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.
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
39 #include <sys/pathname.h>
40 #include <sys/cmn_err.h>
41 #include <sys/errno.h>
43 #include <sys/unistd.h>
44 #include <sys/sunddi.h>
45 #include <sys/random.h>
46 #include <sys/policy.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/fs/zfs.h>
50 #include "fs/fs_subr.h"
53 #include <sys/atomic.h>
54 #include <sys/zfs_ctldir.h>
55 #include <sys/zfs_fuid.h>
57 #include <sys/zfs_sa.h>
59 #include <sys/extdirent.h>
62 * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups
63 * of names after deciding which is the appropriate lookup interface.
66 zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, char *name, boolean_t exact,
67 boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid)
72 matchtype_t mt = MT_FIRST;
73 boolean_t conflict = B_FALSE;
79 bufsz = rpnp->pn_bufsize;
84 * In the non-mixed case we only expect there would ever
85 * be one match, but we need to use the normalizing lookup.
87 error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1,
88 zoid, mt, buf, bufsz, &conflict);
89 if (!error && deflags)
90 *deflags = conflict ? ED_CASE_CONFLICT : 0;
92 error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid);
94 *zoid = ZFS_DIRENT_OBJ(*zoid);
96 if (error == ENOENT && update)
97 dnlc_update(ZTOV(dzp), name, DNLC_NO_VNODE);
103 * Lock a directory entry. A dirlock on <dzp, name> protects that name
104 * in dzp's directory zap object. As long as you hold a dirlock, you can
105 * assume two things: (1) dzp cannot be reaped, and (2) no other thread
106 * can change the zap entry for (i.e. link or unlink) this name.
109 * dzp - znode for directory
110 * name - name of entry to lock
111 * flag - ZNEW: if the entry already exists, fail with EEXIST.
112 * ZEXISTS: if the entry does not exist, fail with ENOENT.
113 * ZSHARED: allow concurrent access with other ZSHARED callers.
114 * ZXATTR: we want dzp's xattr directory
115 * ZCILOOK: On a mixed sensitivity file system,
116 * this lookup should be case-insensitive.
117 * ZCIEXACT: On a purely case-insensitive file system,
118 * this lookup should be case-sensitive.
119 * ZRENAMING: we are locking for renaming, force narrow locks
120 * ZHAVELOCK: Don't grab the z_name_lock for this call. The
121 * current thread already holds it.
124 * zpp - pointer to the znode for the entry (NULL if there isn't one)
125 * dlpp - pointer to the dirlock for this entry (NULL on error)
126 * direntflags - (case-insensitive lookup only)
127 * flags if multiple case-sensitive matches exist in directory
128 * realpnp - (case-insensitive lookup only)
129 * actual name matched within the directory
131 * Return value: 0 on success or errno on failure.
133 * NOTE: Always checks for, and rejects, '.' and '..'.
134 * NOTE: For case-insensitive file systems we take wide locks (see below),
135 * but return znode pointers to a single match.
138 zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp,
139 int flag, int *direntflags, pathname_t *realpnp)
141 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
154 * Verify that we are not trying to lock '.', '..', or '.zfs'
156 if (name[0] == '.' &&
157 (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) ||
158 zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
162 * Case sensitivity and normalization preferences are set when
163 * the file system is created. These are stored in the
164 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices
165 * affect what vnodes can be cached in the DNLC, how we
166 * perform zap lookups, and the "width" of our dirlocks.
168 * A normal dirlock locks a single name. Note that with
169 * normalization a name can be composed multiple ways, but
170 * when normalized, these names all compare equal. A wide
171 * dirlock locks multiple names. We need these when the file
172 * system is supporting mixed-mode access. It is sometimes
173 * necessary to lock all case permutations of file name at
174 * once so that simultaneous case-insensitive/case-sensitive
175 * behaves as rationally as possible.
179 * Decide if exact matches should be requested when performing
180 * a zap lookup on file systems supporting case-insensitive
184 ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) ||
185 ((zfsvfs->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK));
188 * Only look in or update the DNLC if we are looking for the
189 * name on a file system that does not require normalization
190 * or case folding. We can also look there if we happen to be
191 * on a non-normalizing, mixed sensitivity file system IF we
192 * are looking for the exact name.
194 * Maybe can add TO-UPPERed version of name to dnlc in ci-only
195 * case for performance improvement?
197 update = !zfsvfs->z_norm ||
198 ((zfsvfs->z_case == ZFS_CASE_MIXED) &&
199 !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK));
202 * ZRENAMING indicates we are in a situation where we should
203 * take narrow locks regardless of the file system's
204 * preferences for normalizing and case folding. This will
205 * prevent us deadlocking trying to grab the same wide lock
206 * twice if the two names happen to be case-insensitive
209 if (flag & ZRENAMING)
212 cmpflags = zfsvfs->z_norm;
215 * Wait until there are no locks on this name.
217 * Don't grab the the lock if it is already held. However, cannot
218 * have both ZSHARED and ZHAVELOCK together.
220 ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK));
221 if (!(flag & ZHAVELOCK))
222 rw_enter(&dzp->z_name_lock, RW_READER);
224 mutex_enter(&dzp->z_lock);
226 if (dzp->z_unlinked) {
227 mutex_exit(&dzp->z_lock);
228 if (!(flag & ZHAVELOCK))
229 rw_exit(&dzp->z_name_lock);
232 for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) {
233 if ((u8_strcmp(name, dl->dl_name, 0, cmpflags,
234 U8_UNICODE_LATEST, &error) == 0) || error != 0)
238 mutex_exit(&dzp->z_lock);
239 if (!(flag & ZHAVELOCK))
240 rw_exit(&dzp->z_name_lock);
245 * Allocate a new dirlock and add it to the list.
247 dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP);
248 cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL);
254 dl->dl_next = dzp->z_dirlocks;
255 dzp->z_dirlocks = dl;
258 if ((flag & ZSHARED) && dl->dl_sharecnt != 0)
260 cv_wait(&dl->dl_cv, &dzp->z_lock);
264 * If the z_name_lock was NOT held for this dirlock record it.
266 if (flag & ZHAVELOCK)
269 if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) {
271 * We're the second shared reference to dl. Make a copy of
272 * dl_name in case the first thread goes away before we do.
273 * Note that we initialize the new name before storing its
274 * pointer into dl_name, because the first thread may load
275 * dl->dl_name at any time. He'll either see the old value,
276 * which is his, or the new shared copy; either is OK.
278 dl->dl_namesize = strlen(dl->dl_name) + 1;
279 name = kmem_alloc(dl->dl_namesize, KM_SLEEP);
280 bcopy(dl->dl_name, name, dl->dl_namesize);
284 mutex_exit(&dzp->z_lock);
287 * We have a dirlock on the name. (Note that it is the dirlock,
288 * not the dzp's z_lock, that protects the name in the zap object.)
289 * See if there's an object by this name; if so, put a hold on it.
292 error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
295 error = (zoid == 0 ? ENOENT : 0);
298 vp = dnlc_lookup(ZTOV(dzp), name);
299 if (vp == DNLC_NO_VNODE) {
304 zfs_dirent_unlock(dl);
312 error = zfs_match_find(zfsvfs, dzp, name, exact,
313 update, direntflags, realpnp, &zoid);
317 if (error != ENOENT || (flag & ZEXISTS)) {
318 zfs_dirent_unlock(dl);
323 zfs_dirent_unlock(dl);
326 error = zfs_zget(zfsvfs, zoid, zpp);
328 zfs_dirent_unlock(dl);
331 if (!(flag & ZXATTR) && update)
332 dnlc_update(ZTOV(dzp), name, ZTOV(*zpp));
341 * Unlock this directory entry and wake anyone who was waiting for it.
344 zfs_dirent_unlock(zfs_dirlock_t *dl)
346 znode_t *dzp = dl->dl_dzp;
347 zfs_dirlock_t **prev_dl, *cur_dl;
349 mutex_enter(&dzp->z_lock);
351 if (!dl->dl_namelock)
352 rw_exit(&dzp->z_name_lock);
354 if (dl->dl_sharecnt > 1) {
356 mutex_exit(&dzp->z_lock);
359 prev_dl = &dzp->z_dirlocks;
360 while ((cur_dl = *prev_dl) != dl)
361 prev_dl = &cur_dl->dl_next;
362 *prev_dl = dl->dl_next;
363 cv_broadcast(&dl->dl_cv);
364 mutex_exit(&dzp->z_lock);
366 if (dl->dl_namesize != 0)
367 kmem_free(dl->dl_name, dl->dl_namesize);
368 cv_destroy(&dl->dl_cv);
369 kmem_free(dl, sizeof (*dl));
373 * Look up an entry in a directory.
375 * NOTE: '.' and '..' are handled as special cases because
376 * no directory entries are actually stored for them. If this is
377 * the root of a filesystem, then '.zfs' is also treated as a
378 * special pseudo-directory.
381 zfs_dirlook(znode_t *dzp, char *name, vnode_t **vpp, int flags,
382 int *deflg, pathname_t *rpnp)
389 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
392 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
393 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
396 * If we are a snapshot mounted under .zfs, return
397 * the vp for the snapshot directory.
399 if ((error = sa_lookup(dzp->z_sa_hdl,
400 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
402 if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) {
403 error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
404 "snapshot", vpp, NULL, 0, NULL, kcred,
408 rw_enter(&dzp->z_parent_lock, RW_READER);
409 error = zfs_zget(zfsvfs, parent, &zp);
412 rw_exit(&dzp->z_parent_lock);
413 } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
414 *vpp = zfsctl_root(dzp);
418 zf = ZEXISTS | ZSHARED;
419 if (flags & FIGNORECASE)
422 error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp);
425 zfs_dirent_unlock(dl);
426 dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
431 if ((flags & FIGNORECASE) && rpnp && !error)
432 (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize);
438 * unlinked Set (formerly known as the "delete queue") Error Handling
440 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
441 * don't specify the name of the entry that we will be manipulating. We
442 * also fib and say that we won't be adding any new entries to the
443 * unlinked set, even though we might (this is to lower the minimum file
444 * size that can be deleted in a full filesystem). So on the small
445 * chance that the nlink list is using a fat zap (ie. has more than
446 * 2000 entries), we *may* not pre-read a block that's needed.
447 * Therefore it is remotely possible for some of the assertions
448 * regarding the unlinked set below to fail due to i/o error. On a
449 * nondebug system, this will result in the space being leaked.
452 zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx)
454 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
456 ASSERT(zp->z_unlinked);
457 ASSERT(zp->z_links == 0);
460 zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
464 * Clean up any znodes that had no links when we either crashed or
465 * (force) umounted the file system.
468 zfs_unlinked_drain(zfsvfs_t *zfsvfs)
472 dmu_object_info_t doi;
477 * Interate over the contents of the unlinked set.
479 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
480 zap_cursor_retrieve(&zc, &zap) == 0;
481 zap_cursor_advance(&zc)) {
484 * See what kind of object we have in list
487 error = dmu_object_info(zfsvfs->z_os,
488 zap.za_first_integer, &doi);
492 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
493 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
495 * We need to re-mark these list entries for deletion,
496 * so we pull them back into core and set zp->z_unlinked.
498 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
501 * We may pick up znodes that are already marked for deletion.
502 * This could happen during the purge of an extended attribute
503 * directory. All we need to do is skip over them, since they
504 * are already in the system marked z_unlinked.
509 zp->z_unlinked = B_TRUE;
512 zap_cursor_fini(&zc);
516 * Delete the entire contents of a directory. Return a count
517 * of the number of entries that could not be deleted. If we encounter
518 * an error, return a count of at least one so that the directory stays
519 * in the unlinked set.
521 * NOTE: this function assumes that the directory is inactive,
522 * so there is no need to lock its entries before deletion.
523 * Also, it assumes the directory contents is *only* regular
527 zfs_purgedir(znode_t *dzp)
533 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
538 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
539 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
540 zap_cursor_advance(&zc)) {
541 error = zfs_zget(zfsvfs,
542 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
548 ASSERT((ZTOV(xzp)->v_type == VREG) ||
549 (ZTOV(xzp)->v_type == VLNK));
551 tx = dmu_tx_create(zfsvfs->z_os);
552 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
553 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
554 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
555 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
556 /* Is this really needed ? */
557 zfs_sa_upgrade_txholds(tx, xzp);
558 error = dmu_tx_assign(tx, TXG_WAIT);
565 bzero(&dl, sizeof (dl));
567 dl.dl_name = zap.za_name;
569 error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
576 zap_cursor_fini(&zc);
583 zfs_rmnode(znode_t *zp)
585 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
586 objset_t *os = zfsvfs->z_os;
593 ASSERT(zp->z_links == 0);
594 ASSERT(ZTOV(zp)->v_count == 0);
597 * If this is an attribute directory, purge its contents.
599 if (ZTOV(zp)->v_type == VDIR && (zp->z_pflags & ZFS_XATTR)) {
600 if (zfs_purgedir(zp) != 0) {
602 * Not enough space to delete some xattrs.
603 * Leave it in the unlinked set.
605 zfs_znode_dmu_fini(zp);
612 * Free up all the data in the file.
614 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
617 * Not enough space. Leave the file in the unlinked set.
619 zfs_znode_dmu_fini(zp);
625 * If the file has extended attributes, we're going to unlink
628 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
629 &xattr_obj, sizeof (xattr_obj));
630 if (error == 0 && xattr_obj) {
631 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
635 acl_obj = zfs_external_acl(zp);
638 * Set up the final transaction.
640 tx = dmu_tx_create(os);
641 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
642 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
644 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
645 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
648 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
650 zfs_sa_upgrade_txholds(tx, zp);
651 error = dmu_tx_assign(tx, TXG_WAIT);
654 * Not enough space to delete the file. Leave it in the
655 * unlinked set, leaking it until the fs is remounted (at
656 * which point we'll call zfs_unlinked_drain() to process it).
659 zfs_znode_dmu_fini(zp);
666 mutex_enter(&xzp->z_lock);
667 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
668 xzp->z_links = 0; /* no more links to it */
669 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
670 &xzp->z_links, sizeof (xzp->z_links), tx));
671 mutex_exit(&xzp->z_lock);
672 zfs_unlinked_add(xzp, tx);
675 /* Remove this znode from the unlinked set */
677 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
679 zfs_znode_delete(zp, tx);
688 zfs_dirent(znode_t *zp, uint64_t mode)
690 uint64_t de = zp->z_id;
692 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
693 de |= IFTODT(mode) << 60;
698 * Link zp into dl. Can only fail if zp has been unlinked.
701 zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
703 znode_t *dzp = dl->dl_dzp;
704 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
705 vnode_t *vp = ZTOV(zp);
707 int zp_is_dir = (vp->v_type == VDIR);
708 sa_bulk_attr_t bulk[5];
709 uint64_t mtime[2], ctime[2];
713 mutex_enter(&zp->z_lock);
715 if (!(flag & ZRENAMING)) {
716 if (zp->z_unlinked) { /* no new links to unlinked zp */
717 ASSERT(!(flag & (ZNEW | ZEXISTS)));
718 mutex_exit(&zp->z_lock);
722 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
723 &zp->z_links, sizeof (zp->z_links));
726 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
727 &dzp->z_id, sizeof (dzp->z_id));
728 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
729 &zp->z_pflags, sizeof (zp->z_pflags));
731 if (!(flag & ZNEW)) {
732 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
733 ctime, sizeof (ctime));
734 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
737 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
740 mutex_exit(&zp->z_lock);
742 mutex_enter(&dzp->z_lock);
744 dzp->z_links += zp_is_dir;
746 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
747 &dzp->z_size, sizeof (dzp->z_size));
748 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
749 &dzp->z_links, sizeof (dzp->z_links));
750 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
751 mtime, sizeof (mtime));
752 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
753 ctime, sizeof (ctime));
754 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
755 &dzp->z_pflags, sizeof (dzp->z_pflags));
756 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
757 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
759 mutex_exit(&dzp->z_lock);
761 value = zfs_dirent(zp, zp->z_mode);
762 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
766 dnlc_update(ZTOV(dzp), dl->dl_name, vp);
772 zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx,
777 if (zp->z_zfsvfs->z_norm) {
778 if (((zp->z_zfsvfs->z_case == ZFS_CASE_INSENSITIVE) &&
779 (flag & ZCIEXACT)) ||
780 ((zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) &&
782 error = zap_remove_norm(zp->z_zfsvfs->z_os,
783 dzp->z_id, dl->dl_name, MT_EXACT, tx);
785 error = zap_remove_norm(zp->z_zfsvfs->z_os,
786 dzp->z_id, dl->dl_name, MT_FIRST, tx);
788 error = zap_remove(zp->z_zfsvfs->z_os,
789 dzp->z_id, dl->dl_name, tx);
796 * Unlink zp from dl, and mark zp for deletion if this was the last link.
797 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
798 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
799 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
800 * and it's the caller's job to do it.
803 zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
804 boolean_t *unlinkedp)
806 znode_t *dzp = dl->dl_dzp;
807 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
808 vnode_t *vp = ZTOV(zp);
809 int zp_is_dir = (vp->v_type == VDIR);
810 boolean_t unlinked = B_FALSE;
811 sa_bulk_attr_t bulk[5];
812 uint64_t mtime[2], ctime[2];
816 dnlc_remove(ZTOV(dzp), dl->dl_name);
818 if (!(flag & ZRENAMING)) {
819 if (vn_vfswlock(vp)) /* prevent new mounts on zp */
822 if (vn_ismntpt(vp)) { /* don't remove mount point */
827 mutex_enter(&zp->z_lock);
829 if (zp_is_dir && !zfs_dirempty(zp)) {
830 mutex_exit(&zp->z_lock);
836 * If we get here, we are going to try to remove the object.
837 * First try removing the name from the directory; if that
838 * fails, return the error.
840 error = zfs_dropname(dl, zp, dzp, tx, flag);
842 mutex_exit(&zp->z_lock);
847 if (zp->z_links <= zp_is_dir) {
848 zfs_panic_recover("zfs: link count on %s is %u, "
849 "should be at least %u",
850 zp->z_vnode->v_path ? zp->z_vnode->v_path :
851 "<unknown>", (int)zp->z_links,
853 zp->z_links = zp_is_dir + 1;
855 if (--zp->z_links == zp_is_dir) {
856 zp->z_unlinked = B_TRUE;
860 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
861 NULL, &ctime, sizeof (ctime));
862 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
863 NULL, &zp->z_pflags, sizeof (zp->z_pflags));
864 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
867 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
868 NULL, &zp->z_links, sizeof (zp->z_links));
869 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
872 mutex_exit(&zp->z_lock);
875 error = zfs_dropname(dl, zp, dzp, tx, flag);
880 mutex_enter(&dzp->z_lock);
881 dzp->z_size--; /* one dirent removed */
882 dzp->z_links -= zp_is_dir; /* ".." link from zp */
883 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
884 NULL, &dzp->z_links, sizeof (dzp->z_links));
885 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
886 NULL, &dzp->z_size, sizeof (dzp->z_size));
887 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
888 NULL, ctime, sizeof (ctime));
889 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
890 NULL, mtime, sizeof (mtime));
891 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
892 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
893 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
894 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
896 mutex_exit(&dzp->z_lock);
898 if (unlinkedp != NULL)
899 *unlinkedp = unlinked;
901 zfs_unlinked_add(zp, tx);
907 * Indicate whether the directory is empty. Works with or without z_lock
908 * held, but can only be consider a hint in the latter case. Returns true
909 * if only "." and ".." remain and there's no work in progress.
912 zfs_dirempty(znode_t *dzp)
914 return (dzp->z_size == 2 && dzp->z_dirlocks == 0);
918 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
920 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
924 zfs_acl_ids_t acl_ids;
925 boolean_t fuid_dirtied;
930 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
933 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
936 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
937 zfs_acl_ids_free(&acl_ids);
942 tx = dmu_tx_create(zfsvfs->z_os);
943 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
944 ZFS_SA_BASE_ATTR_SIZE);
945 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
946 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
947 fuid_dirtied = zfsvfs->z_fuid_dirty;
949 zfs_fuid_txhold(zfsvfs, tx);
950 error = dmu_tx_assign(tx, TXG_NOWAIT);
952 if (error == ERESTART) {
957 zfs_acl_ids_free(&acl_ids);
961 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
964 zfs_fuid_sync(zfsvfs, tx);
967 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
968 &parent, sizeof (parent));
969 ASSERT(error == 0 && parent == zp->z_id);
972 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
973 sizeof (xzp->z_id), tx));
975 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
976 xzp, "", NULL, acl_ids.z_fuidp, vap);
978 zfs_acl_ids_free(&acl_ids);
987 * Return a znode for the extended attribute directory for zp.
988 * ** If the directory does not already exist, it is created **
990 * IN: zp - znode to obtain attribute directory from
991 * cr - credentials of caller
992 * flags - flags from the VOP_LOOKUP call
994 * OUT: xzpp - pointer to extended attribute znode
996 * RETURN: 0 on success
997 * error number on failure
1000 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
1002 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1008 error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL);
1014 zfs_dirent_unlock(dl);
1019 if (!(flags & CREATE_XATTR_DIR)) {
1020 zfs_dirent_unlock(dl);
1024 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
1025 zfs_dirent_unlock(dl);
1030 * The ability to 'create' files in an attribute
1031 * directory comes from the write_xattr permission on the base file.
1033 * The ability to 'search' an attribute directory requires
1034 * read_xattr permission on the base file.
1036 * Once in a directory the ability to read/write attributes
1037 * is controlled by the permissions on the attribute file.
1039 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
1041 va.va_mode = S_IFDIR | S_ISVTX | 0777;
1042 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
1044 error = zfs_make_xattrdir(zp, &va, xvpp, cr);
1045 zfs_dirent_unlock(dl);
1047 if (error == ERESTART) {
1048 /* NB: we already did dmu_tx_wait() if necessary */
1056 * Decide whether it is okay to remove within a sticky directory.
1058 * In sticky directories, write access is not sufficient;
1059 * you can remove entries from a directory only if:
1061 * you own the directory,
1062 * you own the entry,
1063 * the entry is a plain file and you have write access,
1064 * or you are privileged (checked in secpolicy...).
1066 * The function returns 0 if remove access is granted.
1069 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
1074 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1076 if (zdp->z_zfsvfs->z_replay)
1079 if ((zdp->z_mode & S_ISVTX) == 0)
1082 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
1083 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);
1085 if ((uid = crgetuid(cr)) == downer || uid == fowner ||
1086 (ZTOV(zp)->v_type == VREG &&
1087 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
1090 return (secpolicy_vnode_remove(cr));
1092 #endif /* HAVE_ZPL */