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 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
26 #include <sys/types.h>
27 #include <sys/param.h>
29 #include <sys/systm.h>
30 #include <sys/sysmacros.h>
31 #include <sys/resource.h>
33 #include <sys/vnode.h>
38 #include <sys/pathname.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/zfs_dir.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/fs/zfs.h>
49 #include "fs/fs_subr.h"
52 #include <sys/atomic.h>
53 #include <sys/zfs_ctldir.h>
54 #include <sys/zfs_fuid.h>
56 #include <sys/extdirent.h>
59 * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups
60 * of names after deciding which is the appropriate lookup interface.
63 zfs_match_find(zfsvfs_t
*zfsvfs
, znode_t
*dzp
, char *name
, boolean_t exact
,
64 boolean_t update
, int *deflags
, pathname_t
*rpnp
, uint64_t *zoid
)
69 matchtype_t mt
= MT_FIRST
;
70 boolean_t conflict
= B_FALSE
;
76 bufsz
= rpnp
->pn_bufsize
;
81 * In the non-mixed case we only expect there would ever
82 * be one match, but we need to use the normalizing lookup.
84 error
= zap_lookup_norm(zfsvfs
->z_os
, dzp
->z_id
, name
, 8, 1,
85 zoid
, mt
, buf
, bufsz
, &conflict
);
86 if (!error
&& deflags
)
87 *deflags
= conflict
? ED_CASE_CONFLICT
: 0;
89 error
= zap_lookup(zfsvfs
->z_os
, dzp
->z_id
, name
, 8, 1, zoid
);
91 *zoid
= ZFS_DIRENT_OBJ(*zoid
);
93 if (error
== ENOENT
&& update
)
94 dnlc_update(ZTOV(dzp
), name
, DNLC_NO_VNODE
);
100 * Lock a directory entry. A dirlock on <dzp, name> protects that name
101 * in dzp's directory zap object. As long as you hold a dirlock, you can
102 * assume two things: (1) dzp cannot be reaped, and (2) no other thread
103 * can change the zap entry for (i.e. link or unlink) this name.
106 * dzp - znode for directory
107 * name - name of entry to lock
108 * flag - ZNEW: if the entry already exists, fail with EEXIST.
109 * ZEXISTS: if the entry does not exist, fail with ENOENT.
110 * ZSHARED: allow concurrent access with other ZSHARED callers.
111 * ZXATTR: we want dzp's xattr directory
112 * ZCILOOK: On a mixed sensitivity file system,
113 * this lookup should be case-insensitive.
114 * ZCIEXACT: On a purely case-insensitive file system,
115 * this lookup should be case-sensitive.
116 * ZRENAMING: we are locking for renaming, force narrow locks
119 * zpp - pointer to the znode for the entry (NULL if there isn't one)
120 * dlpp - pointer to the dirlock for this entry (NULL on error)
121 * direntflags - (case-insensitive lookup only)
122 * flags if multiple case-sensitive matches exist in directory
123 * realpnp - (case-insensitive lookup only)
124 * actual name matched within the directory
126 * Return value: 0 on success or errno on failure.
128 * NOTE: Always checks for, and rejects, '.' and '..'.
129 * NOTE: For case-insensitive file systems we take wide locks (see below),
130 * but return znode pointers to a single match.
133 zfs_dirent_lock(zfs_dirlock_t
**dlpp
, znode_t
*dzp
, char *name
, znode_t
**zpp
,
134 int flag
, int *direntflags
, pathname_t
*realpnp
)
136 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
149 * Verify that we are not trying to lock '.', '..', or '.zfs'
151 if (name
[0] == '.' &&
152 (name
[1] == '\0' || (name
[1] == '.' && name
[2] == '\0')) ||
153 zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0)
157 * Case sensitivity and normalization preferences are set when
158 * the file system is created. These are stored in the
159 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices
160 * affect what vnodes can be cached in the DNLC, how we
161 * perform zap lookups, and the "width" of our dirlocks.
163 * A normal dirlock locks a single name. Note that with
164 * normalization a name can be composed multiple ways, but
165 * when normalized, these names all compare equal. A wide
166 * dirlock locks multiple names. We need these when the file
167 * system is supporting mixed-mode access. It is sometimes
168 * necessary to lock all case permutations of file name at
169 * once so that simultaneous case-insensitive/case-sensitive
170 * behaves as rationally as possible.
174 * Decide if exact matches should be requested when performing
175 * a zap lookup on file systems supporting case-insensitive
179 ((zfsvfs
->z_case
== ZFS_CASE_INSENSITIVE
) && (flag
& ZCIEXACT
)) ||
180 ((zfsvfs
->z_case
== ZFS_CASE_MIXED
) && !(flag
& ZCILOOK
));
183 * Only look in or update the DNLC if we are looking for the
184 * name on a file system that does not require normalization
185 * or case folding. We can also look there if we happen to be
186 * on a non-normalizing, mixed sensitivity file system IF we
187 * are looking for the exact name.
189 * Maybe can add TO-UPPERed version of name to dnlc in ci-only
190 * case for performance improvement?
192 update
= !zfsvfs
->z_norm
||
193 ((zfsvfs
->z_case
== ZFS_CASE_MIXED
) &&
194 !(zfsvfs
->z_norm
& ~U8_TEXTPREP_TOUPPER
) && !(flag
& ZCILOOK
));
197 * ZRENAMING indicates we are in a situation where we should
198 * take narrow locks regardless of the file system's
199 * preferences for normalizing and case folding. This will
200 * prevent us deadlocking trying to grab the same wide lock
201 * twice if the two names happen to be case-insensitive
204 if (flag
& ZRENAMING
)
207 cmpflags
= zfsvfs
->z_norm
;
210 * Wait until there are no locks on this name.
212 rw_enter(&dzp
->z_name_lock
, RW_READER
);
213 mutex_enter(&dzp
->z_lock
);
215 if (dzp
->z_unlinked
) {
216 mutex_exit(&dzp
->z_lock
);
217 rw_exit(&dzp
->z_name_lock
);
220 for (dl
= dzp
->z_dirlocks
; dl
!= NULL
; dl
= dl
->dl_next
) {
221 if ((u8_strcmp(name
, dl
->dl_name
, 0, cmpflags
,
222 U8_UNICODE_LATEST
, &error
) == 0) || error
!= 0)
226 mutex_exit(&dzp
->z_lock
);
227 rw_exit(&dzp
->z_name_lock
);
232 * Allocate a new dirlock and add it to the list.
234 dl
= kmem_alloc(sizeof (zfs_dirlock_t
), KM_SLEEP
);
235 cv_init(&dl
->dl_cv
, NULL
, CV_DEFAULT
, NULL
);
240 dl
->dl_next
= dzp
->z_dirlocks
;
241 dzp
->z_dirlocks
= dl
;
244 if ((flag
& ZSHARED
) && dl
->dl_sharecnt
!= 0)
246 cv_wait(&dl
->dl_cv
, &dzp
->z_lock
);
249 if ((flag
& ZSHARED
) && ++dl
->dl_sharecnt
> 1 && dl
->dl_namesize
== 0) {
251 * We're the second shared reference to dl. Make a copy of
252 * dl_name in case the first thread goes away before we do.
253 * Note that we initialize the new name before storing its
254 * pointer into dl_name, because the first thread may load
255 * dl->dl_name at any time. He'll either see the old value,
256 * which is his, or the new shared copy; either is OK.
258 dl
->dl_namesize
= strlen(dl
->dl_name
) + 1;
259 name
= kmem_alloc(dl
->dl_namesize
, KM_SLEEP
);
260 bcopy(dl
->dl_name
, name
, dl
->dl_namesize
);
264 mutex_exit(&dzp
->z_lock
);
267 * We have a dirlock on the name. (Note that it is the dirlock,
268 * not the dzp's z_lock, that protects the name in the zap object.)
269 * See if there's an object by this name; if so, put a hold on it.
272 zoid
= dzp
->z_phys
->zp_xattr
;
273 error
= (zoid
== 0 ? ENOENT
: 0);
276 vp
= dnlc_lookup(ZTOV(dzp
), name
);
277 if (vp
== DNLC_NO_VNODE
) {
282 zfs_dirent_unlock(dl
);
290 error
= zfs_match_find(zfsvfs
, dzp
, name
, exact
,
291 update
, direntflags
, realpnp
, &zoid
);
295 if (error
!= ENOENT
|| (flag
& ZEXISTS
)) {
296 zfs_dirent_unlock(dl
);
301 zfs_dirent_unlock(dl
);
304 error
= zfs_zget(zfsvfs
, zoid
, zpp
);
306 zfs_dirent_unlock(dl
);
309 if (!(flag
& ZXATTR
) && update
)
310 dnlc_update(ZTOV(dzp
), name
, ZTOV(*zpp
));
319 * Unlock this directory entry and wake anyone who was waiting for it.
322 zfs_dirent_unlock(zfs_dirlock_t
*dl
)
324 znode_t
*dzp
= dl
->dl_dzp
;
325 zfs_dirlock_t
**prev_dl
, *cur_dl
;
327 mutex_enter(&dzp
->z_lock
);
328 rw_exit(&dzp
->z_name_lock
);
329 if (dl
->dl_sharecnt
> 1) {
331 mutex_exit(&dzp
->z_lock
);
334 prev_dl
= &dzp
->z_dirlocks
;
335 while ((cur_dl
= *prev_dl
) != dl
)
336 prev_dl
= &cur_dl
->dl_next
;
337 *prev_dl
= dl
->dl_next
;
338 cv_broadcast(&dl
->dl_cv
);
339 mutex_exit(&dzp
->z_lock
);
341 if (dl
->dl_namesize
!= 0)
342 kmem_free(dl
->dl_name
, dl
->dl_namesize
);
343 cv_destroy(&dl
->dl_cv
);
344 kmem_free(dl
, sizeof (*dl
));
348 * Look up an entry in a directory.
350 * NOTE: '.' and '..' are handled as special cases because
351 * no directory entries are actually stored for them. If this is
352 * the root of a filesystem, then '.zfs' is also treated as a
353 * special pseudo-directory.
356 zfs_dirlook(znode_t
*dzp
, char *name
, vnode_t
**vpp
, int flags
,
357 int *deflg
, pathname_t
*rpnp
)
363 if (name
[0] == 0 || (name
[0] == '.' && name
[1] == 0)) {
366 } else if (name
[0] == '.' && name
[1] == '.' && name
[2] == 0) {
367 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
369 * If we are a snapshot mounted under .zfs, return
370 * the vp for the snapshot directory.
372 if (dzp
->z_phys
->zp_parent
== dzp
->z_id
&&
373 zfsvfs
->z_parent
!= zfsvfs
) {
374 error
= zfsctl_root_lookup(zfsvfs
->z_parent
->z_ctldir
,
375 "snapshot", vpp
, NULL
, 0, NULL
, kcred
,
379 rw_enter(&dzp
->z_parent_lock
, RW_READER
);
380 error
= zfs_zget(zfsvfs
, dzp
->z_phys
->zp_parent
, &zp
);
383 rw_exit(&dzp
->z_parent_lock
);
384 } else if (zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0) {
385 *vpp
= zfsctl_root(dzp
);
389 zf
= ZEXISTS
| ZSHARED
;
390 if (flags
& FIGNORECASE
)
393 error
= zfs_dirent_lock(&dl
, dzp
, name
, &zp
, zf
, deflg
, rpnp
);
396 zfs_dirent_unlock(dl
);
397 dzp
->z_zn_prefetch
= B_TRUE
; /* enable prefetching */
402 if ((flags
& FIGNORECASE
) && rpnp
&& !error
)
403 (void) strlcpy(rpnp
->pn_buf
, name
, rpnp
->pn_bufsize
);
409 * unlinked Set (formerly known as the "delete queue") Error Handling
411 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
412 * don't specify the name of the entry that we will be manipulating. We
413 * also fib and say that we won't be adding any new entries to the
414 * unlinked set, even though we might (this is to lower the minimum file
415 * size that can be deleted in a full filesystem). So on the small
416 * chance that the nlink list is using a fat zap (ie. has more than
417 * 2000 entries), we *may* not pre-read a block that's needed.
418 * Therefore it is remotely possible for some of the assertions
419 * regarding the unlinked set below to fail due to i/o error. On a
420 * nondebug system, this will result in the space being leaked.
423 zfs_unlinked_add(znode_t
*zp
, dmu_tx_t
*tx
)
425 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
427 ASSERT(zp
->z_unlinked
);
428 ASSERT3U(zp
->z_phys
->zp_links
, ==, 0);
431 zap_add_int(zfsvfs
->z_os
, zfsvfs
->z_unlinkedobj
, zp
->z_id
, tx
));
435 * Clean up any znodes that had no links when we either crashed or
436 * (force) umounted the file system.
439 zfs_unlinked_drain(zfsvfs_t
*zfsvfs
)
443 dmu_object_info_t doi
;
448 * Interate over the contents of the unlinked set.
450 for (zap_cursor_init(&zc
, zfsvfs
->z_os
, zfsvfs
->z_unlinkedobj
);
451 zap_cursor_retrieve(&zc
, &zap
) == 0;
452 zap_cursor_advance(&zc
)) {
455 * See what kind of object we have in list
458 error
= dmu_object_info(zfsvfs
->z_os
,
459 zap
.za_first_integer
, &doi
);
463 ASSERT((doi
.doi_type
== DMU_OT_PLAIN_FILE_CONTENTS
) ||
464 (doi
.doi_type
== DMU_OT_DIRECTORY_CONTENTS
));
466 * We need to re-mark these list entries for deletion,
467 * so we pull them back into core and set zp->z_unlinked.
469 error
= zfs_zget(zfsvfs
, zap
.za_first_integer
, &zp
);
472 * We may pick up znodes that are already marked for deletion.
473 * This could happen during the purge of an extended attribute
474 * directory. All we need to do is skip over them, since they
475 * are already in the system marked z_unlinked.
480 zp
->z_unlinked
= B_TRUE
;
483 zap_cursor_fini(&zc
);
487 * Delete the entire contents of a directory. Return a count
488 * of the number of entries that could not be deleted. If we encounter
489 * an error, return a count of at least one so that the directory stays
490 * in the unlinked set.
492 * NOTE: this function assumes that the directory is inactive,
493 * so there is no need to lock its entries before deletion.
494 * Also, it assumes the directory contents is *only* regular
498 zfs_purgedir(znode_t
*dzp
)
504 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
509 for (zap_cursor_init(&zc
, zfsvfs
->z_os
, dzp
->z_id
);
510 (error
= zap_cursor_retrieve(&zc
, &zap
)) == 0;
511 zap_cursor_advance(&zc
)) {
512 error
= zfs_zget(zfsvfs
,
513 ZFS_DIRENT_OBJ(zap
.za_first_integer
), &xzp
);
519 ASSERT((ZTOV(xzp
)->v_type
== VREG
) ||
520 (ZTOV(xzp
)->v_type
== VLNK
));
522 tx
= dmu_tx_create(zfsvfs
->z_os
);
523 dmu_tx_hold_bonus(tx
, dzp
->z_id
);
524 dmu_tx_hold_zap(tx
, dzp
->z_id
, FALSE
, zap
.za_name
);
525 dmu_tx_hold_bonus(tx
, xzp
->z_id
);
526 dmu_tx_hold_zap(tx
, zfsvfs
->z_unlinkedobj
, FALSE
, NULL
);
527 error
= dmu_tx_assign(tx
, TXG_WAIT
);
534 bzero(&dl
, sizeof (dl
));
536 dl
.dl_name
= zap
.za_name
;
538 error
= zfs_link_destroy(&dl
, xzp
, tx
, 0, NULL
);
545 zap_cursor_fini(&zc
);
552 zfs_rmnode(znode_t
*zp
)
554 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
555 objset_t
*os
= zfsvfs
->z_os
;
561 ASSERT(ZTOV(zp
)->v_count
== 0);
562 ASSERT(zp
->z_phys
->zp_links
== 0);
565 * If this is an attribute directory, purge its contents.
567 if (ZTOV(zp
)->v_type
== VDIR
&& (zp
->z_phys
->zp_flags
& ZFS_XATTR
)) {
568 if (zfs_purgedir(zp
) != 0) {
570 * Not enough space to delete some xattrs.
571 * Leave it in the unlinked set.
573 zfs_znode_dmu_fini(zp
);
580 * Free up all the data in the file.
582 error
= dmu_free_long_range(os
, zp
->z_id
, 0, DMU_OBJECT_END
);
585 * Not enough space. Leave the file in the unlinked set.
587 zfs_znode_dmu_fini(zp
);
593 * If the file has extended attributes, we're going to unlink
596 if (zp
->z_phys
->zp_xattr
) {
597 error
= zfs_zget(zfsvfs
, zp
->z_phys
->zp_xattr
, &xzp
);
601 acl_obj
= zp
->z_phys
->zp_acl
.z_acl_extern_obj
;
604 * Set up the final transaction.
606 tx
= dmu_tx_create(os
);
607 dmu_tx_hold_free(tx
, zp
->z_id
, 0, DMU_OBJECT_END
);
608 dmu_tx_hold_zap(tx
, zfsvfs
->z_unlinkedobj
, FALSE
, NULL
);
610 dmu_tx_hold_bonus(tx
, xzp
->z_id
);
611 dmu_tx_hold_zap(tx
, zfsvfs
->z_unlinkedobj
, TRUE
, NULL
);
614 dmu_tx_hold_free(tx
, acl_obj
, 0, DMU_OBJECT_END
);
615 error
= dmu_tx_assign(tx
, TXG_WAIT
);
618 * Not enough space to delete the file. Leave it in the
619 * unlinked set, leaking it until the fs is remounted (at
620 * which point we'll call zfs_unlinked_drain() to process it).
623 zfs_znode_dmu_fini(zp
);
629 dmu_buf_will_dirty(xzp
->z_dbuf
, tx
);
630 mutex_enter(&xzp
->z_lock
);
631 xzp
->z_unlinked
= B_TRUE
; /* mark xzp for deletion */
632 xzp
->z_phys
->zp_links
= 0; /* no more links to it */
633 mutex_exit(&xzp
->z_lock
);
634 zfs_unlinked_add(xzp
, tx
);
637 /* Remove this znode from the unlinked set */
639 zap_remove_int(zfsvfs
->z_os
, zfsvfs
->z_unlinkedobj
, zp
->z_id
, tx
));
641 zfs_znode_delete(zp
, tx
);
650 zfs_dirent(znode_t
*zp
)
652 uint64_t de
= zp
->z_id
;
653 if (zp
->z_zfsvfs
->z_version
>= ZPL_VERSION_DIRENT_TYPE
)
654 de
|= IFTODT((zp
)->z_phys
->zp_mode
) << 60;
659 * Link zp into dl. Can only fail if zp has been unlinked.
662 zfs_link_create(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
)
664 znode_t
*dzp
= dl
->dl_dzp
;
665 vnode_t
*vp
= ZTOV(zp
);
667 int zp_is_dir
= (vp
->v_type
== VDIR
);
670 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
671 mutex_enter(&zp
->z_lock
);
673 if (!(flag
& ZRENAMING
)) {
674 if (zp
->z_unlinked
) { /* no new links to unlinked zp */
675 ASSERT(!(flag
& (ZNEW
| ZEXISTS
)));
676 mutex_exit(&zp
->z_lock
);
679 zp
->z_phys
->zp_links
++;
681 zp
->z_phys
->zp_parent
= dzp
->z_id
; /* dzp is now zp's parent */
684 zfs_time_stamper_locked(zp
, STATE_CHANGED
, tx
);
685 mutex_exit(&zp
->z_lock
);
687 dmu_buf_will_dirty(dzp
->z_dbuf
, tx
);
688 mutex_enter(&dzp
->z_lock
);
689 dzp
->z_phys
->zp_size
++; /* one dirent added */
690 dzp
->z_phys
->zp_links
+= zp_is_dir
; /* ".." link from zp */
691 zfs_time_stamper_locked(dzp
, CONTENT_MODIFIED
, tx
);
692 mutex_exit(&dzp
->z_lock
);
694 value
= zfs_dirent(zp
);
695 error
= zap_add(zp
->z_zfsvfs
->z_os
, dzp
->z_id
, dl
->dl_name
,
699 dnlc_update(ZTOV(dzp
), dl
->dl_name
, vp
);
705 * Unlink zp from dl, and mark zp for deletion if this was the last link.
706 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
707 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
708 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
709 * and it's the caller's job to do it.
712 zfs_link_destroy(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
,
713 boolean_t
*unlinkedp
)
715 znode_t
*dzp
= dl
->dl_dzp
;
716 vnode_t
*vp
= ZTOV(zp
);
717 int zp_is_dir
= (vp
->v_type
== VDIR
);
718 boolean_t unlinked
= B_FALSE
;
721 dnlc_remove(ZTOV(dzp
), dl
->dl_name
);
723 if (!(flag
& ZRENAMING
)) {
724 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
726 if (vn_vfswlock(vp
)) /* prevent new mounts on zp */
729 if (vn_ismntpt(vp
)) { /* don't remove mount point */
734 mutex_enter(&zp
->z_lock
);
735 if (zp_is_dir
&& !zfs_dirempty(zp
)) { /* dir not empty */
736 mutex_exit(&zp
->z_lock
);
740 if (zp
->z_phys
->zp_links
<= zp_is_dir
) {
741 zfs_panic_recover("zfs: link count on %s is %u, "
742 "should be at least %u",
743 zp
->z_vnode
->v_path
? zp
->z_vnode
->v_path
:
744 "<unknown>", (int)zp
->z_phys
->zp_links
,
746 zp
->z_phys
->zp_links
= zp_is_dir
+ 1;
748 if (--zp
->z_phys
->zp_links
== zp_is_dir
) {
749 zp
->z_unlinked
= B_TRUE
;
750 zp
->z_phys
->zp_links
= 0;
753 zfs_time_stamper_locked(zp
, STATE_CHANGED
, tx
);
755 mutex_exit(&zp
->z_lock
);
759 dmu_buf_will_dirty(dzp
->z_dbuf
, tx
);
760 mutex_enter(&dzp
->z_lock
);
761 dzp
->z_phys
->zp_size
--; /* one dirent removed */
762 dzp
->z_phys
->zp_links
-= zp_is_dir
; /* ".." link from zp */
763 zfs_time_stamper_locked(dzp
, CONTENT_MODIFIED
, tx
);
764 mutex_exit(&dzp
->z_lock
);
766 if (zp
->z_zfsvfs
->z_norm
) {
767 if (((zp
->z_zfsvfs
->z_case
== ZFS_CASE_INSENSITIVE
) &&
768 (flag
& ZCIEXACT
)) ||
769 ((zp
->z_zfsvfs
->z_case
== ZFS_CASE_MIXED
) &&
771 error
= zap_remove_norm(zp
->z_zfsvfs
->z_os
,
772 dzp
->z_id
, dl
->dl_name
, MT_EXACT
, tx
);
774 error
= zap_remove_norm(zp
->z_zfsvfs
->z_os
,
775 dzp
->z_id
, dl
->dl_name
, MT_FIRST
, tx
);
777 error
= zap_remove(zp
->z_zfsvfs
->z_os
,
778 dzp
->z_id
, dl
->dl_name
, tx
);
782 if (unlinkedp
!= NULL
)
783 *unlinkedp
= unlinked
;
785 zfs_unlinked_add(zp
, tx
);
791 * Indicate whether the directory is empty. Works with or without z_lock
792 * held, but can only be consider a hint in the latter case. Returns true
793 * if only "." and ".." remain and there's no work in progress.
796 zfs_dirempty(znode_t
*dzp
)
798 return (dzp
->z_phys
->zp_size
== 2 && dzp
->z_dirlocks
== 0);
802 zfs_make_xattrdir(znode_t
*zp
, vattr_t
*vap
, vnode_t
**xvpp
, cred_t
*cr
)
804 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
808 zfs_acl_ids_t acl_ids
;
809 boolean_t fuid_dirtied
;
813 if (error
= zfs_zaccess(zp
, ACE_WRITE_NAMED_ATTRS
, 0, B_FALSE
, cr
))
816 if ((error
= zfs_acl_ids_create(zp
, IS_XATTR
, vap
, cr
, NULL
,
819 if (zfs_acl_ids_overquota(zfsvfs
, &acl_ids
)) {
820 zfs_acl_ids_free(&acl_ids
);
824 tx
= dmu_tx_create(zfsvfs
->z_os
);
825 dmu_tx_hold_bonus(tx
, zp
->z_id
);
826 dmu_tx_hold_zap(tx
, DMU_NEW_OBJECT
, FALSE
, NULL
);
827 fuid_dirtied
= zfsvfs
->z_fuid_dirty
;
829 zfs_fuid_txhold(zfsvfs
, tx
);
830 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
832 zfs_acl_ids_free(&acl_ids
);
833 if (error
== ERESTART
)
838 zfs_mknode(zp
, vap
, tx
, cr
, IS_XATTR
, &xzp
, 0, &acl_ids
);
841 zfs_fuid_sync(zfsvfs
, tx
);
843 ASSERT(xzp
->z_phys
->zp_parent
== zp
->z_id
);
844 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
845 zp
->z_phys
->zp_xattr
= xzp
->z_id
;
847 (void) zfs_log_create(zfsvfs
->z_log
, tx
, TX_MKXATTR
, zp
,
848 xzp
, "", NULL
, acl_ids
.z_fuidp
, vap
);
850 zfs_acl_ids_free(&acl_ids
);
859 * Return a znode for the extended attribute directory for zp.
860 * ** If the directory does not already exist, it is created **
862 * IN: zp - znode to obtain attribute directory from
863 * cr - credentials of caller
864 * flags - flags from the VOP_LOOKUP call
866 * OUT: xzpp - pointer to extended attribute znode
868 * RETURN: 0 on success
869 * error number on failure
872 zfs_get_xattrdir(znode_t
*zp
, vnode_t
**xvpp
, cred_t
*cr
, int flags
)
874 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
880 error
= zfs_dirent_lock(&dl
, zp
, "", &xzp
, ZXATTR
, NULL
, NULL
);
886 zfs_dirent_unlock(dl
);
890 ASSERT(zp
->z_phys
->zp_xattr
== 0);
892 if (!(flags
& CREATE_XATTR_DIR
)) {
893 zfs_dirent_unlock(dl
);
897 if (zfsvfs
->z_vfs
->vfs_flag
& VFS_RDONLY
) {
898 zfs_dirent_unlock(dl
);
903 * The ability to 'create' files in an attribute
904 * directory comes from the write_xattr permission on the base file.
906 * The ability to 'search' an attribute directory requires
907 * read_xattr permission on the base file.
909 * Once in a directory the ability to read/write attributes
910 * is controlled by the permissions on the attribute file.
912 va
.va_mask
= AT_TYPE
| AT_MODE
| AT_UID
| AT_GID
;
914 va
.va_mode
= S_IFDIR
| S_ISVTX
| 0777;
915 zfs_fuid_map_ids(zp
, cr
, &va
.va_uid
, &va
.va_gid
);
917 error
= zfs_make_xattrdir(zp
, &va
, xvpp
, cr
);
918 zfs_dirent_unlock(dl
);
920 if (error
== ERESTART
) {
921 /* NB: we already did dmu_tx_wait() if necessary */
929 * Decide whether it is okay to remove within a sticky directory.
931 * In sticky directories, write access is not sufficient;
932 * you can remove entries from a directory only if:
934 * you own the directory,
936 * the entry is a plain file and you have write access,
937 * or you are privileged (checked in secpolicy...).
939 * The function returns 0 if remove access is granted.
942 zfs_sticky_remove_access(znode_t
*zdp
, znode_t
*zp
, cred_t
*cr
)
947 zfsvfs_t
*zfsvfs
= zdp
->z_zfsvfs
;
949 if (zdp
->z_zfsvfs
->z_replay
)
952 if ((zdp
->z_phys
->zp_mode
& S_ISVTX
) == 0)
955 downer
= zfs_fuid_map_id(zfsvfs
, zdp
->z_phys
->zp_uid
, cr
, ZFS_OWNER
);
956 fowner
= zfs_fuid_map_id(zfsvfs
, zp
->z_phys
->zp_uid
, cr
, ZFS_OWNER
);
958 if ((uid
= crgetuid(cr
)) == downer
|| uid
== fowner
||
959 (ZTOV(zp
)->v_type
== VREG
&&
960 zfs_zaccess(zp
, ACE_WRITE_DATA
, 0, B_FALSE
, cr
) == 0))
963 return (secpolicy_vnode_remove(cr
));