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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2013, 2016 by Delphix. All rights reserved.
25 * Copyright 2017 Nexenta Systems, Inc.
28 #include <sys/types.h>
29 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
35 #include <sys/vnode.h>
39 #include <sys/cmn_err.h>
40 #include <sys/errno.h>
42 #include <sys/unistd.h>
43 #include <sys/sunddi.h>
44 #include <sys/random.h>
45 #include <sys/policy.h>
46 #include <sys/kcondvar.h>
47 #include <sys/callb.h>
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_acl.h>
51 #include <sys/fs/zfs.h>
54 #include <sys/atomic.h>
55 #include <sys/zfs_ctldir.h>
56 #include <sys/zfs_fuid.h>
58 #include <sys/zfs_sa.h>
60 #include <sys/extdirent.h>
63 * zfs_match_find() is used by zfs_dirent_lookup() to peform zap lookups
64 * of names after deciding which is the appropriate lookup interface.
67 zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, const char *name,
68 matchtype_t mt, uint64_t *zoid)
75 * In the non-mixed case we only expect there would ever
76 * be one match, but we need to use the normalizing lookup.
78 error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1,
79 zoid, mt, NULL, 0, NULL);
81 error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid);
83 *zoid = ZFS_DIRENT_OBJ(*zoid);
89 * Look up a directory entry under a locked vnode.
90 * dvp being locked gives us a guarantee that there are no concurrent
91 * modification of the directory and, thus, if a node can be found in
92 * the directory, then it must not be unlinked.
95 * dzp - znode for directory
96 * name - name of entry to lock
97 * flag - ZNEW: if the entry already exists, fail with EEXIST.
98 * ZEXISTS: if the entry does not exist, fail with ENOENT.
99 * ZXATTR: we want dzp's xattr directory
102 * zpp - pointer to the znode for the entry (NULL if there isn't one)
104 * Return value: 0 on success or errno on failure.
106 * NOTE: Always checks for, and rejects, '.' and '..'.
109 zfs_dirent_lookup(znode_t *dzp, const char *name, znode_t **zpp, int flag)
111 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
117 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
122 * Verify that we are not trying to lock '.', '..', or '.zfs'
124 if (name[0] == '.' &&
125 (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) ||
126 zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
127 return (SET_ERROR(EEXIST));
130 * Case sensitivity and normalization preferences are set when
131 * the file system is created. These are stored in the
132 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices
133 * affect how we perform zap lookups.
135 * When matching we may need to normalize & change case according to
138 * Note that a normalized match is necessary for a case insensitive
139 * filesystem when the lookup request is not exact because normalization
140 * can fold case independent of normalizing code point sequences.
142 * See the table above zfs_dropname().
144 if (zfsvfs->z_norm != 0) {
148 * Determine if the match needs to honor the case specified in
149 * lookup, and if so keep track of that so that during
150 * normalization we don't fold case.
152 if (zfsvfs->z_case == ZFS_CASE_MIXED) {
158 * Only look in or update the DNLC if we are looking for the
159 * name on a file system that does not require normalization
160 * or case folding. We can also look there if we happen to be
161 * on a non-normalizing, mixed sensitivity file system IF we
162 * are looking for the exact name.
164 * NB: we do not need to worry about this flag for ZFS_CASE_SENSITIVE
165 * because in that case MT_EXACT and MT_FIRST should produce exactly
169 if (dzp->z_unlinked && !(flag & ZXATTR))
172 error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
175 error = (zoid == 0 ? ENOENT : 0);
177 error = zfs_match_find(zfsvfs, dzp, name, mt, &zoid);
180 if (error != ENOENT || (flag & ZEXISTS)) {
185 return (SET_ERROR(EEXIST));
187 error = zfs_zget(zfsvfs, zoid, zpp);
190 ASSERT(!(*zpp)->z_unlinked);
197 zfs_dd_lookup(znode_t *dzp, znode_t **zpp)
199 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
204 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
205 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock));
210 if ((error = sa_lookup(dzp->z_sa_hdl,
211 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
214 error = zfs_zget(zfsvfs, parent, &zp);
221 zfs_dirlook(znode_t *dzp, const char *name, znode_t **zpp)
223 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
227 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
228 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock));
231 return (SET_ERROR(ENOENT));
233 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
235 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
236 error = zfs_dd_lookup(dzp, zpp);
238 error = zfs_dirent_lookup(dzp, name, &zp, ZEXISTS);
240 dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
248 * unlinked Set (formerly known as the "delete queue") Error Handling
250 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
251 * don't specify the name of the entry that we will be manipulating. We
252 * also fib and say that we won't be adding any new entries to the
253 * unlinked set, even though we might (this is to lower the minimum file
254 * size that can be deleted in a full filesystem). So on the small
255 * chance that the nlink list is using a fat zap (ie. has more than
256 * 2000 entries), we *may* not pre-read a block that's needed.
257 * Therefore it is remotely possible for some of the assertions
258 * regarding the unlinked set below to fail due to i/o error. On a
259 * nondebug system, this will result in the space being leaked.
262 zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx)
264 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
266 ASSERT(zp->z_unlinked);
267 ASSERT(zp->z_links == 0);
270 zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
274 * Clean up any znodes that had no links when we either crashed or
275 * (force) umounted the file system.
278 zfs_unlinked_drain(zfsvfs_t *zfsvfs)
282 dmu_object_info_t doi;
288 * Interate over the contents of the unlinked set.
290 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
291 zap_cursor_retrieve(&zc, &zap) == 0;
292 zap_cursor_advance(&zc)) {
295 * See what kind of object we have in list
298 error = dmu_object_info(zfsvfs->z_os,
299 zap.za_first_integer, &doi);
303 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
304 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
306 * We need to re-mark these list entries for deletion,
307 * so we pull them back into core and set zp->z_unlinked.
309 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
312 * We may pick up znodes that are already marked for deletion.
313 * This could happen during the purge of an extended attribute
314 * directory. All we need to do is skip over them, since they
315 * are already in the system marked z_unlinked.
320 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY);
321 zp->z_unlinked = B_TRUE;
322 #if defined(__FreeBSD__)
324 * Due to changes in zfs_rmnode we need to make sure the
325 * link count is set to zero here.
328 tx = dmu_tx_create(zfsvfs->z_os);
329 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
330 VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));
331 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
332 &zp->z_links, sizeof (zp->z_links), tx));
337 zap_cursor_fini(&zc);
341 * Delete the entire contents of a directory. Return a count
342 * of the number of entries that could not be deleted. If we encounter
343 * an error, return a count of at least one so that the directory stays
344 * in the unlinked set.
346 * NOTE: this function assumes that the directory is inactive,
347 * so there is no need to lock its entries before deletion.
348 * Also, it assumes the directory contents is *only* regular
352 zfs_purgedir(znode_t *dzp)
358 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
362 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
363 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
364 zap_cursor_advance(&zc)) {
365 error = zfs_zget(zfsvfs,
366 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
372 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
373 ASSERT((ZTOV(xzp)->v_type == VREG) ||
374 (ZTOV(xzp)->v_type == VLNK));
376 tx = dmu_tx_create(zfsvfs->z_os);
377 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
378 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
379 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
380 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
381 /* Is this really needed ? */
382 zfs_sa_upgrade_txholds(tx, xzp);
383 dmu_tx_mark_netfree(tx);
384 error = dmu_tx_assign(tx, TXG_WAIT);
392 error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL);
399 zap_cursor_fini(&zc);
406 zfs_rmnode(znode_t *zp)
408 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
409 objset_t *os = zfsvfs->z_os;
415 ASSERT(zp->z_links == 0);
416 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
419 * If this is an attribute directory, purge its contents.
421 if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR &&
422 (zp->z_pflags & ZFS_XATTR)) {
423 if (zfs_purgedir(zp) != 0) {
425 * Not enough space to delete some xattrs.
426 * Leave it in the unlinked set.
428 zfs_znode_dmu_fini(zp);
434 * Free up all the data in the file. We don't do this for
435 * XATTR directories because we need truncate and remove to be
436 * in the same tx, like in zfs_znode_delete(). Otherwise, if
437 * we crash here we'll end up with an inconsistent truncated
438 * zap object in the delete queue. Note a truncated file is
439 * harmless since it only contains user data.
441 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
444 * Not enough space or we were interrupted by unmount.
445 * Leave the file in the unlinked set.
447 zfs_znode_dmu_fini(zp);
454 * If the file has extended attributes, we're going to unlink
457 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
458 &xattr_obj, sizeof (xattr_obj));
462 acl_obj = zfs_external_acl(zp);
465 * Set up the final transaction.
467 tx = dmu_tx_create(os);
468 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
469 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
471 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
473 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
475 zfs_sa_upgrade_txholds(tx, zp);
476 error = dmu_tx_assign(tx, TXG_WAIT);
479 * Not enough space to delete the file. Leave it in the
480 * unlinked set, leaking it until the fs is remounted (at
481 * which point we'll call zfs_unlinked_drain() to process it).
484 zfs_znode_dmu_fini(zp);
489 #if defined(__FreeBSD__)
491 * FreeBSD's implemention of zfs_zget requires a vnode to back it.
492 * This means that we could end up calling into getnewvnode while
493 * calling zfs_rmnode as a result of a prior call to getnewvnode
494 * trying to clear vnodes out of the cache. If this repeats we can
495 * recurse enough that we overflow our stack. To avoid this, we
496 * avoid calling zfs_zget on the xattr znode and instead simply add
497 * it to the unlinked set and schedule a call to zfs_unlinked_drain.
500 /* Add extended attribute directory to the unlinked set. */
502 zap_add_int(os, zfsvfs->z_unlinkedobj, xattr_obj, tx));
507 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
508 xzp->z_links = 0; /* no more links to it */
509 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
510 &xzp->z_links, sizeof (xzp->z_links), tx));
511 zfs_unlinked_add(xzp, tx);
515 /* Remove this znode from the unlinked set */
517 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
519 zfs_znode_delete(zp, tx);
525 * We're using the FreeBSD taskqueue API here instead of
526 * the Solaris taskq API since the FreeBSD API allows for a
527 * task to be enqueued multiple times but executed once.
529 taskqueue_enqueue(system_taskq->tq_queue,
530 &zfsvfs->z_unlinked_drain_task);
535 zfs_dirent(znode_t *zp, uint64_t mode)
537 uint64_t de = zp->z_id;
539 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
540 de |= IFTODT(mode) << 60;
545 * Link zp into dzp. Can only fail if zp has been unlinked.
548 zfs_link_create(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
551 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
552 vnode_t *vp = ZTOV(zp);
554 int zp_is_dir = (vp->v_type == VDIR);
555 sa_bulk_attr_t bulk[5];
556 uint64_t mtime[2], ctime[2];
560 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
561 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
564 if (dzp->z_links >= ZFS_LINK_MAX)
565 return (SET_ERROR(EMLINK));
568 if (!(flag & ZRENAMING)) {
569 if (zp->z_unlinked) { /* no new links to unlinked zp */
570 ASSERT(!(flag & (ZNEW | ZEXISTS)));
571 return (SET_ERROR(ENOENT));
574 if (zp->z_links >= ZFS_LINK_MAX - zp_is_dir) {
575 return (SET_ERROR(EMLINK));
579 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
580 &zp->z_links, sizeof (zp->z_links));
583 ASSERT(zp->z_unlinked == 0);
585 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
586 &dzp->z_id, sizeof (dzp->z_id));
587 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
588 &zp->z_pflags, sizeof (zp->z_pflags));
590 if (!(flag & ZNEW)) {
591 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
592 ctime, sizeof (ctime));
593 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
596 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
600 dzp->z_links += zp_is_dir;
602 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
603 &dzp->z_size, sizeof (dzp->z_size));
604 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
605 &dzp->z_links, sizeof (dzp->z_links));
606 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
607 mtime, sizeof (mtime));
608 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
609 ctime, sizeof (ctime));
610 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
611 &dzp->z_pflags, sizeof (dzp->z_pflags));
612 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
613 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
616 value = zfs_dirent(zp, zp->z_mode);
617 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name,
625 * The match type in the code for this function should conform to:
627 * ------------------------------------------------------------------------
628 * fs type | z_norm | lookup type | match type
629 * ---------|-------------|-------------|----------------------------------
630 * CS !norm | 0 | 0 | 0 (exact)
631 * CS norm | formX | 0 | MT_NORMALIZE
632 * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE
633 * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
634 * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE
635 * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
636 * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
637 * CM !norm | upper | ZCILOOK | MT_NORMALIZE
638 * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
639 * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE
642 * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed
643 * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER)
644 * formX = unicode normalization form set on fs creation
647 zfs_dropname(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
652 if (zp->z_zfsvfs->z_norm) {
653 matchtype_t mt = MT_NORMALIZE;
655 if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) {
659 error = zap_remove_norm(zp->z_zfsvfs->z_os, dzp->z_id,
662 error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, name, tx);
669 * Unlink zp from dzp, and mark zp for deletion if this was the last link.
670 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
671 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
672 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
673 * and it's the caller's job to do it.
676 zfs_link_destroy(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
677 int flag, boolean_t *unlinkedp)
679 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
680 vnode_t *vp = ZTOV(zp);
681 int zp_is_dir = (vp->v_type == VDIR);
682 boolean_t unlinked = B_FALSE;
683 sa_bulk_attr_t bulk[5];
684 uint64_t mtime[2], ctime[2];
688 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
689 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
691 if (!(flag & ZRENAMING)) {
693 if (zp_is_dir && !zfs_dirempty(zp)) {
695 return (SET_ERROR(EEXIST));
697 return (SET_ERROR(ENOTEMPTY));
702 * If we get here, we are going to try to remove the object.
703 * First try removing the name from the directory; if that
704 * fails, return the error.
706 error = zfs_dropname(dzp, name, zp, tx, flag);
711 if (zp->z_links <= zp_is_dir) {
712 zfs_panic_recover("zfs: link count on vnode %p is %u, "
713 "should be at least %u", zp->z_vnode,
716 zp->z_links = zp_is_dir + 1;
718 if (--zp->z_links == zp_is_dir) {
719 zp->z_unlinked = B_TRUE;
723 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
724 NULL, &ctime, sizeof (ctime));
725 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
726 NULL, &zp->z_pflags, sizeof (zp->z_pflags));
727 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
730 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
731 NULL, &zp->z_links, sizeof (zp->z_links));
732 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
736 ASSERT(zp->z_unlinked == 0);
737 error = zfs_dropname(dzp, name, zp, tx, flag);
742 dzp->z_size--; /* one dirent removed */
743 dzp->z_links -= zp_is_dir; /* ".." link from zp */
744 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
745 NULL, &dzp->z_links, sizeof (dzp->z_links));
746 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
747 NULL, &dzp->z_size, sizeof (dzp->z_size));
748 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
749 NULL, ctime, sizeof (ctime));
750 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
751 NULL, mtime, sizeof (mtime));
752 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
753 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
754 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
755 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
758 if (unlinkedp != NULL)
759 *unlinkedp = unlinked;
761 zfs_unlinked_add(zp, tx);
767 * Indicate whether the directory is empty.
770 zfs_dirempty(znode_t *dzp)
772 return (dzp->z_size == 2);
776 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
778 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
782 zfs_acl_ids_t acl_ids;
783 boolean_t fuid_dirtied;
789 * In FreeBSD, access checking for creating an EA is being done
790 * in zfs_setextattr(),
792 #ifndef __FreeBSD_kernel__
793 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
797 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
800 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
801 zfs_acl_ids_free(&acl_ids);
802 return (SET_ERROR(EDQUOT));
805 getnewvnode_reserve(1);
807 tx = dmu_tx_create(zfsvfs->z_os);
808 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
809 ZFS_SA_BASE_ATTR_SIZE);
810 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
811 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
812 fuid_dirtied = zfsvfs->z_fuid_dirty;
814 zfs_fuid_txhold(zfsvfs, tx);
815 error = dmu_tx_assign(tx, TXG_WAIT);
817 zfs_acl_ids_free(&acl_ids);
821 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
824 zfs_fuid_sync(zfsvfs, tx);
827 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
828 &parent, sizeof (parent));
829 ASSERT(error == 0 && parent == zp->z_id);
832 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
833 sizeof (xzp->z_id), tx));
835 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
836 xzp, "", NULL, acl_ids.z_fuidp, vap);
838 zfs_acl_ids_free(&acl_ids);
841 getnewvnode_drop_reserve();
849 * Return a znode for the extended attribute directory for zp.
850 * ** If the directory does not already exist, it is created **
852 * IN: zp - znode to obtain attribute directory from
853 * cr - credentials of caller
854 * flags - flags from the VOP_LOOKUP call
856 * OUT: xzpp - pointer to extended attribute znode
858 * RETURN: 0 on success
859 * error number on failure
862 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
864 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
869 error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR);
879 if (!(flags & CREATE_XATTR_DIR)) {
881 return (SET_ERROR(ENOENT));
883 return (SET_ERROR(ENOATTR));
887 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
888 return (SET_ERROR(EROFS));
892 * The ability to 'create' files in an attribute
893 * directory comes from the write_xattr permission on the base file.
895 * The ability to 'search' an attribute directory requires
896 * read_xattr permission on the base file.
898 * Once in a directory the ability to read/write attributes
899 * is controlled by the permissions on the attribute file.
901 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
903 va.va_mode = S_IFDIR | S_ISVTX | 0777;
904 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
906 error = zfs_make_xattrdir(zp, &va, xvpp, cr);
908 if (error == ERESTART) {
909 /* NB: we already did dmu_tx_wait() if necessary */
913 VOP_UNLOCK(*xvpp, 0);
919 * Decide whether it is okay to remove within a sticky directory.
921 * In sticky directories, write access is not sufficient;
922 * you can remove entries from a directory only if:
924 * you own the directory,
926 * the entry is a plain file and you have write access,
927 * or you are privileged (checked in secpolicy...).
929 * The function returns 0 if remove access is granted.
932 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
937 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
939 if (zdp->z_zfsvfs->z_replay)
942 if ((zdp->z_mode & S_ISVTX) == 0)
945 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
946 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);
948 if ((uid = crgetuid(cr)) == downer || uid == fowner ||
949 (ZTOV(zp)->v_type == VREG &&
950 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
953 return (secpolicy_vnode_remove(ZTOV(zp), cr));