2 * Copyright (c) 1994 Jan-Simon Pendry
4 * The Regents of the University of California. All rights reserved.
5 * Copyright (c) 2005, 2006, 2012 Masanori Ozawa <ozawa@ongs.co.jp>, ONGS Inc.
6 * Copyright (c) 2006, 2012 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 * 3. 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. */
250 unp = malloc(sizeof(struct unionfs_node),
251 M_UNIONFSNODE, M_WAITOK | M_ZERO);
253 error = getnewvnode("unionfs", mp, &unionfs_vnodeops, &vp);
255 free(unp, M_UNIONFSNODE);
258 error = insmntque(vp, mp); /* XXX: Too early for mpsafe fs */
260 free(unp, M_UNIONFSNODE);
265 if (uppervp != NULLVP)
267 if (lowervp != NULLVP)
271 unp->un_hashtbl = hashinit(NUNIONFSNODECACHE, M_UNIONFSHASH,
272 &(unp->un_hashmask));
275 unp->un_uppervp = uppervp;
276 unp->un_lowervp = lowervp;
278 if (uppervp != NULLVP)
279 vp->v_vnlock = uppervp->v_vnlock;
281 vp->v_vnlock = lowervp->v_vnlock;
284 unp->un_path = (char *)
285 malloc(cnp->cn_namelen +1, M_UNIONFSPATH, M_WAITOK|M_ZERO);
286 bcopy(cnp->cn_nameptr, unp->un_path, cnp->cn_namelen);
287 unp->un_path[cnp->cn_namelen] = '\0';
292 if ((uppervp != NULLVP && ump->um_uppervp == uppervp) &&
293 (lowervp != NULLVP && ump->um_lowervp == lowervp))
294 vp->v_vflag |= VV_ROOT;
296 if (path != NULL && dvp != NULLVP && vt == VDIR)
297 *vpp = unionfs_ins_cached_vnode(unp, dvp, path);
298 if ((*vpp) != NULLVP) {
301 if (uppervp != NULLVP)
303 if (lowervp != NULLVP)
306 unp->un_uppervp = NULLVP;
307 unp->un_lowervp = NULLVP;
308 unp->un_dvp = NULLVP;
316 if (lkflags & LK_TYPE_MASK)
317 vn_lock(vp, lkflags | LK_RETRY);
323 * Clean up the unionfs node.
326 unionfs_noderem(struct vnode *vp, struct thread *td)
329 struct unionfs_node *unp, *unp_t1, *unp_t2;
330 struct unionfs_node_hashhead *hd;
331 struct unionfs_node_status *unsp, *unsp_tmp;
337 * Use the interlock to protect the clearing of v_data to
338 * prevent faults in unionfs_lock().
341 unp = VTOUNIONFS(vp);
342 lvp = unp->un_lowervp;
343 uvp = unp->un_uppervp;
345 unp->un_lowervp = unp->un_uppervp = NULLVP;
346 vp->v_vnlock = &(vp->v_lock);
352 VOP_UNLOCK(lvp, LK_RELEASE);
354 VOP_UNLOCK(uvp, LK_RELEASE);
356 if (dvp != NULLVP && unp->un_hash.le_prev != NULL)
357 unionfs_rem_cached_vnode(unp, dvp);
359 if (lockmgr(vp->v_vnlock, LK_EXCLUSIVE, VI_MTX(vp)) != 0)
360 panic("the lock for deletion is unacquirable.");
368 unp->un_dvp = NULLVP;
370 if (unp->un_path != NULL) {
371 free(unp->un_path, M_UNIONFSPATH);
375 if (unp->un_hashtbl != NULL) {
376 for (count = 0; count <= unp->un_hashmask; count++) {
377 hd = unp->un_hashtbl + count;
378 LIST_FOREACH_SAFE(unp_t1, hd, un_hash, unp_t2) {
379 LIST_REMOVE(unp_t1, un_hash);
380 unp_t1->un_hash.le_next = NULL;
381 unp_t1->un_hash.le_prev = NULL;
384 hashdestroy(unp->un_hashtbl, M_UNIONFSHASH, unp->un_hashmask);
387 LIST_FOREACH_SAFE(unsp, &(unp->un_unshead), uns_list, unsp_tmp) {
388 LIST_REMOVE(unsp, uns_list);
391 free(unp, M_UNIONFSNODE);
395 * Get the unionfs node status.
396 * You need exclusive lock this vnode.
399 unionfs_get_node_status(struct unionfs_node *unp, struct thread *td,
400 struct unionfs_node_status **unspp)
402 struct unionfs_node_status *unsp;
403 pid_t pid = td->td_proc->p_pid;
405 KASSERT(NULL != unspp, ("null pointer"));
406 ASSERT_VOP_ELOCKED(UNIONFSTOV(unp), "unionfs_get_node_status");
408 LIST_FOREACH(unsp, &(unp->un_unshead), uns_list) {
409 if (unsp->uns_pid == pid) {
415 /* create a new unionfs node status */
416 unsp = malloc(sizeof(struct unionfs_node_status),
417 M_TEMP, M_WAITOK | M_ZERO);
420 LIST_INSERT_HEAD(&(unp->un_unshead), unsp, uns_list);
426 * Remove the unionfs node status, if you can.
427 * You need exclusive lock this vnode.
430 unionfs_tryrem_node_status(struct unionfs_node *unp,
431 struct unionfs_node_status *unsp)
433 KASSERT(NULL != unsp, ("null pointer"));
434 ASSERT_VOP_ELOCKED(UNIONFSTOV(unp), "unionfs_get_node_status");
436 if (0 < unsp->uns_lower_opencnt || 0 < unsp->uns_upper_opencnt)
439 LIST_REMOVE(unsp, uns_list);
444 * Create upper node attr.
447 unionfs_create_uppervattr_core(struct unionfs_mount *ump,
453 uva->va_type = lva->va_type;
454 uva->va_atime = lva->va_atime;
455 uva->va_mtime = lva->va_mtime;
456 uva->va_ctime = lva->va_ctime;
458 switch (ump->um_copymode) {
459 case UNIONFS_TRANSPARENT:
460 uva->va_mode = lva->va_mode;
461 uva->va_uid = lva->va_uid;
462 uva->va_gid = lva->va_gid;
464 case UNIONFS_MASQUERADE:
465 if (ump->um_uid == lva->va_uid) {
466 uva->va_mode = lva->va_mode & 077077;
467 uva->va_mode |= (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile) & 0700;
468 uva->va_uid = lva->va_uid;
469 uva->va_gid = lva->va_gid;
471 uva->va_mode = (lva->va_type == VDIR ? ump->um_udir : ump->um_ufile);
472 uva->va_uid = ump->um_uid;
473 uva->va_gid = ump->um_gid;
476 default: /* UNIONFS_TRADITIONAL */
477 uva->va_mode = 0777 & ~td->td_proc->p_fd->fd_cmask;
478 uva->va_uid = ump->um_uid;
479 uva->va_gid = ump->um_gid;
485 * Create upper node attr.
488 unionfs_create_uppervattr(struct unionfs_mount *ump,
497 if ((error = VOP_GETATTR(lvp, &lva, cred)))
500 unionfs_create_uppervattr_core(ump, &lva, uva, td);
508 * dvp should be locked on entry and will be locked on return.
510 * If an error is returned, *vpp will be invalid, otherwise it will hold a
511 * locked, referenced vnode. If *vpp == dvp then remember that only one
512 * LK_EXCLUSIVE lock is held.
515 unionfs_relookup(struct vnode *dvp, struct vnode **vpp,
516 struct componentname *cnp, struct componentname *cn,
517 struct thread *td, char *path, int pathlen, u_long nameiop)
521 cn->cn_namelen = pathlen;
522 cn->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
523 bcopy(path, cn->cn_pnbuf, pathlen);
524 cn->cn_pnbuf[pathlen] = '\0';
526 cn->cn_nameiop = nameiop;
527 cn->cn_flags = (LOCKPARENT | LOCKLEAF | HASBUF | SAVENAME | ISLASTCN);
528 cn->cn_lkflags = LK_EXCLUSIVE;
530 cn->cn_cred = cnp->cn_cred;
532 cn->cn_nameptr = cn->cn_pnbuf;
534 if (nameiop == DELETE)
535 cn->cn_flags |= (cnp->cn_flags & (DOWHITEOUT | SAVESTART));
536 else if (RENAME == nameiop)
537 cn->cn_flags |= (cnp->cn_flags & SAVESTART);
538 else if (nameiop == CREATE)
539 cn->cn_flags |= NOCACHE;
542 VOP_UNLOCK(dvp, LK_RELEASE);
544 if ((error = relookup(dvp, vpp, cn))) {
545 uma_zfree(namei_zone, cn->cn_pnbuf);
546 cn->cn_flags &= ~HASBUF;
547 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
555 * relookup for CREATE namei operation.
557 * dvp is unionfs vnode. dvp should be locked.
559 * If it called 'unionfs_copyfile' function by unionfs_link etc,
560 * VOP_LOOKUP information is broken.
561 * So it need relookup in order to create link etc.
564 unionfs_relookup_for_create(struct vnode *dvp, struct componentname *cnp,
570 struct componentname cn;
572 udvp = UNIONFSVPTOUPPERVP(dvp);
575 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
576 strlen(cnp->cn_nameptr), CREATE);
589 if (cn.cn_flags & HASBUF) {
590 uma_zfree(namei_zone, cn.cn_pnbuf);
591 cn.cn_flags &= ~HASBUF;
595 cn.cn_flags |= (cnp->cn_flags & HASBUF);
596 cnp->cn_flags = cn.cn_flags;
603 * relookup for DELETE namei operation.
605 * dvp is unionfs vnode. dvp should be locked.
608 unionfs_relookup_for_delete(struct vnode *dvp, struct componentname *cnp,
614 struct componentname cn;
616 udvp = UNIONFSVPTOUPPERVP(dvp);
619 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
620 strlen(cnp->cn_nameptr), DELETE);
633 if (cn.cn_flags & HASBUF) {
634 uma_zfree(namei_zone, cn.cn_pnbuf);
635 cn.cn_flags &= ~HASBUF;
639 cn.cn_flags |= (cnp->cn_flags & HASBUF);
640 cnp->cn_flags = cn.cn_flags;
647 * relookup for RENAME namei operation.
649 * dvp is unionfs vnode. dvp should be locked.
652 unionfs_relookup_for_rename(struct vnode *dvp, struct componentname *cnp,
658 struct componentname cn;
660 udvp = UNIONFSVPTOUPPERVP(dvp);
663 error = unionfs_relookup(udvp, &vp, cnp, &cn, td, cnp->cn_nameptr,
664 strlen(cnp->cn_nameptr), RENAME);
675 if (cn.cn_flags & HASBUF) {
676 uma_zfree(namei_zone, cn.cn_pnbuf);
677 cn.cn_flags &= ~HASBUF;
681 cn.cn_flags |= (cnp->cn_flags & HASBUF);
682 cnp->cn_flags = cn.cn_flags;
690 * Update the unionfs_node.
692 * uvp is new locked upper vnode. unionfs vnode's lock will be exchanged to the
693 * uvp's lock and lower's lock will be unlocked.
696 unionfs_node_update(struct unionfs_node *unp, struct vnode *uvp,
699 unsigned count, lockrec;
704 vp = UNIONFSTOV(unp);
705 lvp = unp->un_lowervp;
706 ASSERT_VOP_ELOCKED(lvp, "unionfs_node_update");
713 unp->un_uppervp = uvp;
714 vp->v_vnlock = uvp->v_vnlock;
716 lockrec = lvp->v_vnlock->lk_recurse;
717 for (count = 0; count < lockrec; count++)
718 vn_lock(uvp, LK_EXCLUSIVE | LK_CANRECURSE | LK_RETRY);
723 if (unp->un_path != NULL && dvp != NULLVP && vp->v_type == VDIR) {
724 static struct unionfs_node_hashhead *hd;
727 hd = unionfs_get_hashhead(dvp, unp->un_path);
728 LIST_REMOVE(unp, un_hash);
729 LIST_INSERT_HEAD(hd, unp, un_hash);
735 * Create a new shadow dir.
737 * udvp should be locked on entry and will be locked on return.
739 * If no error returned, unp will be updated.
742 unionfs_mkshadowdir(struct unionfs_mount *ump, struct vnode *udvp,
743 struct unionfs_node *unp, struct componentname *cnp,
751 struct componentname cn;
754 struct ucred *credbk;
755 struct uidinfo *rootinfo;
757 if (unp->un_uppervp != NULLVP)
760 lvp = unp->un_lowervp;
762 credbk = cnp->cn_cred;
764 /* Authority change to root */
765 rootinfo = uifind((uid_t)0);
766 cred = crdup(cnp->cn_cred);
768 * The calls to chgproccnt() are needed to compensate for change_ruid()
769 * calling chgproccnt().
771 chgproccnt(cred->cr_ruidinfo, 1, 0);
772 change_euid(cred, rootinfo);
773 change_ruid(cred, rootinfo);
774 change_svuid(cred, (uid_t)0);
778 memset(&cn, 0, sizeof(cn));
780 if ((error = VOP_GETATTR(lvp, &lva, cnp->cn_cred)))
781 goto unionfs_mkshadowdir_abort;
783 if ((error = unionfs_relookup(udvp, &uvp, cnp, &cn, td, cnp->cn_nameptr, cnp->cn_namelen, CREATE)))
784 goto unionfs_mkshadowdir_abort;
792 goto unionfs_mkshadowdir_free_out;
795 if ((error = vn_start_write(udvp, &mp, V_WAIT | PCATCH)))
796 goto unionfs_mkshadowdir_free_out;
797 unionfs_create_uppervattr_core(ump, &lva, &va, td);
799 error = VOP_MKDIR(udvp, &uvp, &cn, &va);
802 unionfs_node_update(unp, uvp, td);
805 * XXX The bug which cannot set uid/gid was corrected.
809 VOP_SETATTR(uvp, &va, cn.cn_cred);
811 vn_finished_write(mp);
813 unionfs_mkshadowdir_free_out:
814 if (cn.cn_flags & HASBUF) {
815 uma_zfree(namei_zone, cn.cn_pnbuf);
816 cn.cn_flags &= ~HASBUF;
819 unionfs_mkshadowdir_abort:
820 cnp->cn_cred = credbk;
821 chgproccnt(cred->cr_ruidinfo, -1, 0);
828 * Create a new whiteout.
830 * dvp should be locked on entry and will be locked on return.
833 unionfs_mkwhiteout(struct vnode *dvp, struct componentname *cnp,
834 struct thread *td, char *path)
838 struct componentname cn;
842 path = cnp->cn_nameptr;
845 if ((error = unionfs_relookup(dvp, &wvp, cnp, &cn, td, path, strlen(path), CREATE)))
848 if (cn.cn_flags & HASBUF) {
849 uma_zfree(namei_zone, cn.cn_pnbuf);
850 cn.cn_flags &= ~HASBUF;
860 if ((error = vn_start_write(dvp, &mp, V_WAIT | PCATCH)))
861 goto unionfs_mkwhiteout_free_out;
862 error = VOP_WHITEOUT(dvp, &cn, CREATE);
864 vn_finished_write(mp);
866 unionfs_mkwhiteout_free_out:
867 if (cn.cn_flags & HASBUF) {
868 uma_zfree(namei_zone, cn.cn_pnbuf);
869 cn.cn_flags &= ~HASBUF;
876 * Create a new vnode for create a new shadow file.
878 * If an error is returned, *vpp will be invalid, otherwise it will hold a
879 * locked, referenced and opened vnode.
881 * unp is never updated.
884 unionfs_vn_create_on_upper(struct vnode **vpp, struct vnode *udvp,
885 struct unionfs_node *unp, struct vattr *uvap,
888 struct unionfs_mount *ump;
895 struct componentname cn;
897 ump = MOUNTTOUNIONFSMOUNT(UNIONFSTOV(unp)->v_mount);
899 lvp = unp->un_lowervp;
901 fmode = FFLAGS(O_WRONLY | O_CREAT | O_TRUNC | O_EXCL);
904 if ((error = VOP_GETATTR(lvp, &lva, cred)) != 0)
906 unionfs_create_uppervattr_core(ump, &lva, uvap, td);
908 if (unp->un_path == NULL)
909 panic("unionfs: un_path is null");
911 cn.cn_namelen = strlen(unp->un_path);
912 cn.cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
913 bcopy(unp->un_path, cn.cn_pnbuf, cn.cn_namelen + 1);
914 cn.cn_nameiop = CREATE;
915 cn.cn_flags = (LOCKPARENT | LOCKLEAF | HASBUF | SAVENAME | ISLASTCN);
916 cn.cn_lkflags = LK_EXCLUSIVE;
919 cn.cn_nameptr = cn.cn_pnbuf;
922 if ((error = relookup(udvp, &vp, &cn)) != 0)
923 goto unionfs_vn_create_on_upper_free_out2;
932 goto unionfs_vn_create_on_upper_free_out1;
935 if ((error = VOP_CREATE(udvp, &vp, &cn, uvap)) != 0)
936 goto unionfs_vn_create_on_upper_free_out1;
938 if ((error = VOP_OPEN(vp, fmode, cred, td, NULL)) != 0) {
940 goto unionfs_vn_create_on_upper_free_out1;
942 VOP_ADD_WRITECOUNT(vp, 1);
943 CTR3(KTR_VFS, "%s: vp %p v_writecount increased to %d", __func__, vp,
947 unionfs_vn_create_on_upper_free_out1:
948 VOP_UNLOCK(udvp, LK_RELEASE);
950 unionfs_vn_create_on_upper_free_out2:
951 if (cn.cn_flags & HASBUF) {
952 uma_zfree(namei_zone, cn.cn_pnbuf);
953 cn.cn_flags &= ~HASBUF;
960 * Copy from lvp to uvp.
962 * lvp and uvp should be locked and opened on entry and will be locked and
966 unionfs_copyfile_core(struct vnode *lvp, struct vnode *uvp,
967 struct ucred *cred, struct thread *td)
978 memset(&uio, 0, sizeof(uio));
981 uio.uio_segflg = UIO_SYSSPACE;
984 buf = malloc(MAXBSIZE, M_TEMP, M_WAITOK);
987 offset = uio.uio_offset;
992 iov.iov_len = MAXBSIZE;
993 uio.uio_resid = iov.iov_len;
994 uio.uio_rw = UIO_READ;
996 if ((error = VOP_READ(lvp, &uio, 0, cred)) != 0)
998 if ((count = MAXBSIZE - uio.uio_resid) == 0)
1002 while (bufoffset < count) {
1005 iov.iov_base = buf + bufoffset;
1006 iov.iov_len = count - bufoffset;
1007 uio.uio_offset = offset + bufoffset;
1008 uio.uio_resid = iov.iov_len;
1009 uio.uio_rw = UIO_WRITE;
1011 if ((error = VOP_WRITE(uvp, &uio, 0, cred)) != 0)
1014 bufoffset += (count - bufoffset) - uio.uio_resid;
1017 uio.uio_offset = offset + bufoffset;
1026 * Copy file from lower to upper.
1028 * If you need copy of the contents, set 1 to docopy. Otherwise, set 0 to
1031 * If no error returned, unp will be updated.
1034 unionfs_copyfile(struct unionfs_node *unp, int docopy, struct ucred *cred,
1044 lvp = unp->un_lowervp;
1047 if ((UNIONFSTOV(unp)->v_mount->mnt_flag & MNT_RDONLY))
1049 if (unp->un_dvp == NULLVP)
1051 if (unp->un_uppervp != NULLVP)
1053 udvp = VTOUNIONFS(unp->un_dvp)->un_uppervp;
1056 if ((udvp->v_mount->mnt_flag & MNT_RDONLY))
1059 error = VOP_ACCESS(lvp, VREAD, cred, td);
1063 if ((error = vn_start_write(udvp, &mp, V_WAIT | PCATCH)) != 0)
1065 error = unionfs_vn_create_on_upper(&uvp, udvp, unp, &uva, td);
1067 vn_finished_write(mp);
1072 error = VOP_OPEN(lvp, FREAD, cred, td, NULL);
1074 error = unionfs_copyfile_core(lvp, uvp, cred, td);
1075 VOP_CLOSE(lvp, FREAD, cred, td);
1078 VOP_CLOSE(uvp, FWRITE, cred, td);
1079 VOP_ADD_WRITECOUNT(uvp, -1);
1080 CTR3(KTR_VFS, "%s: vp %p v_writecount decreased to %d", __func__, uvp,
1083 vn_finished_write(mp);
1086 /* Reset the attributes. Ignore errors. */
1088 VOP_SETATTR(uvp, &uva, cred);
1091 unionfs_node_update(unp, uvp, td);
1097 * It checks whether vp can rmdir. (check empty)
1099 * vp is unionfs vnode.
1100 * vp should be locked.
1103 unionfs_check_rmdir(struct vnode *vp, struct ucred *cred, struct thread *td)
1112 struct componentname cn;
1114 * The size of buf needs to be larger than DIRBLKSIZ.
1122 ASSERT_VOP_ELOCKED(vp, "unionfs_check_rmdir");
1125 uvp = UNIONFSVPTOUPPERVP(vp);
1126 lvp = UNIONFSVPTOLOWERVP(vp);
1129 if ((error = VOP_GETATTR(uvp, &va, cred)) != 0)
1131 if (va.va_flags & OPAQUE)
1136 if ((error = mac_vnode_check_open(cred, vp, VEXEC|VREAD)) != 0)
1139 if ((error = VOP_ACCESS(vp, VEXEC|VREAD, cred, td)) != 0)
1141 if ((error = VOP_OPEN(vp, FREAD, cred, td, NULL)) != 0)
1144 uio.uio_rw = UIO_READ;
1145 uio.uio_segflg = UIO_SYSSPACE;
1150 error = mac_vnode_check_readdir(td->td_ucred, lvp);
1152 while (!error && !eofflag) {
1154 iov.iov_len = sizeof(buf);
1157 uio.uio_resid = iov.iov_len;
1159 error = VOP_READDIR(lvp, &uio, cred, &eofflag, NULL, NULL);
1162 if (eofflag == 0 && uio.uio_resid == sizeof(buf)) {
1164 panic("bad readdir response from lower FS.");
1169 edp = (struct dirent*)&buf[sizeof(buf) - uio.uio_resid];
1170 for (dp = (struct dirent*)buf; !error && dp < edp;
1171 dp = (struct dirent*)((caddr_t)dp + dp->d_reclen)) {
1172 if (dp->d_type == DT_WHT || dp->d_fileno == 0 ||
1173 (dp->d_namlen == 1 && dp->d_name[0] == '.') ||
1174 (dp->d_namlen == 2 && !bcmp(dp->d_name, "..", 2)))
1177 cn.cn_namelen = dp->d_namlen;
1179 cn.cn_nameptr = dp->d_name;
1180 cn.cn_nameiop = LOOKUP;
1181 cn.cn_flags = (LOCKPARENT | LOCKLEAF | SAVENAME | RDONLY | ISLASTCN);
1182 cn.cn_lkflags = LK_EXCLUSIVE;
1187 * check entry in lower.
1188 * Sometimes, readdir function returns
1191 lookuperr = VOP_LOOKUP(lvp, &tvp, &cn);
1196 continue; /* skip entry */
1200 * If it has no exist/whiteout entry in upper,
1201 * directory is not empty.
1203 cn.cn_flags = (LOCKPARENT | LOCKLEAF | SAVENAME | RDONLY | ISLASTCN);
1204 lookuperr = VOP_LOOKUP(uvp, &tvp, &cn);
1209 /* ignore exist or whiteout entry */
1211 (lookuperr == ENOENT && (cn.cn_flags & ISWHITEOUT)))
1219 VOP_CLOSE(vp, FREAD, cred, td);
1227 unionfs_checkuppervp(struct vnode *vp, char *fil, int lno)
1229 struct unionfs_node *unp;
1231 unp = VTOUNIONFS(vp);
1234 if (vp->v_op != unionfs_vnodeop_p) {
1235 printf("unionfs_checkuppervp: on non-unionfs-node.\n");
1237 kdb_enter(KDB_WHY_UNIONFS,
1238 "unionfs_checkuppervp: on non-unionfs-node.\n");
1240 panic("unionfs_checkuppervp");
1243 return (unp->un_uppervp);
1247 unionfs_checklowervp(struct vnode *vp, char *fil, int lno)
1249 struct unionfs_node *unp;
1251 unp = VTOUNIONFS(vp);
1254 if (vp->v_op != unionfs_vnodeop_p) {
1255 printf("unionfs_checklowervp: on non-unionfs-node.\n");
1257 kdb_enter(KDB_WHY_UNIONFS,
1258 "unionfs_checklowervp: on non-unionfs-node.\n");
1260 panic("unionfs_checklowervp");
1263 return (unp->un_lowervp);