4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 #include <sys/types.h>
26 #include <sys/param.h>
28 #include <sys/systm.h>
29 #include <sys/sysmacros.h>
30 #include <sys/resource.h>
32 #include <sys/vnode.h>
36 #include <sys/cmn_err.h>
37 #include <sys/errno.h>
39 #include <sys/unistd.h>
40 #include <sys/sunddi.h>
41 #include <sys/random.h>
42 #include <sys/policy.h>
43 #include <sys/kcondvar.h>
44 #include <sys/callb.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/fs/zfs.h>
51 #include <sys/atomic.h>
52 #include <sys/zfs_ctldir.h>
53 #include <sys/zfs_fuid.h>
55 #include <sys/zfs_sa.h>
57 #include <sys/extdirent.h>
60 * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups
61 * of names after deciding which is the appropriate lookup interface.
64 zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, char *name, boolean_t exact,
65 boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid)
70 matchtype_t mt = MT_FIRST;
71 boolean_t conflict = B_FALSE;
77 bufsz = rpnp->pn_bufsize;
82 * In the non-mixed case we only expect there would ever
83 * be one match, but we need to use the normalizing lookup.
85 error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1,
86 zoid, mt, buf, bufsz, &conflict);
87 if (!error && deflags)
88 *deflags = conflict ? ED_CASE_CONFLICT : 0;
90 error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid);
92 *zoid = ZFS_DIRENT_OBJ(*zoid);
94 if (error == ENOENT && update)
95 dnlc_update(ZTOV(dzp), name, DNLC_NO_VNODE);
101 * Lock a directory entry. A dirlock on <dzp, name> protects that name
102 * in dzp's directory zap object. As long as you hold a dirlock, you can
103 * assume two things: (1) dzp cannot be reaped, and (2) no other thread
104 * can change the zap entry for (i.e. link or unlink) this name.
107 * dzp - znode for directory
108 * name - name of entry to lock
109 * flag - ZNEW: if the entry already exists, fail with EEXIST.
110 * ZEXISTS: if the entry does not exist, fail with ENOENT.
111 * ZSHARED: allow concurrent access with other ZSHARED callers.
112 * ZXATTR: we want dzp's xattr directory
113 * ZCILOOK: On a mixed sensitivity file system,
114 * this lookup should be case-insensitive.
115 * ZCIEXACT: On a purely case-insensitive file system,
116 * this lookup should be case-sensitive.
117 * ZRENAMING: we are locking for renaming, force narrow locks
118 * ZHAVELOCK: Don't grab the z_name_lock for this call. The
119 * current thread already holds it.
122 * zpp - pointer to the znode for the entry (NULL if there isn't one)
123 * dlpp - pointer to the dirlock for this entry (NULL on error)
124 * direntflags - (case-insensitive lookup only)
125 * flags if multiple case-sensitive matches exist in directory
126 * realpnp - (case-insensitive lookup only)
127 * actual name matched within the directory
129 * Return value: 0 on success or errno on failure.
131 * NOTE: Always checks for, and rejects, '.' and '..'.
132 * NOTE: For case-insensitive file systems we take wide locks (see below),
133 * but return znode pointers to a single match.
136 zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp,
137 int flag, int *direntflags, pathname_t *realpnp)
139 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
152 * Verify that we are not trying to lock '.', '..', or '.zfs'
154 if (name[0] == '.' &&
155 (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) ||
156 zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
160 * Case sensitivity and normalization preferences are set when
161 * the file system is created. These are stored in the
162 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices
163 * affect what vnodes can be cached in the DNLC, how we
164 * perform zap lookups, and the "width" of our dirlocks.
166 * A normal dirlock locks a single name. Note that with
167 * normalization a name can be composed multiple ways, but
168 * when normalized, these names all compare equal. A wide
169 * dirlock locks multiple names. We need these when the file
170 * system is supporting mixed-mode access. It is sometimes
171 * necessary to lock all case permutations of file name at
172 * once so that simultaneous case-insensitive/case-sensitive
173 * behaves as rationally as possible.
177 * Decide if exact matches should be requested when performing
178 * a zap lookup on file systems supporting case-insensitive
182 ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) ||
183 ((zfsvfs->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK));
186 * Only look in or update the DNLC if we are looking for the
187 * name on a file system that does not require normalization
188 * or case folding. We can also look there if we happen to be
189 * on a non-normalizing, mixed sensitivity file system IF we
190 * are looking for the exact name.
192 * Maybe can add TO-UPPERed version of name to dnlc in ci-only
193 * case for performance improvement?
195 update = !zfsvfs->z_norm ||
196 ((zfsvfs->z_case == ZFS_CASE_MIXED) &&
197 !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK));
200 * ZRENAMING indicates we are in a situation where we should
201 * take narrow locks regardless of the file system's
202 * preferences for normalizing and case folding. This will
203 * prevent us deadlocking trying to grab the same wide lock
204 * twice if the two names happen to be case-insensitive
207 if (flag & ZRENAMING)
210 cmpflags = zfsvfs->z_norm;
213 * Wait until there are no locks on this name.
215 * Don't grab the the lock if it is already held. However, cannot
216 * have both ZSHARED and ZHAVELOCK together.
218 ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK));
219 if (!(flag & ZHAVELOCK))
220 rw_enter(&dzp->z_name_lock, RW_READER);
222 mutex_enter(&dzp->z_lock);
224 if (dzp->z_unlinked) {
225 mutex_exit(&dzp->z_lock);
226 if (!(flag & ZHAVELOCK))
227 rw_exit(&dzp->z_name_lock);
230 for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) {
231 if ((u8_strcmp(name, dl->dl_name, 0, cmpflags,
232 U8_UNICODE_LATEST, &error) == 0) || error != 0)
236 mutex_exit(&dzp->z_lock);
237 if (!(flag & ZHAVELOCK))
238 rw_exit(&dzp->z_name_lock);
245 * Allocate a new dirlock and add it to the list.
247 namesize = strlen(name) + 1;
248 dl = kmem_alloc(sizeof (zfs_dirlock_t) + namesize,
250 cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL);
251 dl->dl_name = (char *)(dl + 1);
252 bcopy(name, dl->dl_name, namesize);
255 dl->dl_namesize = namesize;
257 dl->dl_next = dzp->z_dirlocks;
258 dzp->z_dirlocks = dl;
261 if ((flag & ZSHARED) && dl->dl_sharecnt != 0)
263 cv_wait(&dl->dl_cv, &dzp->z_lock);
267 * If the z_name_lock was NOT held for this dirlock record it.
269 if (flag & ZHAVELOCK)
275 mutex_exit(&dzp->z_lock);
278 * We have a dirlock on the name. (Note that it is the dirlock,
279 * not the dzp's z_lock, that protects the name in the zap object.)
280 * See if there's an object by this name; if so, put a hold on it.
283 error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
286 error = (zoid == 0 ? ENOENT : 0);
289 vp = dnlc_lookup(ZTOV(dzp), name);
290 if (vp == DNLC_NO_VNODE) {
295 zfs_dirent_unlock(dl);
303 error = zfs_match_find(zfsvfs, dzp, name, exact,
304 update, direntflags, realpnp, &zoid);
308 if (error != ENOENT || (flag & ZEXISTS)) {
309 zfs_dirent_unlock(dl);
314 zfs_dirent_unlock(dl);
317 error = zfs_zget(zfsvfs, zoid, zpp);
319 zfs_dirent_unlock(dl);
322 if (!(flag & ZXATTR) && update)
323 dnlc_update(ZTOV(dzp), name, ZTOV(*zpp));
332 * Unlock this directory entry and wake anyone who was waiting for it.
335 zfs_dirent_unlock(zfs_dirlock_t *dl)
337 znode_t *dzp = dl->dl_dzp;
338 zfs_dirlock_t **prev_dl, *cur_dl;
340 mutex_enter(&dzp->z_lock);
342 if (!dl->dl_namelock)
343 rw_exit(&dzp->z_name_lock);
345 if (dl->dl_sharecnt > 1) {
347 mutex_exit(&dzp->z_lock);
350 prev_dl = &dzp->z_dirlocks;
351 while ((cur_dl = *prev_dl) != dl)
352 prev_dl = &cur_dl->dl_next;
353 *prev_dl = dl->dl_next;
354 cv_broadcast(&dl->dl_cv);
355 mutex_exit(&dzp->z_lock);
357 cv_destroy(&dl->dl_cv);
358 kmem_free(dl, sizeof (*dl) + dl->dl_namesize);
362 * Look up an entry in a directory.
364 * NOTE: '.' and '..' are handled as special cases because
365 * no directory entries are actually stored for them. If this is
366 * the root of a filesystem, then '.zfs' is also treated as a
367 * special pseudo-directory.
370 zfs_dirlook(znode_t *dzp, char *name, vnode_t **vpp, int flags,
371 int *deflg, pathname_t *rpnp)
379 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
380 mutex_enter(&dzp->z_lock);
381 unlinked = dzp->z_unlinked;
382 mutex_exit(&dzp->z_lock);
388 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
389 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
392 * If we are a snapshot mounted under .zfs, return
393 * the vp for the snapshot directory.
395 if ((error = sa_lookup(dzp->z_sa_hdl,
396 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
398 if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) {
399 error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
400 "snapshot", vpp, NULL, 0, NULL, kcred,
405 mutex_enter(&dzp->z_lock);
406 unlinked = dzp->z_unlinked;
407 mutex_exit(&dzp->z_lock);
411 rw_enter(&dzp->z_parent_lock, RW_READER);
412 error = zfs_zget(zfsvfs, parent, &zp);
415 rw_exit(&dzp->z_parent_lock);
416 } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
417 *vpp = zfsctl_root(dzp);
421 zf = ZEXISTS | ZSHARED;
422 if (flags & FIGNORECASE)
425 error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp);
428 zfs_dirent_unlock(dl);
429 dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
434 if ((flags & FIGNORECASE) && rpnp && !error)
435 (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize);
441 * unlinked Set (formerly known as the "delete queue") Error Handling
443 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
444 * don't specify the name of the entry that we will be manipulating. We
445 * also fib and say that we won't be adding any new entries to the
446 * unlinked set, even though we might (this is to lower the minimum file
447 * size that can be deleted in a full filesystem). So on the small
448 * chance that the nlink list is using a fat zap (ie. has more than
449 * 2000 entries), we *may* not pre-read a block that's needed.
450 * Therefore it is remotely possible for some of the assertions
451 * regarding the unlinked set below to fail due to i/o error. On a
452 * nondebug system, this will result in the space being leaked.
455 zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx)
457 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
459 ASSERT(zp->z_unlinked);
460 ASSERT(zp->z_links == 0);
463 zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
467 * Clean up any znodes that had no links when we either crashed or
468 * (force) umounted the file system.
471 zfs_unlinked_drain(zfsvfs_t *zfsvfs)
475 dmu_object_info_t doi;
480 * Interate over the contents of the unlinked set.
482 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
483 zap_cursor_retrieve(&zc, &zap) == 0;
484 zap_cursor_advance(&zc)) {
487 * See what kind of object we have in list
490 error = dmu_object_info(zfsvfs->z_os,
491 zap.za_first_integer, &doi);
495 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
496 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
498 * We need to re-mark these list entries for deletion,
499 * so we pull them back into core and set zp->z_unlinked.
501 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
504 * We may pick up znodes that are already marked for deletion.
505 * This could happen during the purge of an extended attribute
506 * directory. All we need to do is skip over them, since they
507 * are already in the system marked z_unlinked.
512 zp->z_unlinked = B_TRUE;
515 zap_cursor_fini(&zc);
519 * Delete the entire contents of a directory. Return a count
520 * of the number of entries that could not be deleted. If we encounter
521 * an error, return a count of at least one so that the directory stays
522 * in the unlinked set.
524 * NOTE: this function assumes that the directory is inactive,
525 * so there is no need to lock its entries before deletion.
526 * Also, it assumes the directory contents is *only* regular
530 zfs_purgedir(znode_t *dzp)
536 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
541 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
542 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
543 zap_cursor_advance(&zc)) {
544 error = zfs_zget(zfsvfs,
545 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
551 ASSERT((ZTOV(xzp)->v_type == VREG) ||
552 (ZTOV(xzp)->v_type == VLNK));
554 tx = dmu_tx_create(zfsvfs->z_os);
555 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
556 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
557 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
558 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
559 /* Is this really needed ? */
560 zfs_sa_upgrade_txholds(tx, xzp);
561 error = dmu_tx_assign(tx, TXG_WAIT);
568 bzero(&dl, sizeof (dl));
570 dl.dl_name = zap.za_name;
572 error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
579 zap_cursor_fini(&zc);
586 zfs_rmnode(znode_t *zp)
588 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
589 objset_t *os = zfsvfs->z_os;
596 ASSERT(zp->z_links == 0);
599 * If this is an attribute directory, purge its contents.
601 if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR &&
602 (zp->z_pflags & ZFS_XATTR)) {
603 if (zfs_purgedir(zp) != 0) {
605 * Not enough space to delete some xattrs.
606 * Leave it in the unlinked set.
608 zfs_znode_dmu_fini(zp);
615 * Free up all the data in the file.
617 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
620 * Not enough space. Leave the file in the unlinked set.
622 zfs_znode_dmu_fini(zp);
628 * If the file has extended attributes, we're going to unlink
631 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
632 &xattr_obj, sizeof (xattr_obj));
633 if (error == 0 && xattr_obj) {
634 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
638 acl_obj = zfs_external_acl(zp);
641 * Set up the final transaction.
643 tx = dmu_tx_create(os);
644 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
645 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
647 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
648 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
651 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
653 zfs_sa_upgrade_txholds(tx, zp);
654 error = dmu_tx_assign(tx, TXG_WAIT);
657 * Not enough space to delete the file. Leave it in the
658 * unlinked set, leaking it until the fs is remounted (at
659 * which point we'll call zfs_unlinked_drain() to process it).
662 zfs_znode_dmu_fini(zp);
669 mutex_enter(&xzp->z_lock);
670 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
671 xzp->z_links = 0; /* no more links to it */
672 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
673 &xzp->z_links, sizeof (xzp->z_links), tx));
674 mutex_exit(&xzp->z_lock);
675 zfs_unlinked_add(xzp, tx);
678 /* Remove this znode from the unlinked set */
680 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
682 zfs_znode_delete(zp, tx);
691 zfs_dirent(znode_t *zp, uint64_t mode)
693 uint64_t de = zp->z_id;
695 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
696 de |= IFTODT(mode) << 60;
701 * Link zp into dl. Can only fail if zp has been unlinked.
704 zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
706 znode_t *dzp = dl->dl_dzp;
707 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
708 vnode_t *vp = ZTOV(zp);
710 int zp_is_dir = (vp->v_type == VDIR);
711 sa_bulk_attr_t bulk[5];
712 uint64_t mtime[2], ctime[2];
716 mutex_enter(&zp->z_lock);
718 if (!(flag & ZRENAMING)) {
719 if (zp->z_unlinked) { /* no new links to unlinked zp */
720 ASSERT(!(flag & (ZNEW | ZEXISTS)));
721 mutex_exit(&zp->z_lock);
725 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
726 &zp->z_links, sizeof (zp->z_links));
729 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
730 &dzp->z_id, sizeof (dzp->z_id));
731 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
732 &zp->z_pflags, sizeof (zp->z_pflags));
734 if (!(flag & ZNEW)) {
735 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
736 ctime, sizeof (ctime));
737 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
740 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
743 mutex_exit(&zp->z_lock);
745 mutex_enter(&dzp->z_lock);
747 dzp->z_links += zp_is_dir;
749 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
750 &dzp->z_size, sizeof (dzp->z_size));
751 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
752 &dzp->z_links, sizeof (dzp->z_links));
753 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
754 mtime, sizeof (mtime));
755 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
756 ctime, sizeof (ctime));
757 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
758 &dzp->z_pflags, sizeof (dzp->z_pflags));
759 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
760 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
762 mutex_exit(&dzp->z_lock);
764 value = zfs_dirent(zp, zp->z_mode);
765 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
769 dnlc_update(ZTOV(dzp), dl->dl_name, vp);
775 zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx,
780 if (zp->z_zfsvfs->z_norm) {
781 if (((zp->z_zfsvfs->z_case == ZFS_CASE_INSENSITIVE) &&
782 (flag & ZCIEXACT)) ||
783 ((zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) &&
785 error = zap_remove_norm(zp->z_zfsvfs->z_os,
786 dzp->z_id, dl->dl_name, MT_EXACT, tx);
788 error = zap_remove_norm(zp->z_zfsvfs->z_os,
789 dzp->z_id, dl->dl_name, MT_FIRST, tx);
791 error = zap_remove(zp->z_zfsvfs->z_os,
792 dzp->z_id, dl->dl_name, tx);
799 * Unlink zp from dl, and mark zp for deletion if this was the last link.
800 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
801 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
802 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
803 * and it's the caller's job to do it.
806 zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
807 boolean_t *unlinkedp)
809 znode_t *dzp = dl->dl_dzp;
810 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
811 vnode_t *vp = ZTOV(zp);
812 int zp_is_dir = (vp->v_type == VDIR);
813 boolean_t unlinked = B_FALSE;
814 sa_bulk_attr_t bulk[5];
815 uint64_t mtime[2], ctime[2];
819 dnlc_remove(ZTOV(dzp), dl->dl_name);
821 if (!(flag & ZRENAMING)) {
822 if (vn_vfswlock(vp)) /* prevent new mounts on zp */
825 if (vn_ismntpt(vp)) { /* don't remove mount point */
830 mutex_enter(&zp->z_lock);
832 if (zp_is_dir && !zfs_dirempty(zp)) {
833 mutex_exit(&zp->z_lock);
839 * If we get here, we are going to try to remove the object.
840 * First try removing the name from the directory; if that
841 * fails, return the error.
843 error = zfs_dropname(dl, zp, dzp, tx, flag);
845 mutex_exit(&zp->z_lock);
850 if (zp->z_links <= zp_is_dir) {
851 zfs_panic_recover("zfs: link count on vnode %p is %u, "
852 "should be at least %u", zp->z_vnode,
855 zp->z_links = zp_is_dir + 1;
857 if (--zp->z_links == zp_is_dir) {
858 zp->z_unlinked = B_TRUE;
862 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
863 NULL, &ctime, sizeof (ctime));
864 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
865 NULL, &zp->z_pflags, sizeof (zp->z_pflags));
866 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
869 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
870 NULL, &zp->z_links, sizeof (zp->z_links));
871 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
874 mutex_exit(&zp->z_lock);
877 error = zfs_dropname(dl, zp, dzp, tx, flag);
882 mutex_enter(&dzp->z_lock);
883 dzp->z_size--; /* one dirent removed */
884 dzp->z_links -= zp_is_dir; /* ".." link from zp */
885 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
886 NULL, &dzp->z_links, sizeof (dzp->z_links));
887 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
888 NULL, &dzp->z_size, sizeof (dzp->z_size));
889 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
890 NULL, ctime, sizeof (ctime));
891 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
892 NULL, mtime, sizeof (mtime));
893 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
894 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
895 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
896 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
898 mutex_exit(&dzp->z_lock);
900 if (unlinkedp != NULL)
901 *unlinkedp = unlinked;
903 zfs_unlinked_add(zp, tx);
909 * Indicate whether the directory is empty. Works with or without z_lock
910 * held, but can only be consider a hint in the latter case. Returns true
911 * if only "." and ".." remain and there's no work in progress.
914 zfs_dirempty(znode_t *dzp)
916 return (dzp->z_size == 2 && dzp->z_dirlocks == 0);
920 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
922 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
926 zfs_acl_ids_t acl_ids;
927 boolean_t fuid_dirtied;
933 * In FreeBSD, access checking for creating an EA is being done
934 * in zfs_setextattr(),
937 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
941 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
944 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
945 zfs_acl_ids_free(&acl_ids);
950 tx = dmu_tx_create(zfsvfs->z_os);
951 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
952 ZFS_SA_BASE_ATTR_SIZE);
953 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
954 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
955 fuid_dirtied = zfsvfs->z_fuid_dirty;
957 zfs_fuid_txhold(zfsvfs, tx);
958 error = dmu_tx_assign(tx, TXG_NOWAIT);
960 if (error == ERESTART) {
965 zfs_acl_ids_free(&acl_ids);
969 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
972 zfs_fuid_sync(zfsvfs, tx);
975 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
976 &parent, sizeof (parent));
977 ASSERT(error == 0 && parent == zp->z_id);
980 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
981 sizeof (xzp->z_id), tx));
983 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
984 xzp, "", NULL, acl_ids.z_fuidp, vap);
986 zfs_acl_ids_free(&acl_ids);
995 * Return a znode for the extended attribute directory for zp.
996 * ** If the directory does not already exist, it is created **
998 * IN: zp - znode to obtain attribute directory from
999 * cr - credentials of caller
1000 * flags - flags from the VOP_LOOKUP call
1002 * OUT: xzpp - pointer to extended attribute znode
1004 * RETURN: 0 on success
1005 * error number on failure
1008 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
1010 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1016 error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL);
1022 zfs_dirent_unlock(dl);
1027 if (!(flags & CREATE_XATTR_DIR)) {
1028 zfs_dirent_unlock(dl);
1036 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
1037 zfs_dirent_unlock(dl);
1042 * The ability to 'create' files in an attribute
1043 * directory comes from the write_xattr permission on the base file.
1045 * The ability to 'search' an attribute directory requires
1046 * read_xattr permission on the base file.
1048 * Once in a directory the ability to read/write attributes
1049 * is controlled by the permissions on the attribute file.
1051 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
1053 va.va_mode = S_IFDIR | S_ISVTX | 0777;
1054 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
1056 error = zfs_make_xattrdir(zp, &va, xvpp, cr);
1057 zfs_dirent_unlock(dl);
1059 if (error == ERESTART) {
1060 /* NB: we already did dmu_tx_wait() if necessary */
1064 VOP_UNLOCK(*xvpp, 0);
1070 * Decide whether it is okay to remove within a sticky directory.
1072 * In sticky directories, write access is not sufficient;
1073 * you can remove entries from a directory only if:
1075 * you own the directory,
1076 * you own the entry,
1077 * the entry is a plain file and you have write access,
1078 * or you are privileged (checked in secpolicy...).
1080 * The function returns 0 if remove access is granted.
1083 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
1088 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1090 if (zdp->z_zfsvfs->z_replay)
1093 if ((zdp->z_mode & S_ISVTX) == 0)
1096 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
1097 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);
1099 if ((uid = crgetuid(cr)) == downer || uid == fowner ||
1100 (ZTOV(zp)->v_type == VREG &&
1101 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
1104 return (secpolicy_vnode_remove(ZTOV(zp), cr));