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.
23 * Copyright (c) 2013 by Delphix. 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/zfs_vnops.h>
50 #include <sys/fs/zfs.h>
51 #include "fs/fs_subr.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_lock() to peform zap lookups
64 * of names after deciding which is the appropriate lookup interface.
67 zfs_match_find(zfs_sb_t
*zsb
, znode_t
*dzp
, char *name
, boolean_t exact
,
68 boolean_t update
, int *deflags
, pathname_t
*rpnp
, uint64_t *zoid
)
70 boolean_t conflict
= B_FALSE
;
74 matchtype_t mt
= MT_FIRST
;
80 bufsz
= rpnp
->pn_bufsize
;
85 * In the non-mixed case we only expect there would ever
86 * be one match, but we need to use the normalizing lookup.
88 error
= zap_lookup_norm(zsb
->z_os
, dzp
->z_id
, name
, 8, 1,
89 zoid
, mt
, buf
, bufsz
, &conflict
);
91 error
= zap_lookup(zsb
->z_os
, dzp
->z_id
, name
, 8, 1, zoid
);
95 * Allow multiple entries provided the first entry is
96 * the object id. Non-zpl consumers may safely make
97 * use of the additional space.
99 * XXX: This should be a feature flag for compatibility
101 if (error
== EOVERFLOW
)
104 if (zsb
->z_norm
&& !error
&& deflags
)
105 *deflags
= conflict
? ED_CASE_CONFLICT
: 0;
107 *zoid
= ZFS_DIRENT_OBJ(*zoid
);
110 if (error
== ENOENT
&& update
)
111 dnlc_update(ZTOI(dzp
), name
, DNLC_NO_VNODE
);
112 #endif /* HAVE_DNLC */
118 * Lock a directory entry. A dirlock on <dzp, name> protects that name
119 * in dzp's directory zap object. As long as you hold a dirlock, you can
120 * assume two things: (1) dzp cannot be reaped, and (2) no other thread
121 * can change the zap entry for (i.e. link or unlink) this name.
124 * dzp - znode for directory
125 * name - name of entry to lock
126 * flag - ZNEW: if the entry already exists, fail with EEXIST.
127 * ZEXISTS: if the entry does not exist, fail with ENOENT.
128 * ZSHARED: allow concurrent access with other ZSHARED callers.
129 * ZXATTR: we want dzp's xattr directory
130 * ZCILOOK: On a mixed sensitivity file system,
131 * this lookup should be case-insensitive.
132 * ZCIEXACT: On a purely case-insensitive file system,
133 * this lookup should be case-sensitive.
134 * ZRENAMING: we are locking for renaming, force narrow locks
135 * ZHAVELOCK: Don't grab the z_name_lock for this call. The
136 * current thread already holds it.
139 * zpp - pointer to the znode for the entry (NULL if there isn't one)
140 * dlpp - pointer to the dirlock for this entry (NULL on error)
141 * direntflags - (case-insensitive lookup only)
142 * flags if multiple case-sensitive matches exist in directory
143 * realpnp - (case-insensitive lookup only)
144 * actual name matched within the directory
146 * Return value: 0 on success or errno on failure.
148 * NOTE: Always checks for, and rejects, '.' and '..'.
149 * NOTE: For case-insensitive file systems we take wide locks (see below),
150 * but return znode pointers to a single match.
153 zfs_dirent_lock(zfs_dirlock_t
**dlpp
, znode_t
*dzp
, char *name
, znode_t
**zpp
,
154 int flag
, int *direntflags
, pathname_t
*realpnp
)
156 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
163 #endif /* HAVE_DNLC */
171 * Verify that we are not trying to lock '.', '..', or '.zfs'
173 if ((name
[0] == '.' &&
174 (name
[1] == '\0' || (name
[1] == '.' && name
[2] == '\0'))) ||
175 (zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0))
176 return (SET_ERROR(EEXIST
));
179 * Case sensitivity and normalization preferences are set when
180 * the file system is created. These are stored in the
181 * zsb->z_case and zsb->z_norm fields. These choices
182 * affect what vnodes can be cached in the DNLC, how we
183 * perform zap lookups, and the "width" of our dirlocks.
185 * A normal dirlock locks a single name. Note that with
186 * normalization a name can be composed multiple ways, but
187 * when normalized, these names all compare equal. A wide
188 * dirlock locks multiple names. We need these when the file
189 * system is supporting mixed-mode access. It is sometimes
190 * necessary to lock all case permutations of file name at
191 * once so that simultaneous case-insensitive/case-sensitive
192 * behaves as rationally as possible.
196 * Decide if exact matches should be requested when performing
197 * a zap lookup on file systems supporting case-insensitive
201 ((zsb
->z_case
== ZFS_CASE_INSENSITIVE
) && (flag
& ZCIEXACT
)) ||
202 ((zsb
->z_case
== ZFS_CASE_MIXED
) && !(flag
& ZCILOOK
));
205 * Only look in or update the DNLC if we are looking for the
206 * name on a file system that does not require normalization
207 * or case folding. We can also look there if we happen to be
208 * on a non-normalizing, mixed sensitivity file system IF we
209 * are looking for the exact name.
211 * Maybe can add TO-UPPERed version of name to dnlc in ci-only
212 * case for performance improvement?
214 update
= !zsb
->z_norm
||
215 ((zsb
->z_case
== ZFS_CASE_MIXED
) &&
216 !(zsb
->z_norm
& ~U8_TEXTPREP_TOUPPER
) && !(flag
& ZCILOOK
));
219 * ZRENAMING indicates we are in a situation where we should
220 * take narrow locks regardless of the file system's
221 * preferences for normalizing and case folding. This will
222 * prevent us deadlocking trying to grab the same wide lock
223 * twice if the two names happen to be case-insensitive
226 if (flag
& ZRENAMING
)
229 cmpflags
= zsb
->z_norm
;
232 * Wait until there are no locks on this name.
234 * Don't grab the the lock if it is already held. However, cannot
235 * have both ZSHARED and ZHAVELOCK together.
237 ASSERT(!(flag
& ZSHARED
) || !(flag
& ZHAVELOCK
));
238 if (!(flag
& ZHAVELOCK
))
239 rw_enter(&dzp
->z_name_lock
, RW_READER
);
241 mutex_enter(&dzp
->z_lock
);
243 if (dzp
->z_unlinked
) {
244 mutex_exit(&dzp
->z_lock
);
245 if (!(flag
& ZHAVELOCK
))
246 rw_exit(&dzp
->z_name_lock
);
247 return (SET_ERROR(ENOENT
));
249 for (dl
= dzp
->z_dirlocks
; dl
!= NULL
; dl
= dl
->dl_next
) {
250 if ((u8_strcmp(name
, dl
->dl_name
, 0, cmpflags
,
251 U8_UNICODE_LATEST
, &error
) == 0) || error
!= 0)
255 mutex_exit(&dzp
->z_lock
);
256 if (!(flag
& ZHAVELOCK
))
257 rw_exit(&dzp
->z_name_lock
);
258 return (SET_ERROR(ENOENT
));
262 * Allocate a new dirlock and add it to the list.
264 dl
= kmem_alloc(sizeof (zfs_dirlock_t
), KM_SLEEP
);
265 cv_init(&dl
->dl_cv
, NULL
, CV_DEFAULT
, NULL
);
271 dl
->dl_next
= dzp
->z_dirlocks
;
272 dzp
->z_dirlocks
= dl
;
275 if ((flag
& ZSHARED
) && dl
->dl_sharecnt
!= 0)
277 cv_wait(&dl
->dl_cv
, &dzp
->z_lock
);
281 * If the z_name_lock was NOT held for this dirlock record it.
283 if (flag
& ZHAVELOCK
)
286 if ((flag
& ZSHARED
) && ++dl
->dl_sharecnt
> 1 && dl
->dl_namesize
== 0) {
288 * We're the second shared reference to dl. Make a copy of
289 * dl_name in case the first thread goes away before we do.
290 * Note that we initialize the new name before storing its
291 * pointer into dl_name, because the first thread may load
292 * dl->dl_name at any time. He'll either see the old value,
293 * which is his, or the new shared copy; either is OK.
295 dl
->dl_namesize
= strlen(dl
->dl_name
) + 1;
296 name
= kmem_alloc(dl
->dl_namesize
, KM_SLEEP
);
297 bcopy(dl
->dl_name
, name
, dl
->dl_namesize
);
301 mutex_exit(&dzp
->z_lock
);
304 * We have a dirlock on the name. (Note that it is the dirlock,
305 * not the dzp's z_lock, that protects the name in the zap object.)
306 * See if there's an object by this name; if so, put a hold on it.
309 error
= sa_lookup(dzp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
), &zoid
,
312 error
= (zoid
== 0 ? SET_ERROR(ENOENT
) : 0);
316 vp
= dnlc_lookup(ZTOI(dzp
), name
);
317 if (vp
== DNLC_NO_VNODE
) {
319 error
= SET_ERROR(ENOENT
);
322 zfs_dirent_unlock(dl
);
324 return (SET_ERROR(EEXIST
));
330 error
= zfs_match_find(zsb
, dzp
, name
, exact
,
331 update
, direntflags
, realpnp
, &zoid
);
334 error
= zfs_match_find(zsb
, dzp
, name
, exact
,
335 update
, direntflags
, realpnp
, &zoid
);
336 #endif /* HAVE_DNLC */
339 if (error
!= ENOENT
|| (flag
& ZEXISTS
)) {
340 zfs_dirent_unlock(dl
);
345 zfs_dirent_unlock(dl
);
346 return (SET_ERROR(EEXIST
));
348 error
= zfs_zget(zsb
, zoid
, zpp
);
350 zfs_dirent_unlock(dl
);
354 if (!(flag
& ZXATTR
) && update
)
355 dnlc_update(ZTOI(dzp
), name
, ZTOI(*zpp
));
356 #endif /* HAVE_DNLC */
365 * Unlock this directory entry and wake anyone who was waiting for it.
368 zfs_dirent_unlock(zfs_dirlock_t
*dl
)
370 znode_t
*dzp
= dl
->dl_dzp
;
371 zfs_dirlock_t
**prev_dl
, *cur_dl
;
373 mutex_enter(&dzp
->z_lock
);
375 if (!dl
->dl_namelock
)
376 rw_exit(&dzp
->z_name_lock
);
378 if (dl
->dl_sharecnt
> 1) {
380 mutex_exit(&dzp
->z_lock
);
383 prev_dl
= &dzp
->z_dirlocks
;
384 while ((cur_dl
= *prev_dl
) != dl
)
385 prev_dl
= &cur_dl
->dl_next
;
386 *prev_dl
= dl
->dl_next
;
387 cv_broadcast(&dl
->dl_cv
);
388 mutex_exit(&dzp
->z_lock
);
390 if (dl
->dl_namesize
!= 0)
391 kmem_free(dl
->dl_name
, dl
->dl_namesize
);
392 cv_destroy(&dl
->dl_cv
);
393 kmem_free(dl
, sizeof (*dl
));
397 * Look up an entry in a directory.
399 * NOTE: '.' and '..' are handled as special cases because
400 * no directory entries are actually stored for them. If this is
401 * the root of a filesystem, then '.zfs' is also treated as a
402 * special pseudo-directory.
405 zfs_dirlook(znode_t
*dzp
, char *name
, struct inode
**ipp
, int flags
,
406 int *deflg
, pathname_t
*rpnp
)
413 if (name
[0] == 0 || (name
[0] == '.' && name
[1] == 0)) {
416 } else if (name
[0] == '.' && name
[1] == '.' && name
[2] == 0) {
417 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
420 * If we are a snapshot mounted under .zfs, return
421 * the inode pointer for the snapshot directory.
423 if ((error
= sa_lookup(dzp
->z_sa_hdl
,
424 SA_ZPL_PARENT(zsb
), &parent
, sizeof (parent
))) != 0)
427 if (parent
== dzp
->z_id
&& zsb
->z_parent
!= zsb
) {
428 error
= zfsctl_root_lookup(zsb
->z_parent
->z_ctldir
,
429 "snapshot", ipp
, 0, kcred
, NULL
, NULL
);
432 rw_enter(&dzp
->z_parent_lock
, RW_READER
);
433 error
= zfs_zget(zsb
, parent
, &zp
);
436 rw_exit(&dzp
->z_parent_lock
);
437 } else if (zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0) {
438 *ipp
= zfsctl_root(dzp
);
442 zf
= ZEXISTS
| ZSHARED
;
443 if (flags
& FIGNORECASE
)
446 error
= zfs_dirent_lock(&dl
, dzp
, name
, &zp
, zf
, deflg
, rpnp
);
449 zfs_dirent_unlock(dl
);
450 dzp
->z_zn_prefetch
= B_TRUE
; /* enable prefetching */
455 if ((flags
& FIGNORECASE
) && rpnp
&& !error
)
456 (void) strlcpy(rpnp
->pn_buf
, name
, rpnp
->pn_bufsize
);
462 * unlinked Set (formerly known as the "delete queue") Error Handling
464 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
465 * don't specify the name of the entry that we will be manipulating. We
466 * also fib and say that we won't be adding any new entries to the
467 * unlinked set, even though we might (this is to lower the minimum file
468 * size that can be deleted in a full filesystem). So on the small
469 * chance that the nlink list is using a fat zap (ie. has more than
470 * 2000 entries), we *may* not pre-read a block that's needed.
471 * Therefore it is remotely possible for some of the assertions
472 * regarding the unlinked set below to fail due to i/o error. On a
473 * nondebug system, this will result in the space being leaked.
476 zfs_unlinked_add(znode_t
*zp
, dmu_tx_t
*tx
)
478 zfs_sb_t
*zsb
= ZTOZSB(zp
);
480 ASSERT(zp
->z_unlinked
);
481 ASSERT(zp
->z_links
== 0);
484 zap_add_int(zsb
->z_os
, zsb
->z_unlinkedobj
, zp
->z_id
, tx
));
488 * Clean up any znodes that had no links when we either crashed or
489 * (force) umounted the file system.
492 zfs_unlinked_drain(zfs_sb_t
*zsb
)
496 dmu_object_info_t doi
;
501 * Iterate over the contents of the unlinked set.
503 for (zap_cursor_init(&zc
, zsb
->z_os
, zsb
->z_unlinkedobj
);
504 zap_cursor_retrieve(&zc
, &zap
) == 0;
505 zap_cursor_advance(&zc
)) {
508 * See what kind of object we have in list
511 error
= dmu_object_info(zsb
->z_os
, zap
.za_first_integer
, &doi
);
515 ASSERT((doi
.doi_type
== DMU_OT_PLAIN_FILE_CONTENTS
) ||
516 (doi
.doi_type
== DMU_OT_DIRECTORY_CONTENTS
));
518 * We need to re-mark these list entries for deletion,
519 * so we pull them back into core and set zp->z_unlinked.
521 error
= zfs_zget(zsb
, zap
.za_first_integer
, &zp
);
524 * We may pick up znodes that are already marked for deletion.
525 * This could happen during the purge of an extended attribute
526 * directory. All we need to do is skip over them, since they
527 * are already in the system marked z_unlinked.
532 zp
->z_unlinked
= B_TRUE
;
535 zap_cursor_fini(&zc
);
539 * Delete the entire contents of a directory. Return a count
540 * of the number of entries that could not be deleted. If we encounter
541 * an error, return a count of at least one so that the directory stays
542 * in the unlinked set.
544 * NOTE: this function assumes that the directory is inactive,
545 * so there is no need to lock its entries before deletion.
546 * Also, it assumes the directory contents is *only* regular
550 zfs_purgedir(znode_t
*dzp
)
556 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
561 for (zap_cursor_init(&zc
, zsb
->z_os
, dzp
->z_id
);
562 (error
= zap_cursor_retrieve(&zc
, &zap
)) == 0;
563 zap_cursor_advance(&zc
)) {
564 error
= zfs_zget(zsb
,
565 ZFS_DIRENT_OBJ(zap
.za_first_integer
), &xzp
);
571 ASSERT(S_ISREG(ZTOI(xzp
)->i_mode
) ||
572 S_ISLNK(ZTOI(xzp
)->i_mode
));
574 tx
= dmu_tx_create(zsb
->z_os
);
575 dmu_tx_hold_sa(tx
, dzp
->z_sa_hdl
, B_FALSE
);
576 dmu_tx_hold_zap(tx
, dzp
->z_id
, FALSE
, zap
.za_name
);
577 dmu_tx_hold_sa(tx
, xzp
->z_sa_hdl
, B_FALSE
);
578 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, FALSE
, NULL
);
579 /* Is this really needed ? */
580 zfs_sa_upgrade_txholds(tx
, xzp
);
581 error
= dmu_tx_assign(tx
, TXG_WAIT
);
584 zfs_iput_async(ZTOI(xzp
));
588 bzero(&dl
, sizeof (dl
));
590 dl
.dl_name
= zap
.za_name
;
592 error
= zfs_link_destroy(&dl
, xzp
, tx
, 0, NULL
);
596 set_nlink(ZTOI(xzp
), xzp
->z_links
);
597 zfs_iput_async(ZTOI(xzp
));
599 zap_cursor_fini(&zc
);
606 zfs_rmnode(znode_t
*zp
)
608 zfs_sb_t
*zsb
= ZTOZSB(zp
);
609 objset_t
*os
= zsb
->z_os
;
616 ASSERT(zp
->z_links
== 0);
617 ASSERT(atomic_read(&ZTOI(zp
)->i_count
) == 0);
620 * If this is an attribute directory, purge its contents.
622 if (S_ISDIR(ZTOI(zp
)->i_mode
) && (zp
->z_pflags
& ZFS_XATTR
)) {
623 if (zfs_purgedir(zp
) != 0) {
625 * Not enough space to delete some xattrs.
626 * Leave it in the unlinked set.
628 zfs_znode_dmu_fini(zp
);
635 * Free up all the data in the file. We don't do this for directories
636 * because we need truncate and remove to be in the same tx, like in
637 * zfs_znode_delete(). Otherwise, if we crash here we'll end up with
638 * an inconsistent truncated zap object in the delete queue. Note a
639 * truncated file is harmless since it only contains user data.
641 if (S_ISREG(ZTOI(zp
)->i_mode
)) {
642 error
= dmu_free_long_range(os
, zp
->z_id
, 0, DMU_OBJECT_END
);
645 * Not enough space. Leave the file in the unlinked
648 zfs_znode_dmu_fini(zp
);
654 * If the file has extended attributes, we're going to unlink
657 error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
),
658 &xattr_obj
, sizeof (xattr_obj
));
659 if (error
== 0 && xattr_obj
) {
660 error
= zfs_zget(zsb
, xattr_obj
, &xzp
);
664 acl_obj
= zfs_external_acl(zp
);
667 * Set up the final transaction.
669 tx
= dmu_tx_create(os
);
670 dmu_tx_hold_free(tx
, zp
->z_id
, 0, DMU_OBJECT_END
);
671 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, FALSE
, NULL
);
673 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, TRUE
, NULL
);
674 dmu_tx_hold_sa(tx
, xzp
->z_sa_hdl
, B_FALSE
);
677 dmu_tx_hold_free(tx
, acl_obj
, 0, DMU_OBJECT_END
);
679 zfs_sa_upgrade_txholds(tx
, zp
);
680 error
= dmu_tx_assign(tx
, TXG_WAIT
);
683 * Not enough space to delete the file. Leave it in the
684 * unlinked set, leaking it until the fs is remounted (at
685 * which point we'll call zfs_unlinked_drain() to process it).
688 zfs_znode_dmu_fini(zp
);
694 mutex_enter(&xzp
->z_lock
);
695 xzp
->z_unlinked
= B_TRUE
; /* mark xzp for deletion */
696 xzp
->z_links
= 0; /* no more links to it */
697 set_nlink(ZTOI(xzp
), 0); /* this will let iput purge us */
698 VERIFY(0 == sa_update(xzp
->z_sa_hdl
, SA_ZPL_LINKS(zsb
),
699 &xzp
->z_links
, sizeof (xzp
->z_links
), tx
));
700 mutex_exit(&xzp
->z_lock
);
701 zfs_unlinked_add(xzp
, tx
);
704 /* Remove this znode from the unlinked set */
706 zap_remove_int(zsb
->z_os
, zsb
->z_unlinkedobj
, zp
->z_id
, tx
));
708 zfs_znode_delete(zp
, tx
);
713 zfs_iput_async(ZTOI(xzp
));
717 zfs_dirent(znode_t
*zp
, uint64_t mode
)
719 uint64_t de
= zp
->z_id
;
721 if (ZTOZSB(zp
)->z_version
>= ZPL_VERSION_DIRENT_TYPE
)
722 de
|= IFTODT(mode
) << 60;
727 * Link zp into dl. Can only fail if zp has been unlinked.
730 zfs_link_create(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
)
732 znode_t
*dzp
= dl
->dl_dzp
;
733 zfs_sb_t
*zsb
= ZTOZSB(zp
);
735 int zp_is_dir
= S_ISDIR(ZTOI(zp
)->i_mode
);
736 sa_bulk_attr_t bulk
[5];
737 uint64_t mtime
[2], ctime
[2];
741 mutex_enter(&zp
->z_lock
);
743 if (!(flag
& ZRENAMING
)) {
744 if (zp
->z_unlinked
) { /* no new links to unlinked zp */
745 ASSERT(!(flag
& (ZNEW
| ZEXISTS
)));
746 mutex_exit(&zp
->z_lock
);
747 return (SET_ERROR(ENOENT
));
750 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
751 &zp
->z_links
, sizeof (zp
->z_links
));
754 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zsb
), NULL
,
755 &dzp
->z_id
, sizeof (dzp
->z_id
));
756 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
757 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
759 if (!(flag
& ZNEW
)) {
760 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
761 ctime
, sizeof (ctime
));
762 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, mtime
,
765 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
768 mutex_exit(&zp
->z_lock
);
770 mutex_enter(&dzp
->z_lock
);
772 dzp
->z_links
+= zp_is_dir
;
774 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
,
775 &dzp
->z_size
, sizeof (dzp
->z_size
));
776 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
777 &dzp
->z_links
, sizeof (dzp
->z_links
));
778 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
), NULL
,
779 mtime
, sizeof (mtime
));
780 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
781 ctime
, sizeof (ctime
));
782 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
783 &dzp
->z_pflags
, sizeof (dzp
->z_pflags
));
784 zfs_tstamp_update_setup(dzp
, CONTENT_MODIFIED
, mtime
, ctime
);
785 error
= sa_bulk_update(dzp
->z_sa_hdl
, bulk
, count
, tx
);
787 mutex_exit(&dzp
->z_lock
);
789 value
= zfs_dirent(zp
, zp
->z_mode
);
790 error
= zap_add(ZTOZSB(zp
)->z_os
, dzp
->z_id
, dl
->dl_name
,
798 zfs_dropname(zfs_dirlock_t
*dl
, znode_t
*zp
, znode_t
*dzp
, dmu_tx_t
*tx
,
803 if (ZTOZSB(zp
)->z_norm
) {
804 if (((ZTOZSB(zp
)->z_case
== ZFS_CASE_INSENSITIVE
) &&
805 (flag
& ZCIEXACT
)) ||
806 ((ZTOZSB(zp
)->z_case
== ZFS_CASE_MIXED
) &&
808 error
= zap_remove_norm(ZTOZSB(zp
)->z_os
,
809 dzp
->z_id
, dl
->dl_name
, MT_EXACT
, tx
);
811 error
= zap_remove_norm(ZTOZSB(zp
)->z_os
,
812 dzp
->z_id
, dl
->dl_name
, MT_FIRST
, tx
);
814 error
= zap_remove(ZTOZSB(zp
)->z_os
,
815 dzp
->z_id
, dl
->dl_name
, tx
);
822 * Unlink zp from dl, and mark zp for deletion if this was the last link. Can
823 * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY).
824 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
825 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
826 * and it's the caller's job to do it.
829 zfs_link_destroy(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
,
830 boolean_t
*unlinkedp
)
832 znode_t
*dzp
= dl
->dl_dzp
;
833 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
834 int zp_is_dir
= S_ISDIR(ZTOI(zp
)->i_mode
);
835 boolean_t unlinked
= B_FALSE
;
836 sa_bulk_attr_t bulk
[5];
837 uint64_t mtime
[2], ctime
[2];
842 dnlc_remove(ZTOI(dzp
), dl
->dl_name
);
843 #endif /* HAVE_DNLC */
845 if (!(flag
& ZRENAMING
)) {
846 mutex_enter(&zp
->z_lock
);
848 if (zp_is_dir
&& !zfs_dirempty(zp
)) {
849 mutex_exit(&zp
->z_lock
);
850 return (SET_ERROR(ENOTEMPTY
));
854 * If we get here, we are going to try to remove the object.
855 * First try removing the name from the directory; if that
856 * fails, return the error.
858 error
= zfs_dropname(dl
, zp
, dzp
, tx
, flag
);
860 mutex_exit(&zp
->z_lock
);
864 if (zp
->z_links
<= zp_is_dir
) {
865 zfs_panic_recover("zfs: link count on %lu is %u, "
866 "should be at least %u", zp
->z_id
,
867 (int)zp
->z_links
, zp_is_dir
+ 1);
868 zp
->z_links
= zp_is_dir
+ 1;
870 if (--zp
->z_links
== zp_is_dir
) {
871 zp
->z_unlinked
= B_TRUE
;
875 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
),
876 NULL
, &ctime
, sizeof (ctime
));
877 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
878 NULL
, &zp
->z_pflags
, sizeof (zp
->z_pflags
));
879 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, mtime
,
882 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
),
883 NULL
, &zp
->z_links
, sizeof (zp
->z_links
));
884 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
887 mutex_exit(&zp
->z_lock
);
889 error
= zfs_dropname(dl
, zp
, dzp
, tx
, flag
);
894 mutex_enter(&dzp
->z_lock
);
895 dzp
->z_size
--; /* one dirent removed */
896 dzp
->z_links
-= zp_is_dir
; /* ".." link from zp */
897 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
),
898 NULL
, &dzp
->z_links
, sizeof (dzp
->z_links
));
899 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
),
900 NULL
, &dzp
->z_size
, sizeof (dzp
->z_size
));
901 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
),
902 NULL
, ctime
, sizeof (ctime
));
903 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
),
904 NULL
, mtime
, sizeof (mtime
));
905 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
906 NULL
, &dzp
->z_pflags
, sizeof (dzp
->z_pflags
));
907 zfs_tstamp_update_setup(dzp
, CONTENT_MODIFIED
, mtime
, ctime
);
908 error
= sa_bulk_update(dzp
->z_sa_hdl
, bulk
, count
, tx
);
910 mutex_exit(&dzp
->z_lock
);
912 if (unlinkedp
!= NULL
)
913 *unlinkedp
= unlinked
;
915 zfs_unlinked_add(zp
, tx
);
921 * Indicate whether the directory is empty. Works with or without z_lock
922 * held, but can only be consider a hint in the latter case. Returns true
923 * if only "." and ".." remain and there's no work in progress.
926 zfs_dirempty(znode_t
*dzp
)
928 return (dzp
->z_size
== 2 && dzp
->z_dirlocks
== 0);
932 zfs_make_xattrdir(znode_t
*zp
, vattr_t
*vap
, struct inode
**xipp
, cred_t
*cr
)
934 zfs_sb_t
*zsb
= ZTOZSB(zp
);
938 zfs_acl_ids_t acl_ids
;
939 boolean_t fuid_dirtied
;
946 if ((error
= zfs_zaccess(zp
, ACE_WRITE_NAMED_ATTRS
, 0, B_FALSE
, cr
)))
949 if ((error
= zfs_acl_ids_create(zp
, IS_XATTR
, vap
, cr
, NULL
,
952 if (zfs_acl_ids_overquota(zsb
, &acl_ids
)) {
953 zfs_acl_ids_free(&acl_ids
);
954 return (SET_ERROR(EDQUOT
));
957 tx
= dmu_tx_create(zsb
->z_os
);
958 dmu_tx_hold_sa_create(tx
, acl_ids
.z_aclp
->z_acl_bytes
+
959 ZFS_SA_BASE_ATTR_SIZE
);
960 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_TRUE
);
961 dmu_tx_hold_zap(tx
, DMU_NEW_OBJECT
, FALSE
, NULL
);
962 fuid_dirtied
= zsb
->z_fuid_dirty
;
964 zfs_fuid_txhold(zsb
, tx
);
965 error
= dmu_tx_assign(tx
, TXG_WAIT
);
967 zfs_acl_ids_free(&acl_ids
);
971 zfs_mknode(zp
, vap
, tx
, cr
, IS_XATTR
, &xzp
, &acl_ids
);
974 zfs_fuid_sync(zsb
, tx
);
977 error
= sa_lookup(xzp
->z_sa_hdl
, SA_ZPL_PARENT(zsb
),
978 &parent
, sizeof (parent
));
979 ASSERT(error
== 0 && parent
== zp
->z_id
);
982 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
), &xzp
->z_id
,
983 sizeof (xzp
->z_id
), tx
));
985 (void) zfs_log_create(zsb
->z_log
, tx
, TX_MKXATTR
, zp
,
986 xzp
, "", NULL
, acl_ids
.z_fuidp
, vap
);
988 zfs_acl_ids_free(&acl_ids
);
997 * Return a znode for the extended attribute directory for zp.
998 * ** If the directory does not already exist, it is created **
1000 * IN: zp - znode to obtain attribute directory from
1001 * cr - credentials of caller
1002 * flags - flags from the VOP_LOOKUP call
1004 * OUT: xipp - pointer to extended attribute znode
1006 * RETURN: 0 on success
1007 * error number on failure
1010 zfs_get_xattrdir(znode_t
*zp
, struct inode
**xipp
, cred_t
*cr
, int flags
)
1012 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1018 error
= zfs_dirent_lock(&dl
, zp
, "", &xzp
, ZXATTR
, NULL
, NULL
);
1024 zfs_dirent_unlock(dl
);
1028 if (!(flags
& CREATE_XATTR_DIR
)) {
1029 zfs_dirent_unlock(dl
);
1030 return (SET_ERROR(ENOENT
));
1033 if (zfs_is_readonly(zsb
)) {
1034 zfs_dirent_unlock(dl
);
1035 return (SET_ERROR(EROFS
));
1039 * The ability to 'create' files in an attribute
1040 * directory comes from the write_xattr permission on the base file.
1042 * The ability to 'search' an attribute directory requires
1043 * read_xattr permission on the base file.
1045 * Once in a directory the ability to read/write attributes
1046 * is controlled by the permissions on the attribute file.
1048 va
.va_mask
= ATTR_MODE
| ATTR_UID
| ATTR_GID
;
1049 va
.va_mode
= S_IFDIR
| S_ISVTX
| 0777;
1050 zfs_fuid_map_ids(zp
, cr
, &va
.va_uid
, &va
.va_gid
);
1052 va
.va_dentry
= NULL
;
1053 error
= zfs_make_xattrdir(zp
, &va
, xipp
, cr
);
1054 zfs_dirent_unlock(dl
);
1056 if (error
== ERESTART
) {
1057 /* NB: we already did dmu_tx_wait() if necessary */
1065 * Decide whether it is okay to remove within a sticky directory.
1067 * In sticky directories, write access is not sufficient;
1068 * you can remove entries from a directory only if:
1070 * you own the directory,
1071 * you own the entry,
1072 * the entry is a plain file and you have write access,
1073 * or you are privileged (checked in secpolicy...).
1075 * The function returns 0 if remove access is granted.
1078 zfs_sticky_remove_access(znode_t
*zdp
, znode_t
*zp
, cred_t
*cr
)
1083 zfs_sb_t
*zsb
= ZTOZSB(zdp
);
1088 if ((zdp
->z_mode
& S_ISVTX
) == 0)
1091 downer
= zfs_fuid_map_id(zsb
, zdp
->z_uid
, cr
, ZFS_OWNER
);
1092 fowner
= zfs_fuid_map_id(zsb
, zp
->z_uid
, cr
, ZFS_OWNER
);
1094 if ((uid
= crgetuid(cr
)) == downer
|| uid
== fowner
||
1095 (S_ISDIR(ZTOI(zp
)->i_mode
) &&
1096 zfs_zaccess(zp
, ACE_WRITE_DATA
, 0, B_FALSE
, cr
) == 0))
1099 return (secpolicy_vnode_remove(cr
));