2 * Copyright (c) 1994 Jan-Simon Pendry
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>
8 * This code is derived from software contributed to Berkeley by
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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
35 * @(#)union_subr.c 8.20 (Berkeley) 5/20/95
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
43 #include <sys/mutex.h>
44 #include <sys/malloc.h>
45 #include <sys/mount.h>
46 #include <sys/namei.h>
48 #include <sys/vnode.h>
49 #include <sys/dirent.h>
50 #include <sys/fcntl.h>
51 #include <sys/filedesc.h>
53 #include <sys/resourcevar.h>
55 #include <security/mac/mac_framework.h>
59 #include <fs/unionfs/union.h>
61 #define NUNIONFSNODECACHE 16
63 static MALLOC_DEFINE(M_UNIONFSHASH, "UNIONFS hash", "UNIONFS hash table");
64 MALLOC_DEFINE(M_UNIONFSNODE, "UNIONFS node", "UNIONFS vnode private part");
65 MALLOC_DEFINE(M_UNIONFSPATH, "UNIONFS path", "UNIONFS path private part");
71 unionfs_init(struct vfsconf *vfsp)
73 UNIONFSDEBUG("unionfs_init\n"); /* printed during system boot */
81 unionfs_uninit(struct vfsconf *vfsp)
86 static struct unionfs_node_hashhead *
87 unionfs_get_hashhead(struct vnode *dvp, char *path)
91 struct unionfs_node *unp;
94 unp = VTOUNIONFS(dvp);
96 for (count = 0; path[count]; count++)
100 return (&(unp->un_hashtbl[hash & (unp->un_hashmask)]));
104 * Get the cached vnode.
106 static struct vnode *
107 unionfs_get_cached_vnode(struct vnode *uvp, struct vnode *lvp,
108 struct vnode *dvp, char *path)
110 struct unionfs_node_hashhead *hd;
111 struct unionfs_node *unp;
114 KASSERT((uvp == NULLVP || uvp->v_type == VDIR),
115 ("unionfs_get_cached_vnode: v_type != VDIR"));
116 KASSERT((lvp == NULLVP || lvp->v_type == VDIR),
117 ("unionfs_get_cached_vnode: v_type != VDIR"));
120 hd = unionfs_get_hashhead(dvp, path);
121 LIST_FOREACH(unp, hd, un_hash) {
122 if (!strcmp(unp->un_path, path)) {
123 vp = UNIONFSTOV(unp);
124 VI_LOCK_FLAGS(vp, MTX_DUPOK);
126 vp->v_iflag &= ~VI_OWEINACT;
127 if ((vp->v_iflag & (VI_DOOMED | VI_DOINGINACT)) != 0) {
141 * Add the new vnode into cache.
143 static struct vnode *
144 unionfs_ins_cached_vnode(struct unionfs_node *uncp,
145 struct vnode *dvp, char *path)
147 struct unionfs_node_hashhead *hd;
148 struct unionfs_node *unp;
151 KASSERT((uncp->un_uppervp==NULLVP || uncp->un_uppervp->v_type==VDIR),
152 ("unionfs_ins_cached_vnode: v_type != VDIR"));
153 KASSERT((uncp->un_lowervp==NULLVP || uncp->un_lowervp->v_type==VDIR),
154 ("unionfs_ins_cached_vnode: v_type != VDIR"));
157 hd = unionfs_get_hashhead(dvp, path);
158 LIST_FOREACH(unp, hd, un_hash) {
159 if (!strcmp(unp->un_path, path)) {
160 vp = UNIONFSTOV(unp);
161 VI_LOCK_FLAGS(vp, MTX_DUPOK);
162 vp->v_iflag &= ~VI_OWEINACT;
163 if ((vp->v_iflag & (VI_DOOMED | VI_DOINGINACT)) != 0) {
164 LIST_INSERT_HEAD(hd, uncp, un_hash);
174 LIST_INSERT_HEAD(hd, uncp, un_hash);
184 unionfs_rem_cached_vnode(struct unionfs_node *unp, struct vnode *dvp)
186 KASSERT((unp != NULL), ("unionfs_rem_cached_vnode: null node"));
187 KASSERT((dvp != NULLVP),
188 ("unionfs_rem_cached_vnode: null parent vnode"));
189 KASSERT((unp->un_hash.le_prev != NULL),
190 ("unionfs_rem_cached_vnode: null hash"));
193 LIST_REMOVE(unp, un_hash);
194 unp->un_hash.le_next = NULL;
195 unp->un_hash.le_prev = NULL;
200 * Make a new or get existing unionfs node.
202 * uppervp and lowervp should be unlocked. Because if new unionfs vnode is
203 * locked, uppervp or lowervp is locked too. In order to prevent dead lock,
204 * you should not lock plurality simultaneously.
207 unionfs_nodeget(struct mount *mp, struct vnode *uppervp,
208 struct vnode *lowervp, struct vnode *dvp,
209 struct vnode **vpp, struct componentname *cnp,
212 struct unionfs_mount *ump;
213 struct unionfs_node *unp;
220 ump = MOUNTTOUNIONFSMOUNT(mp);
221 lkflags = (cnp ? cnp->cn_lkflags : 0);
222 path = (cnp ? cnp->cn_nameptr : NULL);
225 if (uppervp == NULLVP && lowervp == NULLVP)
226 panic("unionfs_nodeget: upper and lower is null");
228 vt = (uppervp != NULLVP ? uppervp->v_type : lowervp->v_type);
230 /* If it has no ISLASTCN flag, path check is skipped. */
231 if (cnp && !(cnp->cn_flags & ISLASTCN))
234 /* check the cache */
235 if (path != NULL && dvp != NULLVP && vt == VDIR) {
236 vp = unionfs_get_cached_vnode(uppervp, lowervp, dvp, path);
240 goto unionfs_nodeget_out;
244 if ((uppervp == NULLVP || ump->um_uppervp != uppervp) ||
245 (lowervp == NULLVP || ump->um_lowervp != lowervp)) {
246 /* dvp will be NULLVP only in case of root vnode. */
252 * Do the MALLOC before the getnewvnode since doing so afterward
253 * might cause a bogus v_data pointer to get dereferenced elsewhere
254 * if MALLOC should block.
256 unp = malloc(sizeof(struct unionfs_node),
257 M_UNIONFSNODE, M_WAITOK | M_ZERO);
259 error = getnewvnode("unionfs", mp, &unionfs_vnodeops, &vp);
261 free(unp, M_UNIONFSNODE);
264 error = insmntque(vp, mp); /* XXX: Too early for mpsafe fs */
266 free(unp, M_UNIONFSNODE);
271 if (uppervp != NULLVP)
273 if (lowervp != NULLVP)
277 unp->un_hashtbl = hashinit(NUNIONFSNODECACHE, M_UNIONFSHASH,
278 &(unp->un_hashmask));
281 unp->un_uppervp = uppervp;
282 unp->un_lowervp = lowervp;
284 if (uppervp != NULLVP)
285 vp->v_vnlock = uppervp->v_vnlock;
287 vp->v_vnlock = lowervp->v_vnlock;
290 unp->un_path = (char *)
291 malloc(cnp->cn_namelen +1, M_UNIONFSPATH, M_WAITOK|M_ZERO);
292 bcopy(cnp->cn_nameptr, unp->un_path, cnp->cn_namelen);
293 unp->un_path[cnp->cn_namelen] = '\0';
298 if ((uppervp != NULLVP && ump->um_uppervp == uppervp) &&
299 (lowervp != NULLVP && ump->um_lowervp == lowervp))
300 vp->v_vflag |= VV_ROOT;
302 if (path != NULL && dvp != NULLVP && vt == VDIR)
303 *vpp = unionfs_ins_cached_vnode(unp, dvp, path);
304 if ((*vpp) != NULLVP) {
307 if (uppervp != NULLVP)
309 if (lowervp != NULLVP)
312 unp->un_uppervp = NULLVP;
313 unp->un_lowervp = NULLVP;
314 unp->un_dvp = NULLVP;
322 if (lkflags & LK_TYPE_MASK)
323 vn_lock(vp, lkflags | LK_RETRY);
329 * Clean up the unionfs node.
332 unionfs_noderem(struct vnode *vp, struct thread *td)
336 struct unionfs_node *unp, *unp_t1, *unp_t2;
337 struct unionfs_node_hashhead *hd;
338 struct unionfs_node_status *unsp, *unsp_tmp;
344 * Use the interlock to protect the clearing of v_data to
345 * prevent faults in unionfs_lock().
348 unp = VTOUNIONFS(vp);
349 lvp = unp->un_lowervp;
350 uvp = unp->un_uppervp;
352 unp->un_lowervp = unp->un_uppervp = NULLVP;
354 vp->v_vnlock = &(vp->v_lock);
356 lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_INTERLOCK, VI_MTX(vp));
363 if (dvp != NULLVP && unp->un_hash.le_prev != NULL)
364 unionfs_rem_cached_vnode(unp, dvp);
367 vfslocked = VFS_LOCK_GIANT(lvp->v_mount);
369 VFS_UNLOCK_GIANT(vfslocked);
372 vfslocked = VFS_LOCK_GIANT(uvp->v_mount);
374 VFS_UNLOCK_GIANT(vfslocked);
377 vfslocked = VFS_LOCK_GIANT(dvp->v_mount);
379 VFS_UNLOCK_GIANT(vfslocked);
380 unp->un_dvp = NULLVP;
382 if (unp->un_path != NULL) {
383 free(unp->un_path, M_UNIONFSPATH);
387 if (unp->un_hashtbl != NULL) {
388 for (count = 0; count <= unp->un_hashmask; count++) {
389 hd = unp->un_hashtbl + count;
390 LIST_FOREACH_SAFE(unp_t1, hd, un_hash, unp_t2) {
391 LIST_REMOVE(unp_t1, un_hash);
392 unp_t1->un_hash.le_next = NULL;
393 unp_t1->un_hash.le_prev = NULL;
396 hashdestroy(unp->un_hashtbl, M_UNIONFSHASH, unp->un_hashmask);
399 LIST_FOREACH_SAFE(unsp, &(unp->un_unshead), uns_list, unsp_tmp) {
400 LIST_REMOVE(unsp, uns_list);
403 free(unp, M_UNIONFSNODE);
407 * Get the unionfs node status.
408 * You need exclusive lock this vnode.
411 unionfs_get_node_status(struct unionfs_node *unp, struct thread *td,
412 struct unionfs_node_status **unspp)
414 struct unionfs_node_status *unsp;
415 pid_t pid = td->td_proc->p_pid;
417 KASSERT(NULL != unspp, ("null pointer"));
418 ASSERT_VOP_ELOCKED(UNIONFSTOV(unp), "unionfs_get_node_status");
420 LIST_FOREACH(unsp, &(unp->un_unshead), uns_list) {
421 if (unsp->uns_pid == pid) {
427 /* create a new unionfs node status */
428 unsp = malloc(sizeof(struct unionfs_node_status),
429 M_TEMP, M_WAITOK | M_ZERO);
432 LIST_INSERT_HEAD(&(unp->un_unshead), unsp, uns_list);
438 * Remove the unionfs node status, if you can.
439 * You need exclusive lock this vnode.
442 unionfs_tryrem_node_status(struct unionfs_node *unp,
443 struct unionfs_node_status *unsp)
445 KASSERT(NULL != unsp, ("null pointer"));
446 ASSERT_VOP_ELOCKED(UNIONFSTOV(unp), "unionfs_get_node_status");
448 if (0 < unsp->uns_lower_opencnt || 0 < unsp->uns_upper_opencnt)
451 LIST_REMOVE(unsp, uns_list);
456 * Create upper node attr.
459 unionfs_create_uppervattr_core(struct unionfs_mount *ump,
465 uva->va_type = lva->va_type;
466 uva->va_atime = lva->va_atime;
467 uva->va_mtime = lva->va_mtime;
468 uva->va_ctime = lva->va_ctime;
470 switch (ump->um_copymode) {
471 case UNIONFS_TRANSPARENT:
472 uva->va_mode = lva->va_mode;
473 uva->va_uid = lva->va_uid;
474 uva->va_gid = lva->va_gid;
476 case UNIONFS_MASQUERADE:
477 if (ump->um_uid == lva->va_uid) {
478 uva->va_mode = lva->va_mode & 077077;
479 uva->va_mode |= (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile) & 0700;
480 uva->va_uid = lva->va_uid;
481 uva->va_gid = lva->va_gid;
483 uva->va_mode = (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile);
484 uva->va_uid = ump->um_uid;
485 uva->va_gid = ump->um_gid;
488 default: /* UNIONFS_TRADITIONAL */
489 uva->va_mode = 0777 & ~td->td_proc->p_fd->fd_cmask;
490 uva->va_uid = ump->um_uid;
491 uva->va_gid = ump->um_gid;
497 * Create upper node attr.
500 unionfs_create_uppervattr(struct unionfs_mount *ump,
509 if ((error = VOP_GETATTR(lvp, &lva, cred)))
512 unionfs_create_uppervattr_core(ump, &lva, uva, td);
520 * dvp should be locked on entry and will be locked on return.
522 * If an error is returned, *vpp will be invalid, otherwise it will hold a
523 * locked, referenced vnode. If *vpp == dvp then remember that only one
524 * LK_EXCLUSIVE lock is held.
527 unionfs_relookup(struct vnode *dvp, struct vnode **vpp,
528 struct componentname *cnp, struct componentname *cn,
529 struct thread *td, char *path, int pathlen, u_long nameiop)
533 cn->cn_namelen = pathlen;
534 cn->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
535 bcopy(path, cn->cn_pnbuf, pathlen);
536 cn->cn_pnbuf[pathlen] = '\0';
538 cn->cn_nameiop = nameiop;
539 cn->cn_flags = (LOCKPARENT | LOCKLEAF | HASBUF | SAVENAME | ISLASTCN);
540 cn->cn_lkflags = LK_EXCLUSIVE;
542 cn->cn_cred = cnp->cn_cred;
544 cn->cn_nameptr = cn->cn_pnbuf;
545 cn->cn_consume = cnp->cn_consume;
547 if (nameiop == DELETE)
548 cn->cn_flags |= (cnp->cn_flags & (DOWHITEOUT | SAVESTART));
549 else if (RENAME == nameiop)
550 cn->cn_flags |= (cnp->cn_flags & SAVESTART);
555 if ((error = relookup(dvp, vpp, cn))) {
556 uma_zfree(namei_zone, cn->cn_pnbuf);
557 cn->cn_flags &= ~HASBUF;
558 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
566 * relookup for CREATE namei operation.
568 * dvp is unionfs vnode. dvp should be locked.
570 * If it called 'unionfs_copyfile' function by unionfs_link etc,
571 * VOP_LOOKUP information is broken.
572 * So it need relookup in order to create link etc.
575 unionfs_relookup_for_create(struct vnode *dvp, struct componentname *cnp,
581 struct componentname cn;
583 udvp = UNIONFSVPTOUPPERVP(dvp);
586 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
587 strlen(cnp->cn_nameptr), CREATE);
600 if (cn.cn_flags & HASBUF) {
601 uma_zfree(namei_zone, cn.cn_pnbuf);
602 cn.cn_flags &= ~HASBUF;
606 cn.cn_flags |= (cnp->cn_flags & HASBUF);
607 cnp->cn_flags = cn.cn_flags;
614 * relookup for DELETE namei operation.
616 * dvp is unionfs vnode. dvp should be locked.
619 unionfs_relookup_for_delete(struct vnode *dvp, struct componentname *cnp,
625 struct componentname cn;
627 udvp = UNIONFSVPTOUPPERVP(dvp);
630 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
631 strlen(cnp->cn_nameptr), DELETE);
644 if (cn.cn_flags & HASBUF) {
645 uma_zfree(namei_zone, cn.cn_pnbuf);
646 cn.cn_flags &= ~HASBUF;
650 cn.cn_flags |= (cnp->cn_flags & HASBUF);
651 cnp->cn_flags = cn.cn_flags;
658 * relookup for RENAME namei operation.
660 * dvp is unionfs vnode. dvp should be locked.
663 unionfs_relookup_for_rename(struct vnode *dvp, struct componentname *cnp,
669 struct componentname cn;
671 udvp = UNIONFSVPTOUPPERVP(dvp);
674 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
675 strlen(cnp->cn_nameptr), RENAME);
686 if (cn.cn_flags & HASBUF) {
687 uma_zfree(namei_zone, cn.cn_pnbuf);
688 cn.cn_flags &= ~HASBUF;
692 cn.cn_flags |= (cnp->cn_flags & HASBUF);
693 cnp->cn_flags = cn.cn_flags;
701 * Update the unionfs_node.
703 * uvp is new locked upper vnode. unionfs vnode's lock will be exchanged to the
704 * uvp's lock and lower's lock will be unlocked.
707 unionfs_node_update(struct unionfs_node *unp, struct vnode *uvp,
710 unsigned count, lockrec;
715 vp = UNIONFSTOV(unp);
716 lvp = unp->un_lowervp;
717 ASSERT_VOP_ELOCKED(lvp, "unionfs_node_update");
724 unp->un_uppervp = uvp;
725 vp->v_vnlock = uvp->v_vnlock;
727 lockrec = lvp->v_vnlock->lk_recurse;
728 for (count = 0; count < lockrec; count++)
729 vn_lock(uvp, LK_EXCLUSIVE | LK_CANRECURSE | LK_RETRY);
734 if (unp->un_path != NULL && dvp != NULLVP && vp->v_type == VDIR) {
735 static struct unionfs_node_hashhead *hd;
738 hd = unionfs_get_hashhead(dvp, unp->un_path);
739 LIST_REMOVE(unp, un_hash);
740 LIST_INSERT_HEAD(hd, unp, un_hash);
746 * Create a new shadow dir.
748 * udvp should be locked on entry and will be locked on return.
750 * If no error returned, unp will be updated.
753 unionfs_mkshadowdir(struct unionfs_mount *ump, struct vnode *udvp,
754 struct unionfs_node *unp, struct componentname *cnp,
762 struct componentname cn;
765 struct ucred *credbk;
766 struct uidinfo *rootinfo;
768 if (unp->un_uppervp != NULLVP)
771 lvp = unp->un_lowervp;
773 credbk = cnp->cn_cred;
775 /* Authority change to root */
776 rootinfo = uifind((uid_t)0);
777 cred = crdup(cnp->cn_cred);
778 chgproccnt(cred->cr_ruidinfo, 1, 0);
779 change_euid(cred, rootinfo);
780 change_ruid(cred, rootinfo);
781 change_svuid(cred, (uid_t)0);
785 memset(&cn, 0, sizeof(cn));
787 if ((error = VOP_GETATTR(lvp, &lva, cnp->cn_cred)))
788 goto unionfs_mkshadowdir_abort;
790 if ((error = unionfs_relookup(udvp, &uvp, cnp, &cn, td, cnp->cn_nameptr, cnp->cn_namelen, CREATE)))
791 goto unionfs_mkshadowdir_abort;
799 goto unionfs_mkshadowdir_free_out;
802 if ((error = vn_start_write(udvp, &mp, V_WAIT | PCATCH)))
803 goto unionfs_mkshadowdir_free_out;
804 unionfs_create_uppervattr_core(ump, &lva, &va, td);
806 error = VOP_MKDIR(udvp, &uvp, &cn, &va);
809 unionfs_node_update(unp, uvp, td);
812 * XXX The bug which cannot set uid/gid was corrected.
816 VOP_SETATTR(uvp, &va, cn.cn_cred);
818 vn_finished_write(mp);
820 unionfs_mkshadowdir_free_out:
821 if (cn.cn_flags & HASBUF) {
822 uma_zfree(namei_zone, cn.cn_pnbuf);
823 cn.cn_flags &= ~HASBUF;
826 unionfs_mkshadowdir_abort:
827 cnp->cn_cred = credbk;
828 chgproccnt(cred->cr_ruidinfo, -1, 0);
835 * Create a new whiteout.
837 * dvp should be locked on entry and will be locked on return.
840 unionfs_mkwhiteout(struct vnode *dvp, struct componentname *cnp,
841 struct thread *td, char *path)
845 struct componentname cn;
849 path = cnp->cn_nameptr;
852 if ((error = unionfs_relookup(dvp, &wvp, cnp, &cn, td, path, strlen(path), CREATE)))
855 if (cn.cn_flags & HASBUF) {
856 uma_zfree(namei_zone, cn.cn_pnbuf);
857 cn.cn_flags &= ~HASBUF;
867 if ((error = vn_start_write(dvp, &mp, V_WAIT | PCATCH)))
868 goto unionfs_mkwhiteout_free_out;
869 error = VOP_WHITEOUT(dvp, &cn, CREATE);
871 vn_finished_write(mp);
873 unionfs_mkwhiteout_free_out:
874 if (cn.cn_flags & HASBUF) {
875 uma_zfree(namei_zone, cn.cn_pnbuf);
876 cn.cn_flags &= ~HASBUF;
883 * Create a new vnode for create a new shadow file.
885 * If an error is returned, *vpp will be invalid, otherwise it will hold a
886 * locked, referenced and opened vnode.
888 * unp is never updated.
891 unionfs_vn_create_on_upper(struct vnode **vpp, struct vnode *udvp,
892 struct unionfs_node *unp, struct vattr *uvap,
895 struct unionfs_mount *ump;
902 struct componentname cn;
904 ump = MOUNTTOUNIONFSMOUNT(UNIONFSTOV(unp)->v_mount);
906 lvp = unp->un_lowervp;
908 fmode = FFLAGS(O_WRONLY | O_CREAT | O_TRUNC | O_EXCL);
911 if ((error = VOP_GETATTR(lvp, &lva, cred)) != 0)
913 unionfs_create_uppervattr_core(ump, &lva, uvap, td);
915 if (unp->un_path == NULL)
916 panic("unionfs: un_path is null");
918 cn.cn_namelen = strlen(unp->un_path);
919 cn.cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
920 bcopy(unp->un_path, cn.cn_pnbuf, cn.cn_namelen + 1);
921 cn.cn_nameiop = CREATE;
922 cn.cn_flags = (LOCKPARENT | LOCKLEAF | HASBUF | SAVENAME | ISLASTCN);
923 cn.cn_lkflags = LK_EXCLUSIVE;
926 cn.cn_nameptr = cn.cn_pnbuf;
930 if ((error = relookup(udvp, &vp, &cn)) != 0)
931 goto unionfs_vn_create_on_upper_free_out2;
940 goto unionfs_vn_create_on_upper_free_out1;
943 if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
944 goto unionfs_vn_create_on_upper_free_out1;
946 if ((error = VOP_OPEN(vp, fmode, cred, td, NULL)) != 0) {
948 goto unionfs_vn_create_on_upper_free_out1;
953 unionfs_vn_create_on_upper_free_out1:
956 unionfs_vn_create_on_upper_free_out2:
957 if (cn.cn_flags & HASBUF) {
958 uma_zfree(namei_zone, cn.cn_pnbuf);
959 cn.cn_flags &= ~HASBUF;
966 * Copy from lvp to uvp.
968 * lvp and uvp should be locked and opened on entry and will be locked and
972 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
973 struct ucred *cred, struct thread *td)
984 memset(&uio, 0, sizeof(uio));
987 uio.uio_segflg = UIO_SYSSPACE;
990 buf = malloc(MAXBSIZE, M_TEMP, M_WAITOK);
993 offset = uio.uio_offset;
998 iov.iov_len = MAXBSIZE;
999 uio.uio_resid = iov.iov_len;
1000 uio.uio_rw = UIO_READ;
1002 if ((error = VOP_READ(lvp, &uio, 0, cred)) != 0)
1004 if ((count = MAXBSIZE - uio.uio_resid) == 0)
1008 while (bufoffset < count) {
1011 iov.iov_base = buf + bufoffset;
1012 iov.iov_len = count - bufoffset;
1013 uio.uio_offset = offset + bufoffset;
1014 uio.uio_resid = iov.iov_len;
1015 uio.uio_rw = UIO_WRITE;
1017 if ((error = VOP_WRITE(uvp, &uio, 0, cred)) != 0)
1020 bufoffset += (count - bufoffset) - uio.uio_resid;
1023 uio.uio_offset = offset + bufoffset;
1032 * Copy file from lower to upper.
1034 * If you need copy of the contents, set 1 to docopy. Otherwise, set 0 to
1037 * If no error returned, unp will be updated.
1040 unionfs_copyfile(struct unionfs_node *unp, int docopy, struct ucred *cred,
1050 lvp = unp->un_lowervp;
1053 if ((UNIONFSTOV(unp)->v_mount->mnt_flag & MNT_RDONLY))
1055 if (unp->un_dvp == NULLVP)
1057 if (unp->un_uppervp != NULLVP)
1059 udvp = VTOUNIONFS(unp->un_dvp)->un_uppervp;
1062 if ((udvp->v_mount->mnt_flag & MNT_RDONLY))
1065 error = VOP_ACCESS(lvp, VREAD, cred, td);
1069 if ((error = vn_start_write(udvp, &mp, V_WAIT | PCATCH)) != 0)
1071 error = unionfs_vn_create_on_upper(&uvp, udvp, unp, &uva, td);
1073 vn_finished_write(mp);
1078 error = VOP_OPEN(lvp, FREAD, cred, td, NULL);
1080 error = unionfs_copyfile_core(lvp, uvp, cred, td);
1081 VOP_CLOSE(lvp, FREAD, cred, td);
1084 VOP_CLOSE(uvp, FWRITE, cred, td);
1085 uvp->v_writecount--;
1087 vn_finished_write(mp);
1090 /* Reset the attributes. Ignore errors. */
1092 VOP_SETATTR(uvp, &uva, cred);
1095 unionfs_node_update(unp, uvp, td);
1101 * It checks whether vp can rmdir. (check empty)
1103 * vp is unionfs vnode.
1104 * vp should be locked.
1107 unionfs_check_rmdir(struct vnode *vp, struct ucred *cred, struct thread *td)
1116 struct componentname cn;
1118 * The size of buf needs to be larger than DIRBLKSIZ.
1126 ASSERT_VOP_ELOCKED(vp, "unionfs_check_rmdir");
1129 uvp = UNIONFSVPTOUPPERVP(vp);
1130 lvp = UNIONFSVPTOLOWERVP(vp);
1133 if ((error = VOP_GETATTR(uvp, &va, cred)) != 0)
1135 if (va.va_flags & OPAQUE)
1140 if ((error = mac_vnode_check_open(cred, vp, VEXEC|VREAD)) != 0)
1143 if ((error = VOP_ACCESS(vp, VEXEC|VREAD, cred, td)) != 0)
1145 if ((error = VOP_OPEN(vp, FREAD, cred, td, NULL)) != 0)
1148 uio.uio_rw = UIO_READ;
1149 uio.uio_segflg = UIO_SYSSPACE;
1154 error = mac_vnode_check_readdir(td->td_ucred, lvp);
1156 while (!error && !eofflag) {
1158 iov.iov_len = sizeof(buf);
1161 uio.uio_resid = iov.iov_len;
1163 error = VOP_READDIR(lvp, &uio, cred, &eofflag, NULL, NULL);
1166 if (eofflag == 0 && uio.uio_resid == sizeof(buf)) {
1168 panic("bad readdir response from lower FS.");
1173 edp = (struct dirent*)&buf[sizeof(buf) - uio.uio_resid];
1174 for (dp = (struct dirent*)buf; !error && dp < edp;
1175 dp = (struct dirent*)((caddr_t)dp + dp->d_reclen)) {
1176 if (dp->d_type == DT_WHT ||
1177 (dp->d_namlen == 1 && dp->d_name[0] == '.') ||
1178 (dp->d_namlen == 2 && !bcmp(dp->d_name, "..", 2)))
1181 cn.cn_namelen = dp->d_namlen;
1183 cn.cn_nameptr = dp->d_name;
1184 cn.cn_nameiop = LOOKUP;
1185 cn.cn_flags = (LOCKPARENT | LOCKLEAF | SAVENAME | RDONLY | ISLASTCN);
1186 cn.cn_lkflags = LK_EXCLUSIVE;
1192 * check entry in lower.
1193 * Sometimes, readdir function returns
1196 lookuperr = VOP_LOOKUP(lvp, &tvp, &cn);
1201 continue; /* skip entry */
1205 * If it has no exist/whiteout entry in upper,
1206 * directory is not empty.
1208 cn.cn_flags = (LOCKPARENT | LOCKLEAF | SAVENAME | RDONLY | ISLASTCN);
1209 lookuperr = VOP_LOOKUP(uvp, &tvp, &cn);
1214 /* ignore exist or whiteout entry */
1216 (lookuperr == ENOENT && (cn.cn_flags & ISWHITEOUT)))
1224 VOP_CLOSE(vp, FREAD, cred, td);
1232 unionfs_checkuppervp(struct vnode *vp, char *fil, int lno)
1234 struct unionfs_node *unp;
1236 unp = VTOUNIONFS(vp);
1239 if (vp->v_op != unionfs_vnodeop_p) {
1240 printf("unionfs_checkuppervp: on non-unionfs-node.\n");
1242 kdb_enter(KDB_WHY_UNIONFS,
1243 "unionfs_checkuppervp: on non-unionfs-node.\n");
1245 panic("unionfs_checkuppervp");
1248 return (unp->un_uppervp);
1252 unionfs_checklowervp(struct vnode *vp, char *fil, int lno)
1254 struct unionfs_node *unp;
1256 unp = VTOUNIONFS(vp);
1259 if (vp->v_op != unionfs_vnodeop_p) {
1260 printf("unionfs_checklowervp: on non-unionfs-node.\n");
1262 kdb_enter(KDB_WHY_UNIONFS,
1263 "unionfs_checklowervp: on non-unionfs-node.\n");
1265 panic("unionfs_checklowervp");
1268 return (unp->un_lowervp);