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 /* Portions Copyright 2007 Jeremy Teo */
29 #include <sys/types.h>
30 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/sysmacros.h>
34 #include <sys/resource.h>
35 #include <sys/mntent.h>
36 #include <sys/mkdev.h>
37 #include <sys/u8_textprep.h>
38 #include <sys/dsl_dataset.h>
40 #include <sys/vfs_opreg.h>
41 #include <sys/vnode.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
47 #include <sys/atomic.h>
49 #include "fs/fs_subr.h"
50 #include <sys/zfs_dir.h>
51 #include <sys/zfs_acl.h>
52 #include <sys/zfs_ioctl.h>
53 #include <sys/zfs_rlock.h>
54 #include <sys/zfs_fuid.h>
55 #include <sys/fs/zfs.h>
56 #include <sys/kidmap.h>
60 #include <sys/refcount.h>
63 #include <sys/zfs_znode.h>
68 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
69 * turned on when DEBUG is also defined.
76 #define ZNODE_STAT_ADD(stat) ((stat)++)
78 #define ZNODE_STAT_ADD(stat) /* nothing */
79 #endif /* ZNODE_STATS */
81 #define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3))
82 #define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1))
85 * Functions needed for userland (ie: libzpool) are not put under
86 * #ifdef_KERNEL; the rest of the functions have dependencies
87 * (such as VFS logic) that will not compile easily in userland.
90 static kmem_cache_t *znode_cache = NULL;
94 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
97 * We should never drop all dbuf refs without first clearing
98 * the eviction callback.
100 panic("evicting znode %p\n", user_ptr);
105 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
109 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
111 zp->z_vnode = vn_alloc(kmflags);
112 if (zp->z_vnode == NULL) {
115 ZTOV(zp)->v_data = zp;
117 list_link_init(&zp->z_link_node);
119 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
120 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
121 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
122 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
124 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
125 avl_create(&zp->z_range_avl, zfs_range_compare,
126 sizeof (rl_t), offsetof(rl_t, r_node));
129 zp->z_dirlocks = NULL;
135 zfs_znode_cache_destructor(void *buf, void *arg)
139 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
140 ASSERT(ZTOV(zp)->v_data == zp);
142 ASSERT(!list_link_active(&zp->z_link_node));
143 mutex_destroy(&zp->z_lock);
144 rw_destroy(&zp->z_parent_lock);
145 rw_destroy(&zp->z_name_lock);
146 mutex_destroy(&zp->z_acl_lock);
147 avl_destroy(&zp->z_range_avl);
148 mutex_destroy(&zp->z_range_lock);
150 ASSERT(zp->z_dbuf == NULL);
151 ASSERT(zp->z_dirlocks == NULL);
156 uint64_t zms_zfsvfs_invalid;
157 uint64_t zms_zfsvfs_unmounted;
158 uint64_t zms_zfsvfs_recheck_invalid;
159 uint64_t zms_obj_held;
160 uint64_t zms_vnode_locked;
161 uint64_t zms_not_only_dnlc;
163 #endif /* ZNODE_STATS */
166 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
171 nzp->z_zfsvfs = ozp->z_zfsvfs;
175 nzp->z_vnode = ozp->z_vnode;
176 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
177 ZTOV(ozp)->v_data = ozp;
178 ZTOV(nzp)->v_data = nzp;
180 nzp->z_id = ozp->z_id;
181 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
182 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
183 nzp->z_unlinked = ozp->z_unlinked;
184 nzp->z_atime_dirty = ozp->z_atime_dirty;
185 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
186 nzp->z_blksz = ozp->z_blksz;
187 nzp->z_seq = ozp->z_seq;
188 nzp->z_mapcnt = ozp->z_mapcnt;
189 nzp->z_last_itx = ozp->z_last_itx;
190 nzp->z_gen = ozp->z_gen;
191 nzp->z_sync_cnt = ozp->z_sync_cnt;
192 nzp->z_phys = ozp->z_phys;
193 nzp->z_dbuf = ozp->z_dbuf;
195 /* Update back pointers. */
196 (void) dmu_buf_update_user(nzp->z_dbuf, ozp, nzp, &nzp->z_phys,
200 * Invalidate the original znode by clearing fields that provide a
201 * pointer back to the znode. Set the low bit of the vfs pointer to
202 * ensure that zfs_znode_move() recognizes the znode as invalid in any
203 * subsequent callback.
206 POINTER_INVALIDATE(&ozp->z_zfsvfs);
210 * Wrapper function for ZFS_ENTER that returns 0 if successful and otherwise
211 * returns a non-zero error code.
214 zfs_enter(zfsvfs_t *zfsvfs)
222 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
224 znode_t *ozp = buf, *nzp = newbuf;
229 * The znode is on the file system's list of known znodes if the vfs
230 * pointer is valid. We set the low bit of the vfs pointer when freeing
231 * the znode to invalidate it, and the memory patterns written by kmem
232 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
233 * created znode sets the vfs pointer last of all to indicate that the
234 * znode is known and in a valid state to be moved by this function.
236 zfsvfs = ozp->z_zfsvfs;
237 if (!POINTER_IS_VALID(zfsvfs)) {
238 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
239 return (KMEM_CBRC_DONT_KNOW);
243 * Ensure that the filesystem is not unmounted during the move.
245 if (zfs_enter(zfsvfs) != 0) { /* ZFS_ENTER */
246 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
247 return (KMEM_CBRC_DONT_KNOW);
250 mutex_enter(&zfsvfs->z_znodes_lock);
252 * Recheck the vfs pointer in case the znode was removed just before
253 * acquiring the lock.
255 if (zfsvfs != ozp->z_zfsvfs) {
256 mutex_exit(&zfsvfs->z_znodes_lock);
258 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck_invalid);
259 return (KMEM_CBRC_DONT_KNOW);
263 * At this point we know that as long as we hold z_znodes_lock, the
264 * znode cannot be freed and fields within the znode can be safely
265 * accessed. Now, prevent a race with zfs_zget().
267 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
268 mutex_exit(&zfsvfs->z_znodes_lock);
270 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
271 return (KMEM_CBRC_LATER);
275 if (mutex_tryenter(&vp->v_lock) == 0) {
276 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
277 mutex_exit(&zfsvfs->z_znodes_lock);
279 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
280 return (KMEM_CBRC_LATER);
283 /* Only move znodes that are referenced _only_ by the DNLC. */
284 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
285 mutex_exit(&vp->v_lock);
286 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
287 mutex_exit(&zfsvfs->z_znodes_lock);
289 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
290 return (KMEM_CBRC_LATER);
294 * The znode is known and in a valid state to move. We're holding the
295 * locks needed to execute the critical section.
297 zfs_znode_move_impl(ozp, nzp);
298 mutex_exit(&vp->v_lock);
299 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
301 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
302 mutex_exit(&zfsvfs->z_znodes_lock);
305 return (KMEM_CBRC_YES);
314 ASSERT(znode_cache == NULL);
315 znode_cache = kmem_cache_create("zfs_znode_cache",
316 sizeof (znode_t), 0, zfs_znode_cache_constructor,
317 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
318 kmem_cache_set_move(znode_cache, zfs_znode_move);
325 * Cleanup vfs & vnode ops
327 zfs_remove_op_tables();
333 kmem_cache_destroy(znode_cache);
337 struct vnodeops *zfs_dvnodeops;
338 struct vnodeops *zfs_fvnodeops;
339 struct vnodeops *zfs_symvnodeops;
340 struct vnodeops *zfs_xdvnodeops;
341 struct vnodeops *zfs_evnodeops;
344 zfs_remove_op_tables()
350 (void) vfs_freevfsops_by_type(zfsfstype);
357 vn_freevnodeops(zfs_dvnodeops);
359 vn_freevnodeops(zfs_fvnodeops);
361 vn_freevnodeops(zfs_symvnodeops);
363 vn_freevnodeops(zfs_xdvnodeops);
365 vn_freevnodeops(zfs_evnodeops);
367 zfs_dvnodeops = NULL;
368 zfs_fvnodeops = NULL;
369 zfs_symvnodeops = NULL;
370 zfs_xdvnodeops = NULL;
371 zfs_evnodeops = NULL;
374 extern const fs_operation_def_t zfs_dvnodeops_template[];
375 extern const fs_operation_def_t zfs_fvnodeops_template[];
376 extern const fs_operation_def_t zfs_xdvnodeops_template[];
377 extern const fs_operation_def_t zfs_symvnodeops_template[];
378 extern const fs_operation_def_t zfs_evnodeops_template[];
381 zfs_create_op_tables()
386 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
387 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
388 * In this case we just return as the ops vectors are already set up.
393 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
398 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
403 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
408 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
413 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
420 * zfs_init_fs - Initialize the zfsvfs struct and the file system
421 * incore "master" object. Verify version compatibility.
424 zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp)
426 extern int zfsfstype;
428 objset_t *os = zfsvfs->z_os;
435 error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
438 } else if (zfsvfs->z_version > ZPL_VERSION) {
439 (void) printf("Mismatched versions: File system "
440 "is version %llu on-disk format, which is "
441 "incompatible with this software version %lld!",
442 (u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
446 if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
448 zfsvfs->z_norm = (int)zval;
449 if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
451 zfsvfs->z_utf8 = (zval != 0);
452 if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
454 zfsvfs->z_case = (uint_t)zval;
456 * Fold case on file systems that are always or sometimes case
459 if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
460 zfsvfs->z_case == ZFS_CASE_MIXED)
461 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
464 * The fsid is 64 bits, composed of an 8-bit fs type, which
465 * separates our fsid from any other filesystem types, and a
466 * 56-bit objset unique ID. The objset unique ID is unique to
467 * all objsets open on this system, provided by unique_create().
468 * The 8-bit fs type must be put in the low bits of fsid[1]
469 * because that's where other Solaris filesystems put it.
471 fsid_guid = dmu_objset_fsid_guid(os);
472 ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
473 zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid;
474 zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
477 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
481 ASSERT(zfsvfs->z_root != 0);
483 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
484 &zfsvfs->z_unlinkedobj);
489 * Initialize zget mutex's
491 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
492 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
494 error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp);
497 * On error, we destroy the mutexes here since it's not
498 * possible for the caller to determine if the mutexes were
499 * initialized properly.
501 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
502 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
505 ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root);
506 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
507 &zfsvfs->z_fuid_obj);
515 * define a couple of values we need available
516 * for both 64 and 32 bit environments.
519 #define NBITSMINOR64 32
522 #define MAXMAJ64 0xffffffffUL
525 #define MAXMIN64 0xffffffffUL
529 * Create special expldev for ZFS private use.
530 * Can't use standard expldev since it doesn't do
531 * what we want. The standard expldev() takes a
532 * dev32_t in LP64 and expands it to a long dev_t.
533 * We need an interface that takes a dev32_t in ILP32
534 * and expands it to a long dev_t.
537 zfs_expldev(dev_t dev)
540 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
541 return (((uint64_t)major << NBITSMINOR64) |
542 ((minor_t)dev & MAXMIN32));
549 * Special cmpldev for ZFS private use.
550 * Can't use standard cmpldev since it takes
551 * a long dev_t and compresses it to dev32_t in
552 * LP64. We need to do a compaction of a long dev_t
553 * to a dev32_t in ILP32.
556 zfs_cmpldev(uint64_t dev)
559 minor_t minor = (minor_t)dev & MAXMIN64;
560 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
562 if (major > MAXMAJ32 || minor > MAXMIN32)
565 return (((dev32_t)major << NBITSMINOR32) | minor);
572 zfs_znode_dmu_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db)
576 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
577 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
579 mutex_enter(&zp->z_lock);
581 ASSERT(zp->z_dbuf == NULL);
583 nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error);
587 * concurrent zgets on this object.
590 panic("existing znode %p for dbuf %p", (void *)nzp, (void *)db);
593 * Slap on VROOT if we are the root znode
595 if (zp->z_id == zfsvfs->z_root)
596 ZTOV(zp)->v_flag |= VROOT;
598 mutex_exit(&zp->z_lock);
603 zfs_znode_dmu_fini(znode_t *zp)
605 dmu_buf_t *db = zp->z_dbuf;
606 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
608 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
609 ASSERT(zp->z_dbuf != NULL);
611 VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL));
612 dmu_buf_rele(db, NULL);
616 * Construct a new znode/vnode and intialize.
618 * This does not do a call to dmu_set_user() that is
619 * up to the caller to do, in case you don't want to
623 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz)
628 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
630 ASSERT(zp->z_dirlocks == NULL);
631 ASSERT(zp->z_dbuf == NULL);
632 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
635 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
636 * the zfs_znode_move() callback.
640 zp->z_atime_dirty = 0;
643 zp->z_id = db->db_object;
645 zp->z_seq = 0x7A4653;
651 zfs_znode_dmu_init(zfsvfs, zp, db);
653 zp->z_gen = zp->z_phys->zp_gen;
655 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
656 vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode);
658 switch (vp->v_type) {
660 if (zp->z_phys->zp_flags & ZFS_XATTR) {
661 vn_setops(vp, zfs_xdvnodeops);
662 vp->v_flag |= V_XATTRDIR;
664 vn_setops(vp, zfs_dvnodeops);
666 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
670 vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev);
675 vn_setops(vp, zfs_fvnodeops);
678 vp->v_flag |= VMODSORT;
679 vn_setops(vp, zfs_fvnodeops);
682 vn_setops(vp, zfs_symvnodeops);
685 vn_setops(vp, zfs_evnodeops);
689 mutex_enter(&zfsvfs->z_znodes_lock);
690 list_insert_tail(&zfsvfs->z_all_znodes, zp);
693 * Everything else must be valid before assigning z_zfsvfs makes the
694 * znode eligible for zfs_znode_move().
696 zp->z_zfsvfs = zfsvfs;
697 mutex_exit(&zfsvfs->z_znodes_lock);
699 VFS_HOLD(zfsvfs->z_vfs);
704 * Create a new DMU object to hold a zfs znode.
706 * IN: dzp - parent directory for new znode
707 * vap - file attributes for new znode
708 * tx - dmu transaction id for zap operations
709 * cr - credentials of caller
711 * IS_ROOT_NODE - new object will be root
712 * IS_XATTR - new object is an attribute
713 * IS_REPLAY - intent log replay
714 * bonuslen - length of bonus buffer
715 * setaclp - File/Dir initial ACL
716 * fuidp - Tracks fuid allocation.
718 * OUT: zpp - allocated znode
722 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
723 uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_t *setaclp,
724 zfs_fuid_info_t **fuidp)
728 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
733 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
735 if (zfsvfs->z_replay) {
736 obj = vap->va_nodeid;
738 now = vap->va_ctime; /* see zfs_replay_create() */
739 gen = vap->va_nblocks; /* ditto */
743 gen = dmu_tx_get_txg(tx);
747 * Create a new DMU object.
750 * There's currently no mechanism for pre-reading the blocks that will
751 * be to needed allocate a new object, so we accept the small chance
752 * that there will be an i/o error and we will fail one of the
755 if (vap->va_type == VDIR) {
756 if (flag & IS_REPLAY) {
757 err = zap_create_claim_norm(zfsvfs->z_os, obj,
758 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
759 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
760 ASSERT3U(err, ==, 0);
762 obj = zap_create_norm(zfsvfs->z_os,
763 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
764 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
767 if (flag & IS_REPLAY) {
768 err = dmu_object_claim(zfsvfs->z_os, obj,
769 DMU_OT_PLAIN_FILE_CONTENTS, 0,
770 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
771 ASSERT3U(err, ==, 0);
773 obj = dmu_object_alloc(zfsvfs->z_os,
774 DMU_OT_PLAIN_FILE_CONTENTS, 0,
775 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
778 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db));
779 dmu_buf_will_dirty(db, tx);
782 * Initialize the znode physical data to zero.
784 ASSERT(db->db_size >= sizeof (znode_phys_t));
785 bzero(db->db_data, db->db_size);
789 * If this is the root, fix up the half-initialized parent pointer
790 * to reference the just-allocated physical data area.
792 if (flag & IS_ROOT_NODE) {
799 * If parent is an xattr, so am I.
801 if (dzp->z_phys->zp_flags & ZFS_XATTR)
804 if (vap->va_type == VBLK || vap->va_type == VCHR) {
805 pzp->zp_rdev = zfs_expldev(vap->va_rdev);
808 if (zfsvfs->z_use_fuids)
809 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
811 if (vap->va_type == VDIR) {
812 pzp->zp_size = 2; /* contents ("." and "..") */
813 pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
816 pzp->zp_parent = dzp->z_id;
818 pzp->zp_flags |= ZFS_XATTR;
822 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
823 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
825 if (vap->va_mask & AT_ATIME) {
826 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
828 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
831 if (vap->va_mask & AT_MTIME) {
832 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
834 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
837 pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode);
838 if (!(flag & IS_ROOT_NODE)) {
839 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
840 *zpp = zfs_znode_alloc(zfsvfs, db, 0);
841 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
844 * If we are creating the root node, the "parent" we
845 * passed in is the znode for the root.
849 zfs_perm_init(*zpp, dzp, flag, vap, tx, cr, setaclp, fuidp);
853 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap)
857 xoap = xva_getxoptattr(xvap);
860 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
861 ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime);
862 XVA_SET_RTN(xvap, XAT_CREATETIME);
864 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
865 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly);
866 XVA_SET_RTN(xvap, XAT_READONLY);
868 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
869 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden);
870 XVA_SET_RTN(xvap, XAT_HIDDEN);
872 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
873 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system);
874 XVA_SET_RTN(xvap, XAT_SYSTEM);
876 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
877 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive);
878 XVA_SET_RTN(xvap, XAT_ARCHIVE);
880 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
881 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable);
882 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
884 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
885 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink);
886 XVA_SET_RTN(xvap, XAT_NOUNLINK);
888 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
889 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly);
890 XVA_SET_RTN(xvap, XAT_APPENDONLY);
892 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
893 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump);
894 XVA_SET_RTN(xvap, XAT_NODUMP);
896 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
897 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque);
898 XVA_SET_RTN(xvap, XAT_OPAQUE);
900 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
901 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
902 xoap->xoa_av_quarantined);
903 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
905 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
906 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified);
907 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
909 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
910 (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp,
911 sizeof (xoap->xoa_av_scanstamp));
912 zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP;
913 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
918 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
920 dmu_object_info_t doi;
927 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
929 err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
931 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
935 dmu_object_info_from_db(db, &doi);
936 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
937 doi.doi_bonus_size < sizeof (znode_phys_t)) {
938 dmu_buf_rele(db, NULL);
939 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
943 zp = dmu_buf_get_user(db);
945 mutex_enter(&zp->z_lock);
948 * Since we do immediate eviction of the z_dbuf, we
949 * should never find a dbuf with a znode that doesn't
950 * know about the dbuf.
952 ASSERT3P(zp->z_dbuf, ==, db);
953 ASSERT3U(zp->z_id, ==, obj_num);
954 if (zp->z_unlinked) {
961 dmu_buf_rele(db, NULL);
962 mutex_exit(&zp->z_lock);
963 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
968 * Not found create new znode/vnode
970 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size);
971 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
977 zfs_rezget(znode_t *zp)
979 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
980 dmu_object_info_t doi;
982 uint64_t obj_num = zp->z_id;
985 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
987 err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
989 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
993 dmu_object_info_from_db(db, &doi);
994 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
995 doi.doi_bonus_size < sizeof (znode_phys_t)) {
996 dmu_buf_rele(db, NULL);
997 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1001 if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) {
1002 dmu_buf_rele(db, NULL);
1003 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1007 zfs_znode_dmu_init(zfsvfs, zp, db);
1008 zp->z_unlinked = (zp->z_phys->zp_links == 0);
1009 zp->z_blksz = doi.doi_data_block_size;
1011 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1017 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1019 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1020 objset_t *os = zfsvfs->z_os;
1021 uint64_t obj = zp->z_id;
1022 uint64_t acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
1024 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1026 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1027 VERIFY(0 == dmu_object_free(os, obj, tx));
1028 zfs_znode_dmu_fini(zp);
1029 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1034 zfs_zinactive(znode_t *zp)
1036 vnode_t *vp = ZTOV(zp);
1037 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1038 uint64_t z_id = zp->z_id;
1040 ASSERT(zp->z_dbuf && zp->z_phys);
1043 * Don't allow a zfs_zget() while were trying to release this znode
1045 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1047 mutex_enter(&zp->z_lock);
1048 mutex_enter(&vp->v_lock);
1050 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1052 * If the hold count is greater than zero, somebody has
1053 * obtained a new reference on this znode while we were
1054 * processing it here, so we are done. If we still have
1055 * mapped pages then we are also done, since we don't
1056 * want to inactivate the znode until the pages get pushed.
1058 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1059 * this seems like it would leave the znode hanging with
1060 * no chance to go inactive...
1062 mutex_exit(&vp->v_lock);
1063 mutex_exit(&zp->z_lock);
1064 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1067 mutex_exit(&vp->v_lock);
1070 * If this was the last reference to a file with no links,
1071 * remove the file from the file system.
1073 if (zp->z_unlinked) {
1074 mutex_exit(&zp->z_lock);
1075 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1079 mutex_exit(&zp->z_lock);
1080 zfs_znode_dmu_fini(zp);
1081 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1086 zfs_znode_free(znode_t *zp)
1088 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1090 vn_invalid(ZTOV(zp));
1092 ASSERT(ZTOV(zp)->v_count == 0);
1094 mutex_enter(&zfsvfs->z_znodes_lock);
1095 POINTER_INVALIDATE(&zp->z_zfsvfs);
1096 list_remove(&zfsvfs->z_all_znodes, zp);
1097 mutex_exit(&zfsvfs->z_znodes_lock);
1099 kmem_cache_free(znode_cache, zp);
1101 VFS_RELE(zfsvfs->z_vfs);
1105 zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx)
1109 ASSERT(MUTEX_HELD(&zp->z_lock));
1114 dmu_buf_will_dirty(zp->z_dbuf, tx);
1115 zp->z_atime_dirty = 0;
1118 zp->z_atime_dirty = 1;
1121 if (flag & AT_ATIME)
1122 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime);
1124 if (flag & AT_MTIME) {
1125 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime);
1126 if (zp->z_zfsvfs->z_use_fuids)
1127 zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED);
1130 if (flag & AT_CTIME) {
1131 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime);
1132 if (zp->z_zfsvfs->z_use_fuids)
1133 zp->z_phys->zp_flags |= ZFS_ARCHIVE;
1138 * Update the requested znode timestamps with the current time.
1139 * If we are in a transaction, then go ahead and mark the znode
1140 * dirty in the transaction so the timestamps will go to disk.
1141 * Otherwise, we will get pushed next time the znode is updated
1142 * in a transaction, or when this znode eventually goes inactive.
1145 * 1 - Only the ACCESS time is ever updated outside of a transaction.
1146 * 2 - Multiple consecutive updates will be collapsed into a single
1147 * znode update by the transaction grouping semantics of the DMU.
1150 zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx)
1152 mutex_enter(&zp->z_lock);
1153 zfs_time_stamper_locked(zp, flag, tx);
1154 mutex_exit(&zp->z_lock);
1158 * Grow the block size for a file.
1160 * IN: zp - znode of file to free data in.
1161 * size - requested block size
1162 * tx - open transaction.
1164 * NOTE: this function assumes that the znode is write locked.
1167 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1172 if (size <= zp->z_blksz)
1175 * If the file size is already greater than the current blocksize,
1176 * we will not grow. If there is more than one block in a file,
1177 * the blocksize cannot change.
1179 if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz)
1182 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1184 if (error == ENOTSUP)
1186 ASSERT3U(error, ==, 0);
1188 /* What blocksize did we actually get? */
1189 dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy);
1193 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1194 * be calling back into the fs for a putpage(). E.g.: when truncating
1195 * a file, the pages being "thrown away* don't need to be written out.
1199 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1200 int flags, cred_t *cr)
1207 * Increase the file length
1209 * IN: zp - znode of file to free data in.
1210 * end - new end-of-file
1212 * RETURN: 0 if success
1213 * error code if failure
1216 zfs_extend(znode_t *zp, uint64_t end)
1218 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1225 * We will change zp_size, lock the whole file.
1227 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1230 * Nothing to do if file already at desired length.
1232 if (end <= zp->z_phys->zp_size) {
1233 zfs_range_unlock(rl);
1237 tx = dmu_tx_create(zfsvfs->z_os);
1238 dmu_tx_hold_bonus(tx, zp->z_id);
1239 if (end > zp->z_blksz &&
1240 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1242 * We are growing the file past the current block size.
1244 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1245 ASSERT(!ISP2(zp->z_blksz));
1246 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1248 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1250 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1255 error = dmu_tx_assign(tx, TXG_NOWAIT);
1257 if (error == ERESTART) {
1263 zfs_range_unlock(rl);
1266 dmu_buf_will_dirty(zp->z_dbuf, tx);
1269 zfs_grow_blocksize(zp, newblksz, tx);
1271 zp->z_phys->zp_size = end;
1273 zfs_range_unlock(rl);
1281 * Free space in a file.
1283 * IN: zp - znode of file to free data in.
1284 * off - start of section to free.
1285 * len - length of section to free.
1287 * RETURN: 0 if success
1288 * error code if failure
1291 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1293 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1298 * Lock the range being freed.
1300 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1303 * Nothing to do if file already at desired length.
1305 if (off >= zp->z_phys->zp_size) {
1306 zfs_range_unlock(rl);
1310 if (off + len > zp->z_phys->zp_size)
1311 len = zp->z_phys->zp_size - off;
1313 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1315 zfs_range_unlock(rl);
1323 * IN: zp - znode of file to free data in.
1324 * end - new end-of-file.
1326 * RETURN: 0 if success
1327 * error code if failure
1330 zfs_trunc(znode_t *zp, uint64_t end)
1332 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1333 vnode_t *vp = ZTOV(zp);
1339 * We will change zp_size, lock the whole file.
1341 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1344 * Nothing to do if file already at desired length.
1346 if (end >= zp->z_phys->zp_size) {
1347 zfs_range_unlock(rl);
1351 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1353 zfs_range_unlock(rl);
1357 tx = dmu_tx_create(zfsvfs->z_os);
1358 dmu_tx_hold_bonus(tx, zp->z_id);
1359 error = dmu_tx_assign(tx, TXG_NOWAIT);
1361 if (error == ERESTART) {
1367 zfs_range_unlock(rl);
1370 dmu_buf_will_dirty(zp->z_dbuf, tx);
1372 zp->z_phys->zp_size = end;
1377 * Clear any mapped pages in the truncated region. This has to
1378 * happen outside of the transaction to avoid the possibility of
1379 * a deadlock with someone trying to push a page that we are
1380 * about to invalidate.
1382 if (vn_has_cached_data(vp)) {
1384 uint64_t start = end & PAGEMASK;
1385 int poff = end & PAGEOFFSET;
1387 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1389 * We need to zero a partial page.
1391 pagezero(pp, poff, PAGESIZE - poff);
1395 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1396 B_INVAL | B_TRUNC, NULL);
1400 zfs_range_unlock(rl);
1406 * Free space in a file
1408 * IN: zp - znode of file to free data in.
1409 * off - start of range
1410 * len - end of range (0 => EOF)
1411 * flag - current file open mode flags.
1412 * log - TRUE if this action should be logged
1414 * RETURN: 0 if success
1415 * error code if failure
1418 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1420 vnode_t *vp = ZTOV(zp);
1422 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1423 zilog_t *zilog = zfsvfs->z_log;
1426 if (off > zp->z_phys->zp_size) {
1427 error = zfs_extend(zp, off+len);
1428 if (error == 0 && log)
1435 * Check for any locks in the region to be freed.
1437 if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) {
1438 uint64_t length = (len ? len : zp->z_phys->zp_size - off);
1439 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1444 error = zfs_trunc(zp, off);
1446 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1447 off + len > zp->z_phys->zp_size)
1448 error = zfs_extend(zp, off+len);
1453 tx = dmu_tx_create(zfsvfs->z_os);
1454 dmu_tx_hold_bonus(tx, zp->z_id);
1455 error = dmu_tx_assign(tx, TXG_NOWAIT);
1457 if (error == ERESTART) {
1466 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
1467 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1474 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1477 uint64_t moid, doid, version;
1478 uint64_t sense = ZFS_CASE_SENSITIVE;
1482 znode_t *rootzp = NULL;
1488 * First attempt to create master node.
1491 * In an empty objset, there are no blocks to read and thus
1492 * there can be no i/o errors (which we assert below).
1494 moid = MASTER_NODE_OBJ;
1495 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1496 DMU_OT_NONE, 0, tx);
1500 * Set starting attributes.
1502 if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
1503 version = ZPL_VERSION;
1505 version = ZPL_VERSION_FUID - 1;
1506 error = zap_update(os, moid, ZPL_VERSION_STR,
1507 8, 1, &version, tx);
1509 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1510 /* For the moment we expect all zpl props to be uint64_ts */
1514 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1515 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1516 name = nvpair_name(elem);
1517 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1519 error = zap_update(os, moid, ZPL_VERSION_STR,
1520 8, 1, &version, tx);
1522 error = zap_update(os, moid, name, 8, 1, &val, tx);
1525 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1527 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1530 ASSERT(version != 0);
1533 * Create a delete queue.
1535 doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1537 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx);
1541 * Create root znode. Create minimal znode/vnode/zfsvfs
1542 * to allow zfs_mknode to work.
1544 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1545 vattr.va_type = VDIR;
1546 vattr.va_mode = S_IFDIR|0755;
1547 vattr.va_uid = crgetuid(cr);
1548 vattr.va_gid = crgetgid(cr);
1550 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1551 rootzp->z_unlinked = 0;
1552 rootzp->z_atime_dirty = 0;
1558 bzero(&zfsvfs, sizeof (zfsvfs_t));
1561 zfsvfs.z_parent = &zfsvfs;
1562 zfsvfs.z_version = version;
1563 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1564 zfsvfs.z_norm = norm;
1566 * Fold case on file systems that are always or sometimes case
1569 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1570 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1572 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1573 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1574 offsetof(znode_t, z_link_node));
1576 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1577 rootzp->z_zfsvfs = &zfsvfs;
1578 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, NULL, NULL);
1579 ASSERT3P(zp, ==, rootzp);
1580 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1581 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1583 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1585 ZTOV(rootzp)->v_count = 0;
1586 dmu_buf_rele(rootzp->z_dbuf, NULL);
1587 rootzp->z_dbuf = NULL;
1588 kmem_cache_free(znode_cache, rootzp);
1591 #endif /* _KERNEL */
1593 * Given an object number, return its parent object number and whether
1594 * or not the object is an extended attribute directory.
1597 zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir)
1600 dmu_object_info_t doi;
1604 if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0)
1607 dmu_object_info_from_db(db, &doi);
1608 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
1609 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1610 dmu_buf_rele(db, FTAG);
1615 *pobjp = zp->zp_parent;
1616 *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) &&
1617 S_ISDIR(zp->zp_mode);
1618 dmu_buf_rele(db, FTAG);
1624 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1626 char *path = buf + len - 1;
1633 char component[MAXNAMELEN + 2];
1637 if ((error = zfs_obj_to_pobj(osp, obj, &pobj,
1638 &is_xattrdir)) != 0)
1649 (void) sprintf(component + 1, "<xattrdir>");
1651 error = zap_value_search(osp, pobj, obj,
1652 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1657 complen = strlen(component);
1659 ASSERT(path >= buf);
1660 bcopy(component, path, complen);
1665 (void) memmove(buf, path, buf + len - path);