object = targetvp->v_object;
533	}
534
535unionfs_open_abort:
536	if (error != 0)
537		unionfs_tryrem_node_status(unp, unsp);
538
539	UNIONFS_INTERNAL_DEBUG("unionfs_open: leave (%d)\n", error);
540
541	return (error);
542}
543
544static int
545unionfs_close(struct vop_close_args *ap)
546{
547	int		error;
548	int		locked;
549	struct unionfs_node *unp;
550	struct unionfs_node_status *unsp;
551	struct ucred   *cred;
552	struct thread  *td;
553	struct vnode   *vp;
554	struct vnode   *ovp;
555
556	UNIONFS_INTERNAL_DEBUG("unionfs_close: enter\n");
557
558	KASSERT_UNIONFS_VNODE(ap->a_vp);
559
560	locked = 0;
561	vp = ap->a_vp;
562	unp = VTOUNIONFS(vp);
563	cred = ap->a_cred;
564	td = ap->a_td;
565
566	if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
567		if (vn_lock(vp, LK_UPGRADE) != 0)
568			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
569		locked = 1;
570	}
571	unionfs_get_node_status(unp, td, &unsp);
572
573	if (unsp->uns_lower_opencnt <= 0 && unsp->uns_upper_opencnt <= 0) {
574#ifdef DIAGNOSTIC
575		printf("unionfs_close: warning: open count is 0\n");
576#endif
577		if (unp->un_uppervp != NULLVP)
578			ovp = unp->un_uppervp;
579		else
580			ovp = unp->un_lowervp;
581	} else if (unsp->uns_upper_opencnt > 0)
582		ovp = unp->un_uppervp;
583	else
584		ovp = unp->un_lowervp;
585
586	error = VOP_CLOSE(ovp, ap->a_fflag, cred, td);
587
588	if (error != 0)
589		goto unionfs_close_abort;
590
591	vp->v_object = ovp->v_object;
592
593	if (ovp == unp->un_uppervp) {
594		unsp->uns_upper_opencnt--;
595		if (unsp->uns_upper_opencnt == 0) {
596			if (unsp->uns_node_flag & UNS_OPENL_4_READDIR) {
597				VOP_CLOSE(unp->un_lowervp, FREAD, cred, td);
598				unsp->uns_node_flag &= ~UNS_OPENL_4_READDIR;
599				unsp->uns_lower_opencnt--;
600			}
601			if (unsp->uns_lower_opencnt > 0)
602				vp->v_object = unp->un_lowervp->v_object;
603		}
604	} else
605		unsp->uns_lower_opencnt--;
606
607unionfs_close_abort:
608	unionfs_tryrem_node_status(unp, unsp);
609
610	if (locked != 0)
611		vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
612
613	UNIONFS_INTERNAL_DEBUG("unionfs_close: leave (%d)\n", error);
614
615	return (error);
616}
617
618/*
619 * Check the access mode toward shadow file/dir.
620 */
621static int
622unionfs_check_corrected_access(accmode_t accmode,
623			     struct vattr *va,
624			     struct ucred *cred)
625{
626	int		count;
627	uid_t		uid;	/* upper side vnode's uid */
628	gid_t		gid;	/* upper side vnode's gid */
629	u_short		vmode;	/* upper side vnode's mode */
630	u_short		mask;
631
632	mask = 0;
633	uid = va->va_uid;
634	gid = va->va_gid;
635	vmode = va->va_mode;
636
637	/* check owner */
638	if (cred->cr_uid == uid) {
639		if (accmode & VEXEC)
640			mask |= S_IXUSR;
641		if (accmode & VREAD)
642			mask |= S_IRUSR;
643		if (accmode & VWRITE)
644			mask |= S_IWUSR;
645		return ((vmode & mask) == mask ? 0 : EACCES);
646	}
647
648	/* check group */
649	count = 0;
650	if (groupmember(gid, cred)) {
651		if (accmode & VEXEC)
652			mask |= S_IXGRP;
653		if (accmode & VREAD)
654			mask |= S_IRGRP;
655		if (accmode & VWRITE)
656			mask |= S_IWGRP;
657		return ((vmode & mask) == mask ? 0 : EACCES);
658	}
659
660	/* check other */
661	if (accmode & VEXEC)
662		mask |= S_IXOTH;
663	if (accmode & VREAD)
664		mask |= S_IROTH;
665	if (accmode & VWRITE)
666		mask |= S_IWOTH;
667
668	return ((vmode & mask) == mask ? 0 : EACCES);
669}
670
671static int
672unionfs_access(struct vop_access_args *ap)
673{
674	struct unionfs_mount *ump;
675	struct unionfs_node *unp;
676	struct vnode   *uvp;
677	struct vnode   *lvp;
678	struct thread  *td;
679	struct vattr	va;
680	accmode_t	accmode;
681	int		error;
682
683	UNIONFS_INTERNAL_DEBUG("unionfs_access: enter\n");
684
685	KASSERT_UNIONFS_VNODE(ap->a_vp);
686
687	ump = MOUNTTOUNIONFSMOUNT(ap->a_vp->v_mount);
688	unp = VTOUNIONFS(ap->a_vp);
689	uvp = unp->un_uppervp;
690	lvp = unp->un_lowervp;
691	td = ap->a_td;
692	accmode = ap->a_accmode;
693	error = EACCES;
694
695	if ((accmode & VWRITE) &&
696	    (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)) {
697		switch (ap->a_vp->v_type) {
698		case VREG:
699		case VDIR:
700		case VLNK:
701			return (EROFS);
702		default:
703			break;
704		}
705	}
706
707	if (uvp != NULLVP) {
708		error = VOP_ACCESS(uvp, accmode, ap->a_cred, td);
709
710		UNIONFS_INTERNAL_DEBUG("unionfs_access: leave (%d)\n", error);
711
712		return (error);
713	}
714
715	if (lvp != NULLVP) {
716		if (accmode & VWRITE) {
717			if (ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY) {
718				switch (ap->a_vp->v_type) {
719				case VREG:
720				case VDIR:
721				case VLNK:
722					return (EROFS);
723				default:
724					break;
725				}
726			} else if (ap->a_vp->v_type == VREG || ap->a_vp->v_type == VDIR) {
727				/* check shadow file/dir */
728				if (ump->um_copymode != UNIONFS_TRANSPARENT) {
729					error = unionfs_create_uppervattr(ump,
730					    lvp, &va, ap->a_cred, td);
731					if (error != 0)
732						return (error);
733
734					error = unionfs_check_corrected_access(
735					    accmode, &va, ap->a_cred);
736					if (error != 0)
737						return (error);
738				}
739			}
740			accmode &= ~(VWRITE | VAPPEND);
741			accmode |= VREAD; /* will copy to upper */
742		}
743		error = VOP_ACCESS(lvp, accmode, ap->a_cred, td);
744	}
745
746	UNIONFS_INTERNAL_DEBUG("unionfs_access: leave (%d)\n", error);
747
748	return (error);
749}
750
751static int
752unionfs_getattr(struct vop_getattr_args *ap)
753{
754	int		error;
755	struct unionfs_node *unp;
756	struct unionfs_mount *ump;
757	struct vnode   *uvp;
758	struct vnode   *lvp;
759	struct thread  *td;
760	struct vattr	va;
761
762	UNIONFS_INTERNAL_DEBUG("unionfs_getattr: enter\n");
763
764	KASSERT_UNIONFS_VNODE(ap->a_vp);
765
766	unp = VTOUNIONFS(ap->a_vp);
767	ump = MOUNTTOUNIONFSMOUNT(ap->a_vp->v_mount);
768	uvp = unp->un_uppervp;
769	lvp = unp->un_lowervp;
770	td = curthread;
771
772	if (uvp != NULLVP) {
773		if ((error = VOP_GETATTR(uvp, ap->a_vap, ap->a_cred)) == 0)
774			ap->a_vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
775
776		UNIONFS_INTERNAL_DEBUG("unionfs_getattr: leave mode=%o, uid=%d, gid=%d (%d)\n",
777		    ap->a_vap->va_mode, ap->a_vap->va_uid,
778		    ap->a_vap->va_gid, error);
779
780		return (error);
781	}
782
783	error = VOP_GETATTR(lvp, ap->a_vap, ap->a_cred);
784
785	if (error == 0 && !(ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY)) {
786		/* correct the attr toward shadow file/dir. */
787		if (ap->a_vp->v_type == VREG || ap->a_vp->v_type == VDIR) {
788			unionfs_create_uppervattr_core(ump, ap->a_vap, &va, td);
789			ap->a_vap->va_mode = va.va_mode;
790			ap->a_vap->va_uid = va.va_uid;
791			ap->a_vap->va_gid = va.va_gid;
792		}
793	}
794
795	if (error == 0)
796		ap->a_vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
797
798	UNIONFS_INTERNAL_DEBUG("unionfs_getattr: leave mode=%o, uid=%d, gid=%d (%d)\n",
799	    ap->a_vap->va_mode, ap->a_vap->va_uid, ap->a_vap->va_gid, error);
800
801	return (error);
802}
803
804static int
805unionfs_setattr(struct vop_setattr_args *ap)
806{
807	int		error;
808	struct unionfs_node *unp;
809	struct vnode   *uvp;
810	struct vnode   *lvp;
811	struct thread  *td;
812	struct vattr   *vap;
813
814	UNIONFS_INTERNAL_DEBUG("unionfs_setattr: enter\n");
815
816	KASSERT_UNIONFS_VNODE(ap->a_vp);
817
818	error = EROFS;
819	unp = VTOUNIONFS(ap->a_vp);
820	uvp = unp->un_uppervp;
821	lvp = unp->un_lowervp;
822	td = curthread;
823	vap = ap->a_vap;
824
825	if ((ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) &&
826	    (vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL ||
827	     vap->va_gid != (gid_t)VNOVAL || vap->va_atime.tv_sec != VNOVAL ||
828	     vap->va_mtime.tv_sec != VNOVAL || vap->va_mode != (mode_t)VNOVAL))
829		return (EROFS);
830
831	if (uvp == NULLVP && lvp->v_type == VREG) {
832		error = unionfs_copyfile(unp, (vap->va_size != 0),
833		    ap->a_cred, td);
834		if (error != 0)
835			return (error);
836		uvp = unp->un_uppervp;
837	}
838
839	if (uvp != NULLVP)
840		error = VOP_SETATTR(uvp, vap, ap->a_cred);
841
842	UNIONFS_INTERNAL_DEBUG("unionfs_setattr: leave (%d)\n", error);
843
844	return (error);
845}
846
847static int
848unionfs_read(struct vop_read_args *ap)
849{
850	int		error;
851	struct unionfs_node *unp;
852	struct vnode   *tvp;
853
854	/* UNIONFS_INTERNAL_DEBUG("unionfs_read: enter\n"); */
855
856	KASSERT_UNIONFS_VNODE(ap->a_vp);
857
858	unp = VTOUNIONFS(ap->a_vp);
859	tvp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
860
861	error = VOP_READ(tvp, ap->a_uio, ap->a_ioflag, ap->a_cred);
862
863	/* UNIONFS_INTERNAL_DEBUG("unionfs_read: leave (%d)\n", error); */
864
865	return (error);
866}
867
868static int
869unionfs_write(struct vop_write_args *ap)
870{
871	int		error;
872	struct unionfs_node *unp;
873	struct vnode   *tvp;
874
875	/* UNIONFS_INTERNAL_DEBUG("unionfs_write: enter\n"); */
876
877	KASSERT_UNIONFS_VNODE(ap->a_vp);
878
879	unp = VTOUNIONFS(ap->a_vp);
880	tvp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
881
882	error = VOP_WRITE(tvp, ap->a_uio, ap->a_ioflag, ap->a_cred);
883
884	/* UNIONFS_INTERNAL_DEBUG("unionfs_write: leave (%d)\n", error); */
885
886	return (error);
887}
888
889static int
890unionfs_ioctl(struct vop_ioctl_args *ap)
891{
892	int error;
893	struct unionfs_node *unp;
894	struct unionfs_node_status *unsp;
895	struct vnode   *ovp;
896
897	UNIONFS_INTERNAL_DEBUG("unionfs_ioctl: enter\n");
898
899	KASSERT_UNIONFS_VNODE(ap->a_vp);
900
901 	vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
902	unp = VTOUNIONFS(ap->a_vp);
903	unionfs_get_node_status(unp, ap->a_td, &unsp);
904	ovp = (unsp->uns_upper_opencnt ? unp->un_uppervp : unp->un_lowervp);
905	unionfs_tryrem_node_status(unp, unsp);
906	VOP_UNLOCK(ap->a_vp, LK_RELEASE);
907
908	if (ovp == NULLVP)
909		return (EBADF);
910
911	error = VOP_IOCTL(ovp, ap->a_command, ap->a_data, ap->a_fflag,
912	    ap->a_cred, ap->a_td);
913
914	UNIONFS_INTERNAL_DEBUG("unionfs_ioctl: leave (%d)\n", error);
915
916	return (error);
917}
918
919static int
920unionfs_poll(struct vop_poll_args *ap)
921{
922	struct unionfs_node *unp;
923	struct unionfs_node_status *unsp;
924	struct vnode   *ovp;
925
926	KASSERT_UNIONFS_VNODE(ap->a_vp);
927
928 	vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
929	unp = VTOUNIONFS(ap->a_vp);
930	unionfs_get_node_status(unp, ap->a_td, &unsp);
931	ovp = (unsp->uns_upper_opencnt ? unp->un_uppervp : unp->un_lowervp);
932	unionfs_tryrem_node_status(unp, unsp);
933	VOP_UNLOCK(ap->a_vp, LK_RELEASE);
934
935	if (ovp == NULLVP)
936		return (EBADF);
937
938	return (VOP_POLL(ovp, ap->a_events, ap->a_cred, ap->a_td));
939}
940
941static int
942unionfs_fsync(struct vop_fsync_args *ap)
943{
944	struct unionfs_node *unp;
945	struct unionfs_node_status *unsp;
946	struct vnode   *ovp;
947
948	KASSERT_UNIONFS_VNODE(ap->a_vp);
949
950	unp = VTOUNIONFS(ap->a_vp);
951	unionfs_get_node_status(unp, ap->a_td, &unsp);
952	ovp = (unsp->uns_upper_opencnt ? unp->un_uppervp : unp->un_lowervp);
953	unionfs_tryrem_node_status(unp, unsp);
954
955	if (ovp == NULLVP)
956		return (EBADF);
957
958	return (VOP_FSYNC(ovp, ap->a_waitfor, ap->a_td));
959}
960
961static int
962unionfs_remove(struct vop_remove_args *ap)
963{
964	int		error;
965	char	       *path;
966	struct unionfs_node *dunp;
967	struct unionfs_node *unp;
968	struct unionfs_mount *ump;
969	struct vnode   *udvp;
970	struct vnode   *uvp;
971	struct vnode   *lvp;
972	struct vnode   *vp;
973	struct componentname *cnp;
974	struct componentname cn;
975	struct thread  *td;
976
977	UNIONFS_INTERNAL_DEBUG("unionfs_remove: enter\n");
978
979	KASSERT_UNIONFS_VNODE(ap->a_dvp);
980
981	error = 0;
982	dunp = VTOUNIONFS(ap->a_dvp);
983	udvp = dunp->un_uppervp;
984	cnp = ap->a_cnp;
985	td = curthread;
986
987	if (ap->a_vp->v_op != &unionfs_vnodeops) {
988		if (ap->a_vp->v_type != VSOCK)
989			return (EINVAL);
990		ump = NULL;
991		vp = uvp = lvp = NULLVP;
992		/* search vnode */
993		VOP_UNLOCK(ap->a_vp, LK_RELEASE);
994		error = unionfs_relookup(udvp, &vp, cnp, &cn, td,
995		    cnp->cn_nameptr, strlen(cnp->cn_nameptr), DELETE);
996		if (error != 0 && error != ENOENT) {
997			vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
998			return (error);
999		}
1000
1001		if (error == 0 && vp == ap->a_vp) {
1002			/* target vnode in upper */
1003			uvp = vp;
1004			vrele(vp);
1005			path = NULL;
1006		} else {
1007			/* target vnode in lower */
1008			if (vp != NULLVP) {
1009				if (udvp == vp)
1010					vrele(vp);
1011				else
1012					vput(vp);
1013			}
1014			vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
1015			lvp = ap->a_vp;
1016			path = ap->a_cnp->cn_nameptr;
1017		}
1018	} else {
1019		ump = MOUNTTOUNIONFSMOUNT(ap->a_vp->v_mount);
1020		unp = VTOUNIONFS(ap->a_vp);
1021		uvp = unp->un_uppervp;
1022		lvp = unp->un_lowervp;
1023		path = unp->un_path;
1024	}
1025
1026	if (udvp == NULLVP)
1027		return (EROFS);
1028
1029	if (uvp != NULLVP) {
1030		/*
1031		 * XXX: if the vnode type is VSOCK, it will create whiteout
1032		 *      after remove.
1033		 */
1034		if (ump == NULL || ump->um_whitemode == UNIONFS_WHITE_ALWAYS ||
1035		    lvp != NULLVP)
1036			cnp->cn_flags |= DOWHITEOUT;
1037		error = VOP_REMOVE(udvp, uvp, cnp);
1038	} else if (lvp != NULLVP)
1039		error = unionfs_mkwhiteout(udvp, cnp, td, path);
1040
1041	UNIONFS_INTERNAL_DEBUG("unionfs_remove: leave (%d)\n", error);
1042
1043	return (error);
1044}
1045
1046static int
1047unionfs_link(struct vop_link_args *ap)
1048{
1049	int		error;
1050	int		needrelookup;
1051	struct unionfs_node *dunp;
1052	struct unionfs_node *unp;
1053	struct vnode   *udvp;
1054	struct vnode   *uvp;
1055	struct componentname *cnp;
1056	struct thread  *td;
1057
1058	UNIONFS_INTERNAL_DEBUG("unionfs_link: enter\n");
1059
1060	KASSERT_UNIONFS_VNODE(ap->a_tdvp);
1061	KASSERT_UNIONFS_VNODE(ap->a_vp);
1062
1063	error = 0;
1064	needrelookup = 0;
1065	dunp = VTOUNIONFS(ap->a_tdvp);
1066	unp = NULL;
1067	udvp = dunp->un_uppervp;
1068	uvp = NULLVP;
1069	cnp = ap->a_cnp;
1070	td = curthread;
1071
1072	if (udvp == NULLVP)
1073		return (EROFS);
1074
1075	if (ap->a_vp->v_op != &unionfs_vnodeops)
1076		uvp = ap->a_vp;
1077	else {
1078		unp = VTOUNIONFS(ap->a_vp);
1079
1080		if (unp->un_uppervp == NULLVP) {
1081			if (ap->a_vp->v_type != VREG)
1082				return (EOPNOTSUPP);
1083
1084			error = unionfs_copyfile(unp, 1, cnp->cn_cred, td);
1085			if (error != 0)
1086				return (error);
1087			needrelookup = 1;
1088		}
1089		uvp = unp->un_uppervp;
1090	}
1091
1092	if (needrelookup != 0)
1093		error = unionfs_relookup_for_create(ap->a_tdvp, cnp, td);
1094
1095	if (error == 0)
1096		error = VOP_LINK(udvp, uvp, cnp);
1097
1098	UNIONFS_INTERNAL_DEBUG("unionfs_link: leave (%d)\n", error);
1099
1100	return (error);
1101}
1102
1103static int
1104unionfs_rename(struct vop_rename_args *ap)
1105{
1106	int		error;
1107	struct vnode   *fdvp;
1108	struct vnode   *fvp;
1109	struct componentname *fcnp;
1110	struct vnode   *tdvp;
1111	struct vnode   *tvp;
1112	struct componentname *tcnp;
1113	struct vnode   *ltdvp;
1114	struct vnode   *ltvp;
1115	struct thread  *td;
1116
1117	/* rename target vnodes */
1118	struct vnode   *rfdvp;
1119	struct vnode   *rfvp;
1120	struct vnode   *rtdvp;
1121	struct vnode   *rtvp;
1122
1123	int		needrelookup;
1124	struct unionfs_mount *ump;
1125	struct unionfs_node *unp;
1126
1127	UNIONFS_INTERNAL_DEBUG("unionfs_rename: enter\n");
1128
1129	error = 0;
1130	fdvp = ap->a_fdvp;
1131	fvp = ap->a_fvp;
1132	fcnp = ap->a_fcnp;
1133	tdvp = ap->a_tdvp;
1134	tvp = ap->a_tvp;
1135	tcnp = ap->a_tcnp;
1136	ltdvp = NULLVP;
1137	ltvp = NULLVP;
1138	td = curthread;
1139	rfdvp = fdvp;
1140	rfvp = fvp;
1141	rtdvp = tdvp;
1142	rtvp = tvp;
1143	needrelookup = 0;
1144
1145#ifdef DIAGNOSTIC
1146	if (!(fcnp->cn_flags & HASBUF) || !(tcnp->cn_flags & HASBUF))
1147		panic("unionfs_rename: no name");
1148#endif
1149
1150	/* check for cross device rename */
1151	if (fvp->v_mount != tdvp->v_mount ||
1152	    (tvp != NULLVP && fvp->v_mount != tvp->v_mount)) {
1153		if (fvp->v_op != &unionfs_vnodeops)
1154			error = ENODEV;
1155		else
1156			error = EXDEV;
1157		goto unionfs_rename_abort;
1158	}
1159
1160	/* Renaming a file to itself has no effect. */
1161	if (fvp == tvp)
1162		goto unionfs_rename_abort;
1163
1164	/*
1165	 * from/to vnode is unionfs node.
1166	 */
1167
1168	KASSERT_UNIONFS_VNODE(fdvp);
1169	KASSERT_UNIONFS_VNODE(fvp);
1170	KASSERT_UNIONFS_VNODE(tdvp);
1171	if (tvp != NULLVP)
1172		KASSERT_UNIONFS_VNODE(tvp);
1173
1174	unp = VTOUNIONFS(fdvp);
1175#ifdef UNIONFS_IDBG_RENAME
1176	UNIONFS_INTERNAL_DEBUG("fdvp=%p, ufdvp=%p, lfdvp=%p\n", fdvp, unp->un_uppervp, unp->un_lowervp);
1177#endif
1178	if (unp->un_uppervp == NULLVP) {
1179		error = ENODEV;
1180		goto unionfs_rename_abort;
1181	}
1182	rfdvp = unp->un_uppervp;
1183	vref(rfdvp);
1184
1185	unp = VTOUNIONFS(fvp);
1186#ifdef UNIONFS_IDBG_RENAME
1187	UNIONFS_INTERNAL_DEBUG("fvp=%p, ufvp=%p, lfvp=%p\n", fvp, unp->un_uppervp, unp->un_lowervp);
1188#endif
1189	ump = MOUNTTOUNIONFSMOUNT(fvp->v_mount);
1190	if (unp->un_uppervp == NULLVP) {
1191		switch (fvp->v_type) {
1192		case VREG:
1193			if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
1194				goto unionfs_rename_abort;
1195			error = unionfs_copyfile(unp, 1, fcnp->cn_cred, td);
1196			VOP_UNLOCK(fvp, LK_RELEASE);
1197			if (error != 0)
1198				goto unionfs_rename_abort;
1199			break;
1200		case VDIR:
1201			if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
1202				goto unionfs_rename_abort;
1203			error = unionfs_mkshadowdir(ump, rfdvp, unp, fcnp, td);
1204			VOP_UNLOCK(fvp, LK_RELEASE);
1205			if (error != 0)
1206				goto unionfs_rename_abort;
1207			break;
1208		default:
1209			error = ENODEV;
1210			goto unionfs_rename_abort;
1211		}
1212
1213		needrelookup = 1;
1214	}
1215
1216	if (unp->un_lowervp != NULLVP)
1217		fcnp->cn_flags |= DOWHITEOUT;
1218	rfvp = unp->un_uppervp;
1219	vref(rfvp);
1220
1221	unp = VTOUNIONFS(tdvp);
1222#ifdef UNIONFS_IDBG_RENAME
1223	UNIONFS_INTERNAL_DEBUG("tdvp=%p, utdvp=%p, ltdvp=%p\n", tdvp, unp->un_uppervp, unp->un_lowervp);
1224#endif
1225	if (unp->un_uppervp == NULLVP) {
1226		error = ENODEV;
1227		goto unionfs_rename_abort;
1228	}
1229	rtdvp = unp->un_uppervp;
1230	ltdvp = unp->un_lowervp;
1231	vref(rtdvp);
1232
1233	if (tdvp == tvp) {
1234		rtvp = rtdvp;
1235		vref(rtvp);
1236	} else if (tvp != NULLVP) {
1237		unp = VTOUNIONFS(tvp);
1238#ifdef UNIONFS_IDBG_RENAME
1239		UNIONFS_INTERNAL_DEBUG("tvp=%p, utvp=%p, ltvp=%p\n", tvp, unp->un_uppervp, unp->un_lowervp);
1240#endif
1241		if (unp->un_uppervp == NULLVP)
1242			rtvp = NULLVP;
1243		else {
1244			if (tvp->v_type == VDIR) {
1245				error = EINVAL;
1246				goto unionfs_rename_abort;
1247			}
1248			rtvp = unp->un_uppervp;
1249			ltvp = unp->un_lowervp;
1250			vref(rtvp);
1251		}
1252	}
1253
1254	if (rfvp == rtvp)
1255		goto unionfs_rename_abort;
1256
1257	if (needrelookup != 0) {
1258		if ((error = vn_lock(fdvp, LK_EXCLUSIVE)) != 0)
1259			goto unionfs_rename_abort;
1260		error = unionfs_relookup_for_delete(fdvp, fcnp, td);
1261		VOP_UNLOCK(fdvp, LK_RELEASE);
1262		if (error != 0)
1263			goto unionfs_rename_abort;
1264
1265		/* Locke of tvp is canceled in order to avoid recursive lock. */
1266		if (tvp != NULLVP && tvp != tdvp)
1267			VOP_UNLOCK(tvp, LK_RELEASE);
1268		error = unionfs_relookup_for_rename(tdvp, tcnp, td);
1269		if (tvp != NULLVP && tvp != tdvp)
1270			vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY);
1271		if (error != 0)
1272			goto unionfs_rename_abort;
1273	}
1274
1275	error = VOP_RENAME(rfdvp, rfvp, fcnp, rtdvp, rtvp, tcnp);
1276
1277	if (error == 0) {
1278		if (rtvp != NULLVP && rtvp->v_type == VDIR)
1279			cache_purge(tdvp);
1280		if (fvp->v_type == VDIR && fdvp != tdvp)
1281			cache_purge(fdvp);
1282	}
1283
1284	if (ltdvp != NULLVP)
1285		VOP_UNLOCK(ltdvp, LK_RELEASE);
1286	if (tdvp != rtdvp)
1287		vrele(tdvp);
1288	if (ltvp != NULLVP)
1289		VOP_UNLOCK(ltvp, LK_RELEASE);
1290	if (tvp != rtvp && tvp != NULLVP) {
1291		if (rtvp == NULLVP)
1292			vput(tvp);
1293		else
1294			vrele(tvp);
1295	}
1296	if (fdvp != rfdvp)
1297		vrele(fdvp);
1298	if (fvp != rfvp)
1299		vrele(fvp);
1300
1301	UNIONFS_INTERNAL_DEBUG("unionfs_rename: leave (%d)\n", error);
1302
1303	return (error);
1304
1305unionfs_rename_abort:
1306	vput(tdvp);
1307	if (tdvp != rtdvp)
1308		vrele(rtdvp);
1309	if (tvp != NULLVP) {
1310		if (tdvp != tvp)
1311			vput(tvp);
1312		else
1313			vrele(tvp);
1314	}
1315	if (tvp != rtvp && rtvp != NULLVP)
1316		vrele(rtvp);
1317	if (fdvp != rfdvp)
1318		vrele(rfdvp);
1319	if (fvp != rfvp)
1320		vrele(rfvp);
1321	vrele(fdvp);
1322	vrele(fvp);
1323
1324	UNIONFS_INTERNAL_DEBUG("unionfs_rename: leave (%d)\n", error);
1325
1326	return (error);
1327}
1328
1329static int
1330unionfs_mkdir(struct vop_mkdir_args *ap)
1331{
1332	int		error;
1333	int		lkflags;
1334	struct unionfs_node *dunp;
1335	struct componentname *cnp;
1336	struct thread  *td;
1337	struct vnode   *udvp;
1338	struct vnode   *uvp;
1339	struct vattr	va;
1340
1341	UNIONFS_INTERNAL_DEBUG("unionfs_mkdir: enter\n");
1342
1343	KASSERT_UNIONFS_VNODE(ap->a_dvp);
1344
1345	error = EROFS;
1346	dunp = VTOUNIONFS(ap->a_dvp);
1347	cnp = ap->a_cnp;
1348	lkflags = cnp->cn_lkflags;
1349	td = curthread;
1350	udvp = dunp->un_uppervp;
1351
1352	if (udvp != NULLVP) {
1353		/* check opaque */
1354		if (!(cnp->cn_flags & ISWHITEOUT)) {
1355			error = VOP_GETATTR(udvp, &va, cnp->cn_cred);
1356			if (error != 0)
1357				return (error);
1358			if (va.va_flags & OPAQUE)
1359				cnp->cn_flags |= ISWHITEOUT;
1360		}
1361
1362		if ((error = VOP_MKDIR(udvp, &uvp, cnp, ap->a_vap)) == 0) {
1363			VOP_UNLOCK(uvp, LK_RELEASE);
1364			cnp->cn_lkflags = LK_EXCLUSIVE;
1365			error = unionfs_nodeget(ap->a_dvp->v_mount, uvp, NULLVP,
1366			    ap->a_dvp, ap->a_vpp, cnp, td);
1367			cnp->cn_lkflags = lkflags;
1368			vrele(uvp);
1369		}
1370	}
1371
1372	UNIONFS_INTERNAL_DEBUG("unionfs_mkdir: leave (%d)\n", error);
1373
1374	return (error);
1375}
1376
1377static int
1378unionfs_rmdir(struct vop_rmdir_args *ap)
1379{
1380	int		error;
1381	struct unionfs_node *dunp;
1382	struct unionfs_node *unp;
1383	struct unionfs_mount *ump;
1384	struct componentname *cnp;
1385	struct thread  *td;
1386	struct vnode   *udvp;
1387	struct vnode   *uvp;
1388	struct vnode   *lvp;
1389
1390	UNIONFS_INTERNAL_DEBUG("unionfs_rmdir: enter\n");
1391
1392	KASSERT_UNIONFS_VNODE(ap->a_dvp);
1393	KASSERT_UNIONFS_VNODE(ap->a_vp);
1394
1395	error = 0;
1396	dunp = VTOUNIONFS(ap->a_dvp);
1397	unp = VTOUNIONFS(ap->a_vp);
1398	cnp = ap->a_cnp;
1399	td = curthread;
1400	udvp = dunp->un_uppervp;
1401	uvp = unp->un_uppervp;
1402	lvp = unp->un_lowervp;
1403
1404	if (udvp == NULLVP)
1405		return (EROFS);
1406
1407	if (udvp == uvp)
1408		return (EOPNOTSUPP);
1409
1410	if (uvp != NULLVP) {
1411		if (lvp != NULLVP) {
1412			error = unionfs_check_rmdir(ap->a_vp, cnp->cn_cred, td);
1413			if (error != 0)
1414				return (error);
1415		}
1416		ump = MOUNTTOUNIONFSMOUNT(ap->a_vp->v_mount);
1417		if (ump->um_whitemode == UNIONFS_WHITE_ALWAYS || lvp != NULLVP)
1418			cnp->cn_flags |= DOWHITEOUT;
1419		error = unionfs_relookup_for_delete(ap->a_dvp, cnp, td);
1420		if (!error)
1421			error = VOP_RMDIR(udvp, uvp, cnp);
1422	}
1423	else if (lvp != NULLVP)
1424		error = unionfs_mkwhiteout(udvp, cnp, td, unp->un_path);
1425
1426	if (error == 0) {
1427		cache_purge(ap->a_dvp);
1428		cache_purge(ap->a_vp);
1429	}
1430
1431	UNIONFS_INTERNAL_DEBUG("unionfs_rmdir: leave (%d)\n", error);
1432
1433	return (error);
1434}
1435
1436static int
1437unionfs_symlink(struct vop_symlink_args *ap)
1438{
1439	int		error;
1440	int		lkflags;
1441	struct unionfs_node *dunp;
1442	struct componentname *cnp;
1443	struct thread  *td;
1444	struct vnode   *udvp;
1445	struct vnode   *uvp;
1446
1447	UNIONFS_INTERNAL_DEBUG("unionfs_symlink: enter\n");
1448
1449	KASSERT_UNIONFS_VNODE(ap->a_dvp);
1450
1451	error = EROFS;
1452	dunp = VTOUNIONFS(ap->a_dvp);
1453	cnp = ap->a_cnp;
1454	lkflags = cnp->cn_lkflags;
1455	td = curthread;
1456	udvp = dunp->un_uppervp;
1457
1458	if (udvp != NULLVP) {
1459		error = VOP_SYMLINK(udvp, &uvp, cnp, ap->a_vap, ap->a_target);
1460		if (error == 0) {
1461			VOP_UNLOCK(uvp, LK_RELEASE);
1462			cnp->cn_lkflags = LK_EXCLUSIVE;
1463			error = unionfs_nodeget(ap->a_dvp->v_mount, uvp, NULLVP,
1464			    ap->a_dvp, ap->a_vpp, cnp, td);
1465			cnp->cn_lkflags = lkflags;
1466			vrele(uvp);
1467		}
1468	}
1469
1470	UNIONFS_INTERNAL_DEBUG("unionfs_symlink: leave (%d)\n", error);
1471
1472	return (error);
1473}
1474
1475static int
1476unionfs_readdir(struct vop_readdir_args *ap)
1477{
1478	int		error;
1479	int		eofflag;
1480	int		locked;
1481	int		uio_offset_bk;
1482	struct unionfs_node *unp;
1483	struct unionfs_node_status *unsp;
1484	struct uio     *uio;
1485	struct vnode   *vp;
1486	struct vnode   *uvp;
1487	struct vnode   *lvp;
1488	struct thread  *td;
1489	struct vattr    va;
1490
1491	int		ncookies_bk;
1492	u_long         *cookies_bk;
1493
1494	UNIONFS_INTERNAL_DEBUG("unionfs_readdir: enter\n");
1495
1496	KASSERT_UNIONFS_VNODE(ap->a_vp);
1497
1498	error = 0;
1499	eofflag = 0;
1500	locked = 0;
1501	uio_offset_bk = 0;
1502	uio = ap->a_uio;
1503	uvp = NULLVP;
1504	lvp = NULLVP;
1505	td = uio->uio_td;
1506	ncookies_bk = 0;
1507	cookies_bk = NULL;
1508
1509	vp = ap->a_vp;
1510	if (vp->v_type != VDIR)
1511		return (ENOTDIR);
1512
1513	/* check the open count. unionfs needs to open before readdir. */
1514	if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
1515		if (vn_lock(vp, LK_UPGRADE) != 0)
1516			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1517		locked = 1;
1518	}
1519	unp = VTOUNIONFS(vp);
1520	if (unp == NULL)
1521		error = EBADF;
1522	else {
1523		uvp = unp->un_uppervp;
1524		lvp = unp->un_lowervp;
1525		unionfs_get_node_status(unp, td, &unsp);
1526		if ((uvp != NULLVP && unsp->uns_upper_opencnt <= 0) ||
1527			(lvp != NULLVP && unsp->uns_lower_opencnt <= 0)) {
1528			unionfs_tryrem_node_status(unp, unsp);
1529			error = EBADF;
1530		}
1531	}
1532	if (locked)
1533		vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
1534	if (error != 0)
1535		goto unionfs_readdir_exit;
1536
1537	/* check opaque */
1538	if (uvp != NULLVP && lvp != NULLVP) {
1539		if ((error = VOP_GETATTR(uvp, &va, ap->a_cred)) != 0)
1540			goto unionfs_readdir_exit;
1541		if (va.va_flags & OPAQUE)
1542			lvp = NULLVP;
1543	}
1544
1545	/* upper only */
1546	if (uvp != NULLVP && lvp == NULLVP) {
1547		error = VOP_READDIR(uvp, uio, ap->a_cred, ap->a_eofflag,
1548		    ap->a_ncookies, ap->a_cookies);
1549		unsp->uns_readdir_status = 0;
1550
1551		goto unionfs_readdir_exit;
1552	}
1553
1554	/* lower only */
1555	if (uvp == NULLVP && lvp != NULLVP) {
1556		error = VOP_READDIR(lvp, uio, ap->a_cred, ap->a_eofflag,
1557		    ap->a_ncookies, ap->a_cookies);
1558		unsp->uns_readdir_status = 2;
1559
1560		goto unionfs_readdir_exit;
1561	}
1562
1563	/*
1564	 * readdir upper and lower
1565	 */
1566	KASSERT(uvp != NULLVP, ("unionfs_readdir: null upper vp"));
1567	KASSERT(lvp != NULLVP, ("unionfs_readdir: null lower vp"));
1568	if (uio->uio_offset == 0)
1569		unsp->uns_readdir_status = 0;
1570
1571	if (unsp->uns_readdir_status == 0) {
1572		/* read upper */
1573		error = VOP_READDIR(uvp, uio, ap->a_cred, &eofflag,
1574				    ap->a_ncookies, ap->a_cookies);
1575
1576		if (error != 0 || eofflag == 0)
1577			goto unionfs_readdir_exit;
1578		unsp->uns_readdir_status = 1;
1579
1580		/*
1581		 * UFS(and other FS) needs size of uio_resid larger than
1582		 * DIRBLKSIZ.
1583		 * size of DIRBLKSIZ equals DEV_BSIZE.
1584		 * (see: ufs/ufs/ufs_vnops.c ufs_readdir func , ufs/ufs/dir.h)
1585		 */
1586		if (uio->uio_resid <= (uio->uio_resid & (DEV_BSIZE -1)))
1587			goto unionfs_readdir_exit;
1588
1589		/*
1590		 * Backup cookies.
1591		 * It prepares to readdir in lower.
1592		 */
1593		if (ap->a_ncookies != NULL) {
1594			ncookies_bk = *(ap->a_ncookies);
1595			*(ap->a_ncookies) = 0;
1596		}
1597		if (ap->a_cookies != NULL) {
1598			cookies_bk = *(ap->a_cookies);
1599			*(ap->a_cookies) = NULL;
1600		}
1601	}
1602
1603	/* initialize for readdir in lower */
1604	if (unsp->uns_readdir_status == 1) {
1605		unsp->uns_readdir_status = 2;
1606		/*
1607		 * Backup uio_offset. See the comment after the
1608		 * VOP_READDIR call on the lower layer.
1609		 */
1610		uio_offset_bk = uio->uio_offset;
1611		uio->uio_offset = 0;
1612	}
1613
1614	if (lvp == NULLVP) {
1615		error = EBADF;
1616		goto unionfs_readdir_exit;
1617	}
1618	/* read lower */
1619	error = VOP_READDIR(lvp, uio, ap->a_cred, ap->a_eofflag,
1620			    ap->a_ncookies, ap->a_cookies);
1621
1622	/*
1623	 * We can't return an uio_offset of 0: this would trigger an
1624	 * infinite loop, because the next call to unionfs_readdir would
1625	 * always restart with the upper layer (uio_offset == 0) and
1626	 * always return some data.
1627	 *
1628	 * This happens when the lower layer root directory is removed.
1629	 * (A root directory deleting of unionfs should not be permitted.
1630	 *  But current VFS can not do it.)
1631	 */
1632	if (uio->uio_offset == 0)
1633		uio->uio_offset = uio_offset_bk;
1634
1635	if (cookies_bk != NULL) {
1636		/* merge cookies */
1637		int		size;
1638		u_long         *newcookies, *pos;
1639
1640		size = *(ap->a_ncookies) + ncookies_bk;
1641		newcookies = (u_long *) malloc(size * sizeof(u_long),
1642		    M_TEMP, M_WAITOK);
1643		pos = newcookies;
1644
1645		memcpy(pos, cookies_bk, ncookies_bk * sizeof(u_long));
1646		pos += ncookies_bk;
1647		memcpy(pos, *(ap->a_cookies), *(ap->a_ncookies) * sizeof(u_long));
1648		free(cookies_bk, M_TEMP);
1649		free(*(ap->a_cookies), M_TEMP);
1650		*(ap->a_ncookies) = size;
1651		*(ap->a_cookies) = newcookies;
1652	}
1653
1654unionfs_readdir_exit:
1655	if (error != 0 && ap->a_eofflag != NULL)
1656		*(ap->a_eofflag) = 1;
1657
1658	UNIONFS_INTERNAL_DEBUG("unionfs_readdir: leave (%d)\n", error);
1659
1660	return (error);
1661}
1662
1663static int
1664unionfs_readlink(struct vop_readlink_args *ap)
1665{
1666	int error;
1667	struct unionfs_node *unp;
1668	struct vnode   *vp;
1669
1670	UNIONFS_INTERNAL_DEBUG("unionfs_readlink: enter\n");
1671
1672	KASSERT_UNIONFS_VNODE(ap->a_vp);
1673
1674	unp = VTOUNIONFS(ap->a_vp);
1675	vp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
1676
1677	error = VOP_READLINK(vp, ap->a_uio, ap->a_cred);
1678
1679	UNIONFS_INTERNAL_DEBUG("unionfs_readlink: leave (%d)\n", error);
1680
1681	return (error);
1682}
1683
1684static int
1685unionfs_getwritemount(struct vop_getwritemount_args *ap)
1686{
1687	int		error;
1688	struct vnode   *uvp;
1689	struct vnode   *vp;
1690
1691	UNIONFS_INTERNAL_DEBUG("unionfs_getwritemount: enter\n");
1692
1693	error = 0;
1694	vp = ap->a_vp;
1695
1696	if (vp == NULLVP || (vp->v_mount->mnt_flag & MNT_RDONLY))
1697		return (EACCES);
1698
1699	KASSERT_UNIONFS_VNODE(vp);
1700
1701	uvp = UNIONFSVPTOUPPERVP(vp);
1702	if (uvp == NULLVP && VREG == vp->v_type)
1703		uvp = UNIONFSVPTOUPPERVP(VTOUNIONFS(vp)->un_dvp);
1704
1705	if (uvp != NULLVP)
1706		error = VOP_GETWRITEMOUNT(uvp, ap->a_mpp);
1707	else {
1708		VI_LOCK(vp);
1709		if (vp->v_iflag & VI_FREE)
1710			error = EOPNOTSUPP;
1711		else
1712			error = EACCES;
1713		VI_UNLOCK(vp);
1714	}
1715
1716	UNIONFS_INTERNAL_DEBUG("unionfs_getwritemount: leave (%d)\n", error);
1717
1718	return (error);
1719}
1720
1721static int
1722unionfs_inactive(struct vop_inactive_args *ap)
1723{
1724	ap->a_vp->v_object = NULL;
1725	vrecycle(ap->a_vp);
1726	return (0);
1727}
1728
1729static int
1730unionfs_reclaim(struct vop_reclaim_args *ap)
1731{
1732	/* UNIONFS_INTERNAL_DEBUG("unionfs_reclaim: enter\n"); */
1733
1734	unionfs_noderem(ap->a_vp, ap->a_td);
1735
1736	/* UNIONFS_INTERNAL_DEBUG("unionfs_reclaim: leave\n"); */
1737
1738	return (0);
1739}
1740
1741static int
1742unionfs_print(struct vop_print_args *ap)
1743{
1744	struct unionfs_node *unp;
1745	/* struct unionfs_node_status *unsp; */
1746
1747	unp = VTOUNIONFS(ap->a_vp);
1748	/* unionfs_get_node_status(unp, curthread, &unsp); */
1749
1750	printf("unionfs_vp=%p, uppervp=%p, lowervp=%p\n",
1751	    ap->a_vp, unp->un_uppervp, unp->un_lowervp);
1752	/*
1753	printf("unionfs opencnt: uppervp=%d, lowervp=%d\n",
1754	    unsp->uns_upper_opencnt, unsp->uns_lower_opencnt);
1755	*/
1756
1757	if (unp->un_uppervp != NULLVP)
1758		vn_printf(unp->un_uppervp, "unionfs: upper ");
1759	if (unp->un_lowervp != NULLVP)
1760		vn_printf(unp->un_lowervp, "unionfs: lower ");
1761
1762	return (0);
1763}
1764
1765static int
1766unionfs_islocked(struct vop_islocked_args *ap)
1767{
1768	struct unionfs_node *unp;
1769
1770	KASSERT_UNIONFS_VNODE(ap->a_vp);
1771
1772	unp = VTOUNIONFS(ap->a_vp);
1773	if (unp == NULL)
1774		return (vop_stdislocked(ap));
1775
1776	if (unp->un_uppervp != NULLVP)
1777		return (VOP_ISLOCKED(unp->un_uppervp));
1778	if (unp->un_lowervp != NULLVP)
1779		return (VOP_ISLOCKED(unp->un_lowervp));
1780	return (vop_stdislocked(ap));
1781}
1782
1783static int
1784unionfs_get_llt_revlock(struct vnode *vp, int flags)
1785{
1786	int revlock;
1787
1788	revlock = 0;
1789
1790	switch (flags & LK_TYPE_MASK) {
1791	case LK_SHARED:
1792		if (VOP_ISLOCKED(vp) == LK_EXCLUSIVE)
1793			revlock = LK_UPGRADE;
1794		else
1795			revlock = LK_RELEASE;
1796		break;
1797	case LK_EXCLUSIVE:
1798	case LK_UPGRADE:
1799		revlock = LK_RELEASE;
1800		break;
1801	case LK_DOWNGRADE:
1802		revlock = LK_UPGRADE;
1803		break;
1804	default:
1805		break;
1806	}
1807
1808	return (revlock);
1809}
1810
1811/*
1812 * The state of an acquired lock is adjusted similarly to
1813 * the time of error generating.
1814 * flags: LK_RELEASE or LK_UPGRADE
1815 */
1816static void
1817unionfs_revlock(struct vnode *vp, int flags)
1818{
1819	if (flags & LK_RELEASE)
1820		VOP_UNLOCK(vp, flags);
1821	else {
1822		/* UPGRADE */
1823		if (vn_lock(vp, flags) != 0)
1824			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1825	}
1826}
1827
1828static int
1829unionfs_lock(struct vop_lock1_args *ap)
1830{
1831	int		error;
1832	int		flags;
1833	int		revlock;
1834	int		interlock;
1835	int		uhold;
1836	struct mount   *mp;
1837	struct unionfs_mount *ump;
1838	struct unionfs_node *unp;
1839	struct vnode   *vp;
1840	struct vnode   *uvp;
1841	struct vnode   *lvp;
1842
1843	KASSERT_UNIONFS_VNODE(ap->a_vp);
1844
1845	error = 0;
1846	interlock = 1;
1847	uhold = 0;
1848	flags = ap->a_flags;
1849	vp = ap->a_vp;
1850
1851	if (LK_RELEASE == (flags & LK_TYPE_MASK) || !(flags & LK_TYPE_MASK))
1852		return (VOP_UNLOCK(vp, flags | LK_RELEASE));
1853
1854	if ((flags & LK_INTERLOCK) == 0)
1855		VI_LOCK(vp);
1856
1857	mp = vp->v_mount;
1858	if (mp == NULL)
1859		goto unionfs_lock_null_vnode;
1860
1861	ump = MOUNTTOUNIONFSMOUNT(mp);
1862	unp = VTOUNIONFS(vp);
1863	if (ump == NULL || unp == NULL)
1864		goto unionfs_lock_null_vnode;
1865	lvp = unp->un_lowervp;
1866	uvp = unp->un_uppervp;
1867
1868	if ((revlock = unionfs_get_llt_revlock(vp, flags)) == 0)
1869		panic("unknown lock type: 0x%x", flags & LK_TYPE_MASK);
1870
1871	if ((vp->v_iflag & VI_OWEINACT) != 0)
1872		flags |= LK_NOWAIT;
1873
1874	/*
1875	 * Sometimes, lower or upper is already exclusive locked.
1876	 * (ex. vfs_domount: mounted vnode is already locked.)
1877	 */
1878	if ((flags & LK_TYPE_MASK) == LK_EXCLUSIVE &&
1879	    vp == ump->um_rootvp)
1880		flags |= LK_CANRECURSE;
1881
1882	if (lvp != NULLVP) {
1883		if (uvp != NULLVP && flags & LK_UPGRADE) {
1884			/* Share Lock is once released and a deadlock is avoided.  */
1885			VI_LOCK_FLAGS(uvp, MTX_DUPOK);
1886			vholdl(uvp);
1887			uhold = 1;
1888			VI_UNLOCK(vp);
1889			VOP_UNLOCK(uvp, LK_RELEASE | LK_INTERLOCK);
1890			VI_LOCK(vp);
1891			unp = VTOUNIONFS(vp);
1892			if (unp == NULL) {
1893				/* vnode is released. */
1894				VI_UNLOCK(vp);
1895				VOP_UNLOCK(lvp, LK_RELEASE);
1896				vdrop(uvp);
1897				return (EBUSY);
1898			}
1899		}
1900		VI_LOCK_FLAGS(lvp, MTX_DUPOK);
1901		flags |= LK_INTERLOCK;
1902		vholdl(lvp);
1903
1904		VI_UNLOCK(vp);
1905		ap->a_flags &= ~LK_INTERLOCK;
1906
1907		error = VOP_LOCK(lvp, flags);
1908
1909		VI_LOCK(vp);
1910		unp = VTOUNIONFS(vp);
1911		if (unp == NULL) {
1912			/* vnode is released. */
1913			VI_UNLOCK(vp);
1914			if (error == 0)
1915				VOP_UNLOCK(lvp, LK_RELEASE);
1916			vdrop(lvp);
1917			if (uhold != 0)
1918				vdrop(uvp);
1919			return (vop_stdlock(ap));
1920		}
1921	}
1922
1923	if (error == 0 && uvp != NULLVP) {
1924		if (uhold && flags & LK_UPGRADE) {
1925			flags &= ~LK_TYPE_MASK;
1926			flags |= LK_EXCLUSIVE;
1927		}
1928		VI_LOCK_FLAGS(uvp, MTX_DUPOK);
1929		flags |= LK_INTERLOCK;
1930		if (uhold == 0) {
1931			vholdl(uvp);
1932			uhold = 1;
1933		}
1934
1935		VI_UNLOCK(vp);
1936		ap->a_flags &= ~LK_INTERLOCK;
1937
1938		error = VOP_LOCK(uvp, flags);
1939
1940		VI_LOCK(vp);
1941		unp = VTOUNIONFS(vp);
1942		if (unp == NULL) {
1943			/* vnode is released. */
1944			VI_UNLOCK(vp);
1945			if (error == 0)
1946				VOP_UNLOCK(uvp, LK_RELEASE);
1947			vdrop(uvp);
1948			if (lvp != NULLVP) {
1949				VOP_UNLOCK(lvp, LK_RELEASE);
1950				vdrop(lvp);
1951			}
1952			return (vop_stdlock(ap));
1953		}
1954		if (error != 0 && lvp != NULLVP) {
1955			/* rollback */
1956			VI_UNLOCK(vp);
1957			unionfs_revlock(lvp, revlock);
1958			interlock = 0;
1959		}
1960	}
1961
1962	if (interlock)
1963		VI_UNLOCK(vp);
1964	if (lvp != NULLVP)
1965		vdrop(lvp);
1966	if (uhold != 0)
1967		vdrop(uvp);
1968
1969	return (error);
1970
1971unionfs_lock_null_vnode:
1972	ap->a_flags |= LK_INTERLOCK;
1973	return (vop_stdlock(ap));
1974}
1975
1976static int
1977unionfs_unlock(struct vop_unlock_args *ap)
1978{
1979	int		error;
1980	int		flags;
1981	int		mtxlkflag;
1982	int		uhold;
1983	struct vnode   *vp;
1984	struct vnode   *lvp;
1985	struct vnode   *uvp;
1986	struct unionfs_node *unp;
1987
1988	KASSERT_UNIONFS_VNODE(ap->a_vp);
1989
1990	error = 0;
1991	mtxlkflag = 0;
1992	uhold = 0;
1993	flags = ap->a_flags | LK_RELEASE;
1994	vp = ap->a_vp;
1995
1996	if ((flags & LK_INTERLOCK) != 0)
1997		mtxlkflag = 1;
1998	else if (mtx_owned(VI_MTX(vp)) == 0) {
1999		VI_LOCK(vp);
2000		mtxlkflag = 2;
2001	}
2002
2003	unp = VTOUNIONFS(vp);
2004	if (unp == NULL)
2005		goto unionfs_unlock_null_vnode;
2006	lvp = unp->un_lowervp;
2007	uvp = unp->un_uppervp;
2008
2009	if (lvp != NULLVP) {
2010		VI_LOCK_FLAGS(lvp, MTX_DUPOK);
2011		flags |= LK_INTERLOCK;
2012		vholdl(lvp);
2013
2014		VI_UNLOCK(vp);
2015		ap->a_flags &= ~LK_INTERLOCK;
2016
2017		error = VOP_UNLOCK(lvp, flags);
2018
2019		VI_LOCK(vp);
2020	}
2021
2022	if (error == 0 && uvp != NULLVP) {
2023		VI_LOCK_FLAGS(uvp, MTX_DUPOK);
2024		flags |= LK_INTERLOCK;
2025		vholdl(uvp);
2026		uhold = 1;
2027
2028		VI_UNLOCK(vp);
2029		ap->a_flags &= ~LK_INTERLOCK;
2030
2031		error = VOP_UNLOCK(uvp, flags);
2032
2033		VI_LOCK(vp);
2034	}
2035
2036	VI_UNLOCK(vp);
2037	if (lvp != NULLVP)
2038		vdrop(lvp);
2039	if (uhold != 0)
2040		vdrop(uvp);
2041	if (mtxlkflag == 0)
2042		VI_LOCK(vp);
2043
2044	return error;
2045
2046unionfs_unlock_null_vnode:
2047	if (mtxlkflag == 2)
2048		VI_UNLOCK(vp);
2049	return (vop_stdunlock(ap));
2050}
2051
2052static int
2053unionfs_pathconf(struct vop_pathconf_args *ap)
2054{
2055	struct unionfs_node *unp;
2056	struct vnode   *vp;
2057
2058	KASSERT_UNIONFS_VNODE(ap->a_vp);
2059
2060	unp = VTOUNIONFS(ap->a_vp);
2061	vp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
2062
2063	return (VOP_PATHCONF(vp, ap->a_name, ap->a_retval));
2064}
2065
2066static int
2067unionfs_advlock(struct vop_advlock_args *ap)
2068{
2069	int error;
2070	struct unionfs_node *unp;
2071	struct unionfs_node_status *unsp;
2072	struct vnode   *vp;
2073	struct vnode   *uvp;
2074	struct thread  *td;
2075
2076	UNIONFS_INTERNAL_DEBUG("unionfs_advlock: enter\n");
2077
2078	KASSERT_UNIONFS_VNODE(ap->a_vp);
2079
2080	vp = ap->a_vp;
2081	td = curthread;
2082
2083	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2084
2085	unp = VTOUNIONFS(ap->a_vp);
2086	uvp = unp->un_uppervp;
2087
2088	if (uvp == NULLVP) {
2089		error = unionfs_copyfile(unp, 1, td->td_ucred, td);
2090		if (error != 0)
2091			goto unionfs_advlock_abort;
2092		uvp = unp->un_uppervp;
2093
2094		unionfs_get_node_status(unp, td, &unsp);
2095		if (unsp->uns_lower_opencnt > 0) {
2096			/* try reopen the vnode */
2097			error = VOP_OPEN(uvp, unsp->uns_lower_openmode,
2098				td->td_ucred, td, NULL);
2099			if (error)
2100				goto unionfs_advlock_abort;
2101			unsp->uns_upper_opencnt++;
2102			VOP_CLOSE(unp->un_lowervp, unsp->uns_lower_openmode, td->td_ucred, td);
2103			unsp->uns_lower_opencnt--;
2104		} else
2105			unionfs_tryrem_node_status(unp, unsp);
2106	}
2107
2108	VOP_UNLOCK(vp, LK_RELEASE);
2109
2110	error = VOP_ADVLOCK(uvp, ap->a_id, ap->a_op, ap->a_fl, ap->a_flags);
2111
2112	UNIONFS_INTERNAL_DEBUG("unionfs_advlock: leave (%d)\n", error);
2113
2114	return error;
2115
2116unionfs_advlock_abort:
2117	VOP_UNLOCK(vp, LK_RELEASE);
2118
2119	UNIONFS_INTERNAL_DEBUG("unionfs_advlock: leave (%d)\n", error);
2120
2121	return error;
2122}
2123
2124static int
2125unionfs_strategy(struct vop_strategy_args *ap)
2126{
2127	struct unionfs_node *unp;
2128	struct vnode   *vp;
2129
2130	KASSERT_UNIONFS_VNODE(ap->a_vp);
2131
2132	unp = VTOUNIONFS(ap->a_vp);
2133	vp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
2134
2135#ifdef DIAGNOSTIC
2136	if (vp == NULLVP)
2137		panic("unionfs_strategy: nullvp");
2138
2139	if (ap->a_bp->b_iocmd == BIO_WRITE && vp == unp->un_lowervp)
2140		panic("unionfs_strategy: writing to lowervp");
2141#endif
2142
2143	return (VOP_STRATEGY(vp, ap->a_bp));
2144}
2145
2146static int
2147unionfs_getacl(struct vop_getacl_args *ap)
2148{
2149	int		error;
2150	struct unionfs_node *unp;
2151	struct vnode   *vp;
2152
2153	KASSERT_UNIONFS_VNODE(ap->a_vp);
2154
2155	unp = VTOUNIONFS(ap->a_vp);
2156	vp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
2157
2158	UNIONFS_INTERNAL_DEBUG("unionfs_getacl: enter\n");
2159
2160	error = VOP_GETACL(vp, ap->a_type, ap->a_aclp, ap->a_cred, ap->a_td);
2161
2162	UNIONFS_INTERNAL_DEBUG("unionfs_getacl: leave (%d)\n", error);
2163
2164	return (error);
2165}
2166
2167static int
2168unionfs_setacl(struct vop_setacl_args *ap)
2169{
2170	int		error;
2171	struct unionfs_node *unp;
2172	struct vnode   *uvp;
2173	struct vnode   *lvp;
2174	struct thread  *td;
2175
2176	UNIONFS_INTERNAL_DEBUG("unionfs_setacl: enter\n");
2177
2178	KASSERT_UNIONFS_VNODE(ap->a_vp);
2179
2180	error = EROFS;
2181	unp = VTOUNIONFS(ap->a_vp);
2182	uvp = unp->un_uppervp;
2183	lvp = unp->un_lowervp;
2184	td = ap->a_td;
2185
2186	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
2187		return (EROFS);
2188
2189	if (uvp == NULLVP && lvp->v_type == VREG) {
2190		if ((error = unionfs_copyfile(unp, 1, ap->a_cred, td)) != 0)
2191			return (error);
2192		uvp = unp->un_uppervp;
2193	}
2194
2195	if (uvp != NULLVP)
2196		error = VOP_SETACL(uvp, ap->a_type, ap->a_aclp, ap->a_cred, td);
2197
2198	UNIONFS_INTERNAL_DEBUG("unionfs_setacl: leave (%d)\n", error);
2199
2200	return (error);
2201}
2202
2203static int
2204unionfs_aclcheck(struct vop_aclcheck_args *ap)
2205{
2206	int		error;
2207	struct unionfs_node *unp;
2208	struct vnode   *vp;
2209
2210	UNIONFS_INTERNAL_DEBUG("unionfs_aclcheck: enter\n");
2211
2212	KASSERT_UNIONFS_VNODE(ap->a_vp);
2213
2214	unp = VTOUNIONFS(ap->a_vp);
2215	vp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
2216
2217	error = VOP_ACLCHECK(vp, ap->a_type, ap->a_aclp, ap->a_cred, ap->a_td);
2218
2219	UNIONFS_INTERNAL_DEBUG("unionfs_aclcheck: leave (%d)\n", error);
2220
2221	return (error);
2222}
2223
2224static int
2225unionfs_openextattr(struct vop_openextattr_args *ap)
2226{
2227	int		error;
2228	struct unionfs_node *unp;
2229	struct vnode   *vp;
2230	struct vnode   *tvp;
2231
2232	KASSERT_UNIONFS_VNODE(ap->a_vp);
2233
2234	vp = ap->a_vp;
2235	unp = VTOUNIONFS(vp);
2236	tvp = (unp->un_uppervp != NULLVP ? unp->un_uppervp : unp->un_lowervp);
2237
2238	if ((tvp == unp->un_uppervp && (unp->un_flag & UNIONFS_OPENEXTU)) ||
2239	    (tvp == unp->un_lowervp && (unp->un_flag & UNIONFS_OPENEXTL)))
2240		return (EBUSY);
2241
2242	error = VOP_OPENEXTATTR(tvp, ap->a_cred, ap->a_td);
2243
2244	if (error == 0) {
2245		if (vn_lock(vp, LK_UPGRADE) != 0)
2246			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2247		if (tvp == unp->un_uppervp)
2248			unp->un_flag |= UNIONFS_OPENEXTU;
2249		else
2250			unp->un_flag |= UNIONFS_OPENEXTL;
2251		vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
2252	}
2253
2254	return (error);
2255}
2256
2257static int
2258unionfs_closeextattr(struct vop_closeextattr_args *ap)
2259{
2260	int		error;
2261	struct unionfs_node *unp;
2262	struct vnode   *vp;
2263	struct vnode   *tvp;
2264
2265	KASSERT_UNIONFS_VNODE(ap->a_vp);
2266
2267	vp = ap->a_vp;
2268	unp = VTOUNIONFS(vp);
2269	tvp = NULLVP;
2270
2271	if (unp->un_flag & UNIONFS_OPENEXTU)
2272		tvp = unp->un_uppervp;
2273	else if (unp->un_flag & UNIONFS_OPENEXTL)
2274		tvp = unp->un_lowervp;
2275
2276	if (tvp == NULLVP)
2277		return (EOPNOTSUPP);
2278
2279	error = VOP_CLOSEEXTATTR(tvp, ap->a_commit, ap->a_cred, ap->a_td);
2280
2281	if (error == 0) {
2282		if (vn_lock(vp, LK_UPGRADE) != 0)
2283			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2284		if (tvp == unp->un_uppervp)
2285			unp->un_flag &= ~UNIONFS_OPENEXTU;
2286		else
2287			unp->un_flag &= ~UNIONFS_OPENEXTL;
2288		vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
2289	}
2290
2291	return (error);
2292}
2293
2294static int
2295unionfs_getextattr(struct vop_getextattr_args *ap)
2296{
2297	struct unionfs_node *unp;
2298	struct vnode   *vp;
2299
2300	KASSERT_UNIONFS_VNODE(ap->a_vp);
2301
2302	unp = VTOUNIONFS(ap->a_vp);
2303	vp = NULLVP;
2304
2305	if (unp->un_flag & UNIONFS_OPENEXTU)
2306		vp = unp->un_uppervp;
2307	else if (unp->un_flag & UNIONFS_OPENEXTL)
2308		vp = unp->un_lowervp;
2309
2310	if (vp == NULLVP)
2311		return (EOPNOTSUPP);
2312
2313	return (VOP_GETEXTATTR(vp, ap->a_attrnamespace, ap->a_name,
2314	    ap->a_uio, ap->a_size, ap->a_cred, ap->a_td));
2315}
2316
2317static int
2318unionfs_setextattr(struct vop_setextattr_args *ap)
2319{
2320	int		error;
2321	struct unionfs_node *unp;
2322	struct vnode   *uvp;
2323	struct vnode   *lvp;
2324	struct vnode   *ovp;
2325	struct ucred   *cred;
2326	struct thread  *td;
2327
2328	KASSERT_UNIONFS_VNODE(ap->a_vp);
2329
2330	error = EROFS;
2331	unp = VTOUNIONFS(ap->a_vp);
2332	uvp = unp->un_uppervp;
2333	lvp = unp->un_lowervp;
2334	ovp = NULLVP;
2335	cred = ap->a_cred;
2336	td = ap->a_td;
2337
2338	UNIONFS_INTERNAL_DEBUG("unionfs_setextattr: enter (un_flag=%x)\n", unp->un_flag);
2339
2340	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
2341		return (EROFS);
2342
2343	if (unp->un_flag & UNIONFS_OPENEXTU)
2344		ovp = unp->un_uppervp;
2345	else if (unp->un_flag & UNIONFS_OPENEXTL)
2346		ovp = unp->un_lowervp;
2347
2348	if (ovp == NULLVP)
2349		return (EOPNOTSUPP);
2350
2351	if (ovp == lvp && lvp->v_type == VREG) {
2352		VOP_CLOSEEXTATTR(lvp, 0, cred, td);
2353		if (uvp == NULLVP &&
2354		    (error = unionfs_copyfile(unp, 1, cred, td)) != 0) {
2355unionfs_setextattr_reopen:
2356			if ((unp->un_flag & UNIONFS_OPENEXTL) &&
2357			    VOP_OPENEXTATTR(lvp, cred, td)) {
2358#ifdef DIAGNOSTIC
2359				panic("unionfs: VOP_OPENEXTATTR failed");
2360#endif
2361				unp->un_flag &= ~UNIONFS_OPENEXTL;
2362			}
2363			goto unionfs_setextattr_abort;
2364		}
2365		uvp = unp->un_uppervp;
2366		if ((error = VOP_OPENEXTATTR(uvp, cred, td)) != 0)
2367			goto unionfs_setextattr_reopen;
2368		unp->un_flag &= ~UNIONFS_OPENEXTL;
2369		unp->un_flag |= UNIONFS_OPENEXTU;
2370		ovp = uvp;
2371	}
2372
2373	if (ovp == uvp)
2374		error = VOP_SETEXTATTR(ovp, ap->a_attrnamespace, ap->a_name,
2375		    ap->a_uio, cred, td);
2376
2377unionfs_setextattr_abort:
2378	UNIONFS_INTERNAL_DEBUG("unionfs_setextattr: leave (%d)\n", error);
2379
2380	return (error);
2381}
2382
2383static int
2384unionfs_listextattr(struct vop_listextattr_args *ap)
2385{
2386	struct unionfs_node *unp;
2387	struct vnode   *vp;
2388
2389	KASSERT_UNIONFS_VNODE(ap->a_vp);
2390
2391	unp = VTOUNIONFS(ap->a_vp);
2392	vp = NULLVP;
2393
2394	if (unp->un_flag & UNIONFS_OPENEXTU)
2395		vp = unp->un_uppervp;
2396	else if (unp->un_flag & UNIONFS_OPENEXTL)
2397		vp = unp->un_lowervp;
2398
2399	if (vp == NULLVP)
2400		return (EOPNOTSUPP);
2401
2402	return (VOP_LISTEXTATTR(vp, ap->a_attrnamespace, ap->a_uio,
2403	    ap->a_size, ap->a_cred, ap->a_td));
2404}
2405
2406static int
2407unionfs_deleteextattr(struct vop_deleteextattr_args *ap)
2408{
2409	int		error;
2410	struct unionfs_node *unp;
2411	struct vnode   *uvp;
2412	struct vnode   *lvp;
2413	struct vnode   *ovp;
2414	struct ucred   *cred;
2415	struct thread  *td;
2416
2417	KASSERT_UNIONFS_VNODE(ap->a_vp);
2418
2419	error = EROFS;
2420	unp = VTOUNIONFS(ap->a_vp);
2421	uvp = unp->un_uppervp;
2422	lvp = unp->un_lowervp;
2423	ovp = NULLVP;
2424	cred = ap->a_cred;
2425	td = ap->a_td;
2426
2427	UNIONFS_INTERNAL_DEBUG("unionfs_deleteextattr: enter (un_flag=%x)\n", unp->un_flag);
2428
2429	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
2430		return (EROFS);
2431
2432	if (unp->un_flag & UNIONFS_OPENEXTU)
2433		ovp = unp->un_uppervp;
2434	else if (unp->un_flag & UNIONFS_OPENEXTL)
2435		ovp = unp->un_lowervp;
2436
2437	if (ovp == NULLVP)
2438		return (EOPNOTSUPP);
2439
2440	if (ovp == lvp && lvp->v_type == VREG) {
2441		VOP_CLOSEEXTATTR(lvp, 0, cred, td);
2442		if (uvp == NULLVP &&
2443		    (error = unionfs_copyfile(unp, 1, cred, td)) != 0) {
2444unionfs_deleteextattr_reopen:
2445			if ((unp->un_flag & UNIONFS_OPENEXTL) &&
2446			    VOP_OPENEXTATTR(lvp, cred, td)) {
2447#ifdef DIAGNOSTIC
2448				panic("unionfs: VOP_OPENEXTATTR failed");
2449#endif
2450				unp->un_flag &= ~UNIONFS_OPENEXTL;
2451			}
2452			goto unionfs_deleteextattr_abort;
2453		}
2454		uvp = unp->un_uppervp;
2455		if ((error = VOP_OPENEXTATTR(uvp, cred, td)) != 0)
2456			goto unionfs_deleteextattr_reopen;
2457		unp->un_flag &= ~UNIONFS_OPENEXTL;
2458		unp->un_flag |= UNIONFS_OPENEXTU;
2459		ovp = uvp;
2460	}
2461
2462	if (ovp == uvp)
2463		error = VOP_DELETEEXTATTR(ovp, ap->a_attrnamespace, ap->a_name,
2464		    ap->a_cred, ap->a_td);
2465
2466unionfs_deleteextattr_abort:
2467	UNIONFS_INTERNAL_DEBUG("unionfs_deleteextattr: leave (%d)\n", error);
2468
2469	return (error);
2470}
2471
2472static int
2473unionfs_setlabel(struct vop_setlabel_args *ap)
2474{
2475	int		error;
2476	struct unionfs_node *unp;
2477	struct vnode   *uvp;
2478	struct vnode   *lvp;
2479	struct thread  *td;
2480
2481	UNIONFS_INTERNAL_DEBUG("unionfs_setlabel: enter\n");
2482
2483	KASSERT_UNIONFS_VNODE(ap->a_vp);
2484
2485	error = EROFS;
2486	unp = VTOUNIONFS(ap->a_vp);
2487	uvp = unp->un_uppervp;
2488	lvp = unp->un_lowervp;
2489	td = ap->a_td;
2490
2491	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
2492		return (EROFS);
2493
2494	if (uvp == NULLVP && lvp->v_type == VREG) {
2495		if ((error = unionfs_copyfile(unp, 1, ap->a_cred, td)) != 0)
2496			return (error);
2497		uvp = unp->un_uppervp;
2498	}
2499
2500	if (uvp != NULLVP)
2501		error = VOP_SETLABEL(uvp, ap->a_label, ap->a_cred, td);
2502
2503	UNIONFS_INTERNAL_DEBUG("unionfs_setlabel: leave (%d)\n", error);
2504
2505	return (error);
2506}
2507
2508static int
2509unionfs_vptofh(struct vop_vptofh_args *ap)
2510{
2511	return (EOPNOTSUPP);
2512}
2513
2514static int
2515unionfs_add_writecount(struct vop_add_writecount_args *ap)
2516{
2517	struct vnode *tvp, *vp;
2518	struct unionfs_node *unp;
2519	int error;
2520
2521	vp = ap->a_vp;
2522	unp = VTOUNIONFS(vp);
2523	tvp = unp->un_uppervp != NULL ? unp->un_uppervp : unp->un_lowervp;
2524	VI_LOCK(vp);
2525	/* text refs are bypassed to lowervp */
2526	VNASSERT(vp->v_writecount >= 0, vp, ("wrong null writecount"));
2527	VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp,
2528	    ("wrong writecount inc %d", ap->a_inc));
2529	if (tvp != NULL)
2530		error = VOP_ADD_WRITECOUNT(tvp, ap->a_inc);
2531	else if (vp->v_writecount < 0)
2532		error = ETXTBSY;
2533	else
2534		error = 0;
2535	if (error == 0)
2536		vp->v_writecount += ap->a_inc;
2537	VI_UNLOCK(vp);
2538	return (error);
2539}
2540
2541struct vop_vector unionfs_vnodeops = {
2542	.vop_default =		&default_vnodeops,
2543
2544	.vop_access =		unionfs_access,
2545	.vop_aclcheck =		unionfs_aclcheck,
2546	.vop_advlock =		unionfs_advlock,
2547	.vop_bmap =		VOP_EOPNOTSUPP,
2548	.vop_cachedlookup =	unionfs_lookup,
2549	.vop_close =		unionfs_close,
2550	.vop_closeextattr =	unionfs_closeextattr,
2551	.vop_create =		unionfs_create,
2552	.vop_deleteextattr =	unionfs_deleteextattr,
2553	.vop_fsync =		unionfs_fsync,
2554	.vop_getacl =		unionfs_getacl,
2555	.vop_getattr =		unionfs_getattr,
2556	.vop_getextattr =	unionfs_getextattr,
2557	.vop_getwritemount =	unionfs_getwritemount,
2558	.vop_inactive =		unionfs_inactive,
2559	.vop_islocked =		unionfs_islocked,
2560	.vop_ioctl =		unionfs_ioctl,
2561	.vop_link =		unionfs_link,
2562	.vop_listextattr =	unionfs_listextattr,
2563	.vop_lock1 =		unionfs_lock,
2564	.vop_lookup =		vfs_cache_lookup,
2565	.vop_mkdir =		unionfs_mkdir,
2566	.vop_mknod =		unionfs_mknod,
2567	.vop_open =		unionfs_open,
2568	.vop_openextattr =	unionfs_openextattr,
2569	.vop_pathconf =		unionfs_pathconf,
2570	.vop_poll =		unionfs_poll,
2571	.vop_print =		unionfs_print,
2572	.vop_read =		unionfs_read,
2573	.vop_readdir =		unionfs_readdir,
2574	.vop_readlink =		unionfs_readlink,
2575	.vop_reclaim =		unionfs_reclaim,
2576	.vop_remove =		unionfs_remove,
2577	.vop_rename =		unionfs_rename,
2578	.vop_rmdir =		unionfs_rmdir,
2579	.vop_setacl =		unionfs_setacl,
2580	.vop_setattr =		unionfs_setattr,
2581	.vop_setextattr =	unionfs_setextattr,
2582	.vop_setlabel =		unionfs_setlabel,
2583	.vop_strategy =		unionfs_strategy,
2584	.vop_symlink =		unionfs_symlink,
2585	.vop_unlock =		unionfs_unlock,
2586	.vop_whiteout =		unionfs_whiteout,
2587	.vop_write =		unionfs_write,
2588	.vop_vptofh =		unionfs_vptofh,
2589	.vop_add_writecount =	unionfs_add_writecount,
2590};
2591