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.
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/zfs_sa.h>
58 #include <sys/extdirent.h>
61 * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups
62 * of names after deciding which is the appropriate lookup interface.
65 zfs_match_find(zfs_sb_t
*zsb
, znode_t
*dzp
, char *name
, boolean_t exact
,
66 boolean_t update
, int *deflags
, pathname_t
*rpnp
, uint64_t *zoid
)
68 boolean_t conflict
= B_FALSE
;
72 matchtype_t mt
= MT_FIRST
;
78 bufsz
= rpnp
->pn_bufsize
;
83 * In the non-mixed case we only expect there would ever
84 * be one match, but we need to use the normalizing lookup.
86 error
= zap_lookup_norm(zsb
->z_os
, dzp
->z_id
, name
, 8, 1,
87 zoid
, mt
, buf
, bufsz
, &conflict
);
89 error
= zap_lookup(zsb
->z_os
, dzp
->z_id
, name
, 8, 1, zoid
);
93 * Allow multiple entries provided the first entry is
94 * the object id. Non-zpl consumers may safely make
95 * use of the additional space.
97 * XXX: This should be a feature flag for compatibility
99 if (error
== EOVERFLOW
)
102 if (zsb
->z_norm
&& !error
&& deflags
)
103 *deflags
= conflict
? ED_CASE_CONFLICT
: 0;
105 *zoid
= ZFS_DIRENT_OBJ(*zoid
);
108 if (error
== ENOENT
&& update
)
109 dnlc_update(ZTOI(dzp
), name
, DNLC_NO_VNODE
);
110 #endif /* HAVE_DNLC */
116 * Lock a directory entry. A dirlock on <dzp, name> protects that name
117 * in dzp's directory zap object. As long as you hold a dirlock, you can
118 * assume two things: (1) dzp cannot be reaped, and (2) no other thread
119 * can change the zap entry for (i.e. link or unlink) this name.
122 * dzp - znode for directory
123 * name - name of entry to lock
124 * flag - ZNEW: if the entry already exists, fail with EEXIST.
125 * ZEXISTS: if the entry does not exist, fail with ENOENT.
126 * ZSHARED: allow concurrent access with other ZSHARED callers.
127 * ZXATTR: we want dzp's xattr directory
128 * ZCILOOK: On a mixed sensitivity file system,
129 * this lookup should be case-insensitive.
130 * ZCIEXACT: On a purely case-insensitive file system,
131 * this lookup should be case-sensitive.
132 * ZRENAMING: we are locking for renaming, force narrow locks
133 * ZHAVELOCK: Don't grab the z_name_lock for this call. The
134 * current thread already holds it.
137 * zpp - pointer to the znode for the entry (NULL if there isn't one)
138 * dlpp - pointer to the dirlock for this entry (NULL on error)
139 * direntflags - (case-insensitive lookup only)
140 * flags if multiple case-sensitive matches exist in directory
141 * realpnp - (case-insensitive lookup only)
142 * actual name matched within the directory
144 * Return value: 0 on success or errno on failure.
146 * NOTE: Always checks for, and rejects, '.' and '..'.
147 * NOTE: For case-insensitive file systems we take wide locks (see below),
148 * but return znode pointers to a single match.
151 zfs_dirent_lock(zfs_dirlock_t
**dlpp
, znode_t
*dzp
, char *name
, znode_t
**zpp
,
152 int flag
, int *direntflags
, pathname_t
*realpnp
)
154 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
161 #endif /* HAVE_DNLC */
169 * Verify that we are not trying to lock '.', '..', or '.zfs'
171 if ((name
[0] == '.' &&
172 (name
[1] == '\0' || (name
[1] == '.' && name
[2] == '\0'))) ||
173 (zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0))
177 * Case sensitivity and normalization preferences are set when
178 * the file system is created. These are stored in the
179 * zsb->z_case and zsb->z_norm fields. These choices
180 * affect what vnodes can be cached in the DNLC, how we
181 * perform zap lookups, and the "width" of our dirlocks.
183 * A normal dirlock locks a single name. Note that with
184 * normalization a name can be composed multiple ways, but
185 * when normalized, these names all compare equal. A wide
186 * dirlock locks multiple names. We need these when the file
187 * system is supporting mixed-mode access. It is sometimes
188 * necessary to lock all case permutations of file name at
189 * once so that simultaneous case-insensitive/case-sensitive
190 * behaves as rationally as possible.
194 * Decide if exact matches should be requested when performing
195 * a zap lookup on file systems supporting case-insensitive
199 ((zsb
->z_case
== ZFS_CASE_INSENSITIVE
) && (flag
& ZCIEXACT
)) ||
200 ((zsb
->z_case
== ZFS_CASE_MIXED
) && !(flag
& ZCILOOK
));
203 * Only look in or update the DNLC if we are looking for the
204 * name on a file system that does not require normalization
205 * or case folding. We can also look there if we happen to be
206 * on a non-normalizing, mixed sensitivity file system IF we
207 * are looking for the exact name.
209 * Maybe can add TO-UPPERed version of name to dnlc in ci-only
210 * case for performance improvement?
212 update
= !zsb
->z_norm
||
213 ((zsb
->z_case
== ZFS_CASE_MIXED
) &&
214 !(zsb
->z_norm
& ~U8_TEXTPREP_TOUPPER
) && !(flag
& ZCILOOK
));
217 * ZRENAMING indicates we are in a situation where we should
218 * take narrow locks regardless of the file system's
219 * preferences for normalizing and case folding. This will
220 * prevent us deadlocking trying to grab the same wide lock
221 * twice if the two names happen to be case-insensitive
224 if (flag
& ZRENAMING
)
227 cmpflags
= zsb
->z_norm
;
230 * Wait until there are no locks on this name.
232 * Don't grab the the lock if it is already held. However, cannot
233 * have both ZSHARED and ZHAVELOCK together.
235 ASSERT(!(flag
& ZSHARED
) || !(flag
& ZHAVELOCK
));
236 if (!(flag
& ZHAVELOCK
))
237 rw_enter(&dzp
->z_name_lock
, RW_READER
);
239 mutex_enter(&dzp
->z_lock
);
241 if (dzp
->z_unlinked
) {
242 mutex_exit(&dzp
->z_lock
);
243 if (!(flag
& ZHAVELOCK
))
244 rw_exit(&dzp
->z_name_lock
);
247 for (dl
= dzp
->z_dirlocks
; dl
!= NULL
; dl
= dl
->dl_next
) {
248 if ((u8_strcmp(name
, dl
->dl_name
, 0, cmpflags
,
249 U8_UNICODE_LATEST
, &error
) == 0) || error
!= 0)
253 mutex_exit(&dzp
->z_lock
);
254 if (!(flag
& ZHAVELOCK
))
255 rw_exit(&dzp
->z_name_lock
);
260 * Allocate a new dirlock and add it to the list.
262 dl
= kmem_alloc(sizeof (zfs_dirlock_t
), KM_SLEEP
);
263 cv_init(&dl
->dl_cv
, NULL
, CV_DEFAULT
, NULL
);
269 dl
->dl_next
= dzp
->z_dirlocks
;
270 dzp
->z_dirlocks
= dl
;
273 if ((flag
& ZSHARED
) && dl
->dl_sharecnt
!= 0)
275 cv_wait(&dl
->dl_cv
, &dzp
->z_lock
);
279 * If the z_name_lock was NOT held for this dirlock record it.
281 if (flag
& ZHAVELOCK
)
284 if ((flag
& ZSHARED
) && ++dl
->dl_sharecnt
> 1 && dl
->dl_namesize
== 0) {
286 * We're the second shared reference to dl. Make a copy of
287 * dl_name in case the first thread goes away before we do.
288 * Note that we initialize the new name before storing its
289 * pointer into dl_name, because the first thread may load
290 * dl->dl_name at any time. He'll either see the old value,
291 * which is his, or the new shared copy; either is OK.
293 dl
->dl_namesize
= strlen(dl
->dl_name
) + 1;
294 name
= kmem_alloc(dl
->dl_namesize
, KM_SLEEP
);
295 bcopy(dl
->dl_name
, name
, dl
->dl_namesize
);
299 mutex_exit(&dzp
->z_lock
);
302 * We have a dirlock on the name. (Note that it is the dirlock,
303 * not the dzp's z_lock, that protects the name in the zap object.)
304 * See if there's an object by this name; if so, put a hold on it.
307 error
= sa_lookup(dzp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
), &zoid
,
310 error
= (zoid
== 0 ? ENOENT
: 0);
314 vp
= dnlc_lookup(ZTOI(dzp
), name
);
315 if (vp
== DNLC_NO_VNODE
) {
320 zfs_dirent_unlock(dl
);
328 error
= zfs_match_find(zsb
, dzp
, name
, exact
,
329 update
, direntflags
, realpnp
, &zoid
);
332 error
= zfs_match_find(zsb
, dzp
, name
, exact
,
333 update
, direntflags
, realpnp
, &zoid
);
334 #endif /* HAVE_DNLC */
337 if (error
!= ENOENT
|| (flag
& ZEXISTS
)) {
338 zfs_dirent_unlock(dl
);
343 zfs_dirent_unlock(dl
);
346 error
= zfs_zget(zsb
, zoid
, zpp
);
348 zfs_dirent_unlock(dl
);
352 if (!(flag
& ZXATTR
) && update
)
353 dnlc_update(ZTOI(dzp
), name
, ZTOI(*zpp
));
354 #endif /* HAVE_DNLC */
363 * Unlock this directory entry and wake anyone who was waiting for it.
366 zfs_dirent_unlock(zfs_dirlock_t
*dl
)
368 znode_t
*dzp
= dl
->dl_dzp
;
369 zfs_dirlock_t
**prev_dl
, *cur_dl
;
371 mutex_enter(&dzp
->z_lock
);
373 if (!dl
->dl_namelock
)
374 rw_exit(&dzp
->z_name_lock
);
376 if (dl
->dl_sharecnt
> 1) {
378 mutex_exit(&dzp
->z_lock
);
381 prev_dl
= &dzp
->z_dirlocks
;
382 while ((cur_dl
= *prev_dl
) != dl
)
383 prev_dl
= &cur_dl
->dl_next
;
384 *prev_dl
= dl
->dl_next
;
385 cv_broadcast(&dl
->dl_cv
);
386 mutex_exit(&dzp
->z_lock
);
388 if (dl
->dl_namesize
!= 0)
389 kmem_free(dl
->dl_name
, dl
->dl_namesize
);
390 cv_destroy(&dl
->dl_cv
);
391 kmem_free(dl
, sizeof (*dl
));
395 * Look up an entry in a directory.
397 * NOTE: '.' and '..' are handled as special cases because
398 * no directory entries are actually stored for them. If this is
399 * the root of a filesystem, then '.zfs' is also treated as a
400 * special pseudo-directory.
403 zfs_dirlook(znode_t
*dzp
, char *name
, struct inode
**ipp
, int flags
,
404 int *deflg
, pathname_t
*rpnp
)
411 if (name
[0] == 0 || (name
[0] == '.' && name
[1] == 0)) {
414 } else if (name
[0] == '.' && name
[1] == '.' && name
[2] == 0) {
415 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
418 * If we are a snapshot mounted under .zfs, return
419 * the inode pointer for the snapshot directory.
421 if ((error
= sa_lookup(dzp
->z_sa_hdl
,
422 SA_ZPL_PARENT(zsb
), &parent
, sizeof (parent
))) != 0)
425 if (parent
== dzp
->z_id
&& zsb
->z_parent
!= zsb
) {
426 error
= zfsctl_root_lookup(zsb
->z_parent
->z_ctldir
,
427 "snapshot", ipp
, 0, kcred
, NULL
, NULL
);
430 rw_enter(&dzp
->z_parent_lock
, RW_READER
);
431 error
= zfs_zget(zsb
, parent
, &zp
);
434 rw_exit(&dzp
->z_parent_lock
);
435 } else if (zfs_has_ctldir(dzp
) && strcmp(name
, ZFS_CTLDIR_NAME
) == 0) {
436 *ipp
= zfsctl_root(dzp
);
440 zf
= ZEXISTS
| ZSHARED
;
441 if (flags
& FIGNORECASE
)
444 error
= zfs_dirent_lock(&dl
, dzp
, name
, &zp
, zf
, deflg
, rpnp
);
447 zfs_dirent_unlock(dl
);
448 dzp
->z_zn_prefetch
= B_TRUE
; /* enable prefetching */
453 if ((flags
& FIGNORECASE
) && rpnp
&& !error
)
454 (void) strlcpy(rpnp
->pn_buf
, name
, rpnp
->pn_bufsize
);
460 * unlinked Set (formerly known as the "delete queue") Error Handling
462 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
463 * don't specify the name of the entry that we will be manipulating. We
464 * also fib and say that we won't be adding any new entries to the
465 * unlinked set, even though we might (this is to lower the minimum file
466 * size that can be deleted in a full filesystem). So on the small
467 * chance that the nlink list is using a fat zap (ie. has more than
468 * 2000 entries), we *may* not pre-read a block that's needed.
469 * Therefore it is remotely possible for some of the assertions
470 * regarding the unlinked set below to fail due to i/o error. On a
471 * nondebug system, this will result in the space being leaked.
474 zfs_unlinked_add(znode_t
*zp
, dmu_tx_t
*tx
)
476 zfs_sb_t
*zsb
= ZTOZSB(zp
);
478 ASSERT(zp
->z_unlinked
);
479 ASSERT(zp
->z_links
== 0);
482 zap_add_int(zsb
->z_os
, zsb
->z_unlinkedobj
, zp
->z_id
, tx
));
486 * Delete the entire contents of a directory. Return a count
487 * of the number of entries that could not be deleted. If we encounter
488 * an error, return a count of at least one so that the directory stays
489 * in the unlinked set.
491 * NOTE: this function assumes that the directory is inactive,
492 * so there is no need to lock its entries before deletion.
493 * Also, it assumes the directory contents is *only* regular
497 zfs_purgedir(znode_t
*dzp
)
503 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
508 for (zap_cursor_init(&zc
, zsb
->z_os
, dzp
->z_id
);
509 (error
= zap_cursor_retrieve(&zc
, &zap
)) == 0;
510 zap_cursor_advance(&zc
)) {
511 error
= zfs_zget(zsb
,
512 ZFS_DIRENT_OBJ(zap
.za_first_integer
), &xzp
);
518 ASSERT(S_ISREG(ZTOI(xzp
)->i_mode
)||S_ISLNK(ZTOI(xzp
)->i_mode
));
520 tx
= dmu_tx_create(zsb
->z_os
);
521 dmu_tx_hold_sa(tx
, dzp
->z_sa_hdl
, B_FALSE
);
522 dmu_tx_hold_zap(tx
, dzp
->z_id
, FALSE
, zap
.za_name
);
523 dmu_tx_hold_sa(tx
, xzp
->z_sa_hdl
, B_FALSE
);
524 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, FALSE
, NULL
);
525 /* Is this really needed ? */
526 zfs_sa_upgrade_txholds(tx
, xzp
);
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 * Clean up any znodes that had no links when we either crashed or
553 * (force) umounted the file system.
556 zfs_unlinked_drain(zfs_sb_t
*zsb
)
560 dmu_object_info_t doi
;
565 * Interate over the contents of the unlinked set.
567 for (zap_cursor_init(&zc
, zsb
->z_os
, zsb
->z_unlinkedobj
);
568 zap_cursor_retrieve(&zc
, &zap
) == 0;
569 zap_cursor_advance(&zc
)) {
572 * See what kind of object we have in list
575 error
= dmu_object_info(zsb
->z_os
, zap
.za_first_integer
, &doi
);
579 ASSERT((doi
.doi_type
== DMU_OT_PLAIN_FILE_CONTENTS
) ||
580 (doi
.doi_type
== DMU_OT_DIRECTORY_CONTENTS
));
582 * We need to re-mark these list entries for deletion,
583 * so we pull them back into core and set zp->z_unlinked.
585 error
= zfs_zget(zsb
, zap
.za_first_integer
, &zp
);
588 * We may pick up znodes that are already marked for deletion.
589 * This could happen during the purge of an extended attribute
590 * directory. All we need to do is skip over them, since they
591 * are already in the system marked z_unlinked.
596 zp
->z_unlinked
= B_TRUE
;
599 * If this is an attribute directory, purge its contents.
601 if (S_ISDIR(ZTOI(zp
)->i_mode
) && (zp
->z_pflags
& ZFS_XATTR
)) {
603 * We don't need to check the return value of
604 * zfs_purgedir here, because zfs_rmnode will just
605 * return this xattr directory to the unlinked set
606 * until all of its xattrs are gone.
608 (void) zfs_purgedir(zp
);
613 zap_cursor_fini(&zc
);
617 zfs_rmnode(znode_t
*zp
)
619 zfs_sb_t
*zsb
= ZTOZSB(zp
);
620 objset_t
*os
= zsb
->z_os
;
628 ASSERT(zp
->z_links
== 0);
629 ASSERT(atomic_read(&ZTOI(zp
)->i_count
) == 0);
632 * If this is an attribute directory, purge its contents.
634 if (S_ISDIR(ZTOI(zp
)->i_mode
) && (zp
->z_pflags
& ZFS_XATTR
)) {
635 error
= zap_count(os
, zp
->z_id
, &count
);
637 zfs_znode_dmu_fini(zp
);
645 * There are still directory entries in this xattr
646 * directory. Let zfs_unlinked_drain() deal with
647 * them to avoid deadlocking this process in the
648 * zfs_purgedir()->zfs_zget()->ilookup() callpath
649 * on the xattr inode's I_FREEING bit.
651 taskq
= dsl_pool_iput_taskq(dmu_objset_pool(os
));
652 taskq_dispatch(taskq
, (task_func_t
*)
653 zfs_unlinked_drain
, zsb
, TQ_SLEEP
);
655 zfs_znode_dmu_fini(zp
);
661 * Free up all the data in the file.
663 error
= dmu_free_long_range(os
, zp
->z_id
, 0, DMU_OBJECT_END
);
666 * Not enough space. Leave the file in the unlinked set.
668 zfs_znode_dmu_fini(zp
);
673 * If the file has extended attributes, we're going to unlink
676 error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
),
677 &xattr_obj
, sizeof (xattr_obj
));
678 if (error
== 0 && xattr_obj
) {
679 error
= zfs_zget(zsb
, xattr_obj
, &xzp
);
683 acl_obj
= zfs_external_acl(zp
);
686 * Set up the final transaction.
688 tx
= dmu_tx_create(os
);
689 dmu_tx_hold_free(tx
, zp
->z_id
, 0, DMU_OBJECT_END
);
690 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, FALSE
, NULL
);
692 dmu_tx_hold_zap(tx
, zsb
->z_unlinkedobj
, TRUE
, NULL
);
693 dmu_tx_hold_sa(tx
, xzp
->z_sa_hdl
, B_FALSE
);
696 dmu_tx_hold_free(tx
, acl_obj
, 0, DMU_OBJECT_END
);
698 zfs_sa_upgrade_txholds(tx
, zp
);
699 error
= dmu_tx_assign(tx
, TXG_WAIT
);
702 * Not enough space to delete the file. Leave it in the
703 * unlinked set, leaking it until the fs is remounted (at
704 * which point we'll call zfs_unlinked_drain() to process it).
707 zfs_znode_dmu_fini(zp
);
713 mutex_enter(&xzp
->z_lock
);
714 xzp
->z_unlinked
= B_TRUE
; /* mark xzp for deletion */
715 xzp
->z_links
= 0; /* no more links to it */
716 VERIFY(0 == sa_update(xzp
->z_sa_hdl
, SA_ZPL_LINKS(zsb
),
717 &xzp
->z_links
, sizeof (xzp
->z_links
), tx
));
718 mutex_exit(&xzp
->z_lock
);
719 zfs_unlinked_add(xzp
, tx
);
722 /* Remove this znode from the unlinked set */
724 zap_remove_int(zsb
->z_os
, zsb
->z_unlinkedobj
, zp
->z_id
, tx
));
726 zfs_znode_delete(zp
, tx
);
735 zfs_dirent(znode_t
*zp
, uint64_t mode
)
737 uint64_t de
= zp
->z_id
;
739 if (ZTOZSB(zp
)->z_version
>= ZPL_VERSION_DIRENT_TYPE
)
740 de
|= IFTODT(mode
) << 60;
745 * Link zp into dl. Can only fail if zp has been unlinked.
748 zfs_link_create(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
)
750 znode_t
*dzp
= dl
->dl_dzp
;
751 zfs_sb_t
*zsb
= ZTOZSB(zp
);
753 int zp_is_dir
= S_ISDIR(ZTOI(zp
)->i_mode
);
754 sa_bulk_attr_t bulk
[5];
755 uint64_t mtime
[2], ctime
[2];
759 mutex_enter(&zp
->z_lock
);
761 if (!(flag
& ZRENAMING
)) {
762 if (zp
->z_unlinked
) { /* no new links to unlinked zp */
763 ASSERT(!(flag
& (ZNEW
| ZEXISTS
)));
764 mutex_exit(&zp
->z_lock
);
768 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
769 &zp
->z_links
, sizeof (zp
->z_links
));
772 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zsb
), NULL
,
773 &dzp
->z_id
, sizeof (dzp
->z_id
));
774 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
775 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
777 if (!(flag
& ZNEW
)) {
778 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
779 ctime
, sizeof (ctime
));
780 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, mtime
,
783 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
786 mutex_exit(&zp
->z_lock
);
788 mutex_enter(&dzp
->z_lock
);
790 dzp
->z_links
+= zp_is_dir
;
792 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
,
793 &dzp
->z_size
, sizeof (dzp
->z_size
));
794 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
795 &dzp
->z_links
, sizeof (dzp
->z_links
));
796 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
), NULL
,
797 mtime
, sizeof (mtime
));
798 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
799 ctime
, sizeof (ctime
));
800 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
801 &dzp
->z_pflags
, sizeof (dzp
->z_pflags
));
802 zfs_tstamp_update_setup(dzp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
803 error
= sa_bulk_update(dzp
->z_sa_hdl
, bulk
, count
, tx
);
805 mutex_exit(&dzp
->z_lock
);
807 value
= zfs_dirent(zp
, zp
->z_mode
);
808 error
= zap_add(ZTOZSB(zp
)->z_os
, dzp
->z_id
, dl
->dl_name
,
816 zfs_dropname(zfs_dirlock_t
*dl
, znode_t
*zp
, znode_t
*dzp
, dmu_tx_t
*tx
,
821 if (ZTOZSB(zp
)->z_norm
) {
822 if (((ZTOZSB(zp
)->z_case
== ZFS_CASE_INSENSITIVE
) &&
823 (flag
& ZCIEXACT
)) ||
824 ((ZTOZSB(zp
)->z_case
== ZFS_CASE_MIXED
) &&
826 error
= zap_remove_norm(ZTOZSB(zp
)->z_os
,
827 dzp
->z_id
, dl
->dl_name
, MT_EXACT
, tx
);
829 error
= zap_remove_norm(ZTOZSB(zp
)->z_os
,
830 dzp
->z_id
, dl
->dl_name
, MT_FIRST
, tx
);
832 error
= zap_remove(ZTOZSB(zp
)->z_os
,
833 dzp
->z_id
, dl
->dl_name
, tx
);
840 * Unlink zp from dl, and mark zp for deletion if this was the last link. Can
841 * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY).
842 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
843 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
844 * and it's the caller's job to do it.
847 zfs_link_destroy(zfs_dirlock_t
*dl
, znode_t
*zp
, dmu_tx_t
*tx
, int flag
,
848 boolean_t
*unlinkedp
)
850 znode_t
*dzp
= dl
->dl_dzp
;
851 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
852 int zp_is_dir
= S_ISDIR(ZTOI(zp
)->i_mode
);
853 boolean_t unlinked
= B_FALSE
;
854 sa_bulk_attr_t bulk
[5];
855 uint64_t mtime
[2], ctime
[2];
860 dnlc_remove(ZTOI(dzp
), dl
->dl_name
);
861 #endif /* HAVE_DNLC */
863 if (!(flag
& ZRENAMING
)) {
864 mutex_enter(&zp
->z_lock
);
866 if (zp_is_dir
&& !zfs_dirempty(zp
)) {
867 mutex_exit(&zp
->z_lock
);
872 * If we get here, we are going to try to remove the object.
873 * First try removing the name from the directory; if that
874 * fails, return the error.
876 error
= zfs_dropname(dl
, zp
, dzp
, tx
, flag
);
878 mutex_exit(&zp
->z_lock
);
882 if (zp
->z_links
<= zp_is_dir
) {
883 zfs_panic_recover("zfs: link count on %lu is %u, "
884 "should be at least %u", zp
->z_id
,
885 (int)zp
->z_links
, zp_is_dir
+ 1);
886 zp
->z_links
= zp_is_dir
+ 1;
888 if (--zp
->z_links
== zp_is_dir
) {
889 zp
->z_unlinked
= B_TRUE
;
893 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
),
894 NULL
, &ctime
, sizeof (ctime
));
895 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
896 NULL
, &zp
->z_pflags
, sizeof (zp
->z_pflags
));
897 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, mtime
, ctime
,
900 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
),
901 NULL
, &zp
->z_links
, sizeof (zp
->z_links
));
902 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
905 mutex_exit(&zp
->z_lock
);
907 error
= zfs_dropname(dl
, zp
, dzp
, tx
, flag
);
912 mutex_enter(&dzp
->z_lock
);
913 dzp
->z_size
--; /* one dirent removed */
914 dzp
->z_links
-= zp_is_dir
; /* ".." link from zp */
915 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
),
916 NULL
, &dzp
->z_links
, sizeof (dzp
->z_links
));
917 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
),
918 NULL
, &dzp
->z_size
, sizeof (dzp
->z_size
));
919 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
),
920 NULL
, ctime
, sizeof (ctime
));
921 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
),
922 NULL
, mtime
, sizeof (mtime
));
923 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
924 NULL
, &dzp
->z_pflags
, sizeof (dzp
->z_pflags
));
925 zfs_tstamp_update_setup(dzp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
926 error
= sa_bulk_update(dzp
->z_sa_hdl
, bulk
, count
, tx
);
928 mutex_exit(&dzp
->z_lock
);
930 if (unlinkedp
!= NULL
)
931 *unlinkedp
= unlinked
;
933 zfs_unlinked_add(zp
, tx
);
939 * Indicate whether the directory is empty. Works with or without z_lock
940 * held, but can only be consider a hint in the latter case. Returns true
941 * if only "." and ".." remain and there's no work in progress.
944 zfs_dirempty(znode_t
*dzp
)
946 return (dzp
->z_size
== 2 && dzp
->z_dirlocks
== 0);
950 zfs_make_xattrdir(znode_t
*zp
, vattr_t
*vap
, struct inode
**xipp
, cred_t
*cr
)
952 zfs_sb_t
*zsb
= ZTOZSB(zp
);
956 zfs_acl_ids_t acl_ids
;
957 boolean_t fuid_dirtied
;
964 if ((error
= zfs_zaccess(zp
, ACE_WRITE_NAMED_ATTRS
, 0, B_FALSE
, cr
)))
967 if ((error
= zfs_acl_ids_create(zp
, IS_XATTR
, vap
, cr
, NULL
,
970 if (zfs_acl_ids_overquota(zsb
, &acl_ids
)) {
971 zfs_acl_ids_free(&acl_ids
);
976 tx
= dmu_tx_create(zsb
->z_os
);
977 dmu_tx_hold_sa_create(tx
, acl_ids
.z_aclp
->z_acl_bytes
+
978 ZFS_SA_BASE_ATTR_SIZE
);
979 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_TRUE
);
980 dmu_tx_hold_zap(tx
, DMU_NEW_OBJECT
, FALSE
, NULL
);
981 fuid_dirtied
= zsb
->z_fuid_dirty
;
983 zfs_fuid_txhold(zsb
, tx
);
984 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
986 if (error
== ERESTART
) {
991 zfs_acl_ids_free(&acl_ids
);
995 zfs_mknode(zp
, vap
, tx
, cr
, IS_XATTR
, &xzp
, &acl_ids
);
998 zfs_fuid_sync(zsb
, tx
);
1001 error
= sa_lookup(xzp
->z_sa_hdl
, SA_ZPL_PARENT(zsb
),
1002 &parent
, sizeof (parent
));
1003 ASSERT(error
== 0 && parent
== zp
->z_id
);
1006 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_XATTR(zsb
), &xzp
->z_id
,
1007 sizeof (xzp
->z_id
), tx
));
1009 (void) zfs_log_create(zsb
->z_log
, tx
, TX_MKXATTR
, zp
,
1010 xzp
, "", NULL
, acl_ids
.z_fuidp
, vap
);
1012 zfs_acl_ids_free(&acl_ids
);
1021 * Return a znode for the extended attribute directory for zp.
1022 * ** If the directory does not already exist, it is created **
1024 * IN: zp - znode to obtain attribute directory from
1025 * cr - credentials of caller
1026 * flags - flags from the VOP_LOOKUP call
1028 * OUT: xipp - pointer to extended attribute znode
1030 * RETURN: 0 on success
1031 * error number on failure
1034 zfs_get_xattrdir(znode_t
*zp
, struct inode
**xipp
, cred_t
*cr
, int flags
)
1036 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1042 error
= zfs_dirent_lock(&dl
, zp
, "", &xzp
, ZXATTR
, NULL
, NULL
);
1048 zfs_dirent_unlock(dl
);
1052 if (!(flags
& CREATE_XATTR_DIR
)) {
1053 zfs_dirent_unlock(dl
);
1057 if (zfs_is_readonly(zsb
)) {
1058 zfs_dirent_unlock(dl
);
1063 * The ability to 'create' files in an attribute
1064 * directory comes from the write_xattr permission on the base file.
1066 * The ability to 'search' an attribute directory requires
1067 * read_xattr permission on the base file.
1069 * Once in a directory the ability to read/write attributes
1070 * is controlled by the permissions on the attribute file.
1072 va
.va_mask
= ATTR_MODE
| ATTR_UID
| ATTR_GID
;
1073 va
.va_mode
= S_IFDIR
| S_ISVTX
| 0777;
1074 zfs_fuid_map_ids(zp
, cr
, &va
.va_uid
, &va
.va_gid
);
1076 va
.va_dentry
= NULL
;
1077 error
= zfs_make_xattrdir(zp
, &va
, xipp
, cr
);
1078 zfs_dirent_unlock(dl
);
1080 if (error
== ERESTART
) {
1081 /* NB: we already did dmu_tx_wait() if necessary */
1089 * Decide whether it is okay to remove within a sticky directory.
1091 * In sticky directories, write access is not sufficient;
1092 * you can remove entries from a directory only if:
1094 * you own the directory,
1095 * you own the entry,
1096 * the entry is a plain file and you have write access,
1097 * or you are privileged (checked in secpolicy...).
1099 * The function returns 0 if remove access is granted.
1102 zfs_sticky_remove_access(znode_t
*zdp
, znode_t
*zp
, cred_t
*cr
)
1107 zfs_sb_t
*zsb
= ZTOZSB(zdp
);
1112 if ((zdp
->z_mode
& S_ISVTX
) == 0)
1115 downer
= zfs_fuid_map_id(zsb
, zdp
->z_uid
, cr
, ZFS_OWNER
);
1116 fowner
= zfs_fuid_map_id(zsb
, zp
->z_uid
, cr
, ZFS_OWNER
);
1118 if ((uid
= crgetuid(cr
)) == downer
|| uid
== fowner
||
1119 (S_ISDIR(ZTOI(zp
)->i_mode
) &&
1120 zfs_zaccess(zp
, ACE_WRITE_DATA
, 0, B_FALSE
, cr
) == 0))
1123 return (secpolicy_vnode_remove(cr
));