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.
91 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
92 * be freed before it can be safely accessed.
94 krwlock_t zfsvfs_lock;
96 static kmem_cache_t *znode_cache = NULL;
100 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
103 * We should never drop all dbuf refs without first clearing
104 * the eviction callback.
106 panic("evicting znode %p\n", user_ptr);
111 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
115 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
117 zp->z_vnode = vn_alloc(kmflags);
118 if (zp->z_vnode == NULL) {
121 ZTOV(zp)->v_data = zp;
123 list_link_init(&zp->z_link_node);
125 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
126 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
127 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
128 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
130 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
131 avl_create(&zp->z_range_avl, zfs_range_compare,
132 sizeof (rl_t), offsetof(rl_t, r_node));
135 zp->z_dirlocks = NULL;
141 zfs_znode_cache_destructor(void *buf, void *arg)
145 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
146 ASSERT(ZTOV(zp)->v_data == zp);
148 ASSERT(!list_link_active(&zp->z_link_node));
149 mutex_destroy(&zp->z_lock);
150 rw_destroy(&zp->z_parent_lock);
151 rw_destroy(&zp->z_name_lock);
152 mutex_destroy(&zp->z_acl_lock);
153 avl_destroy(&zp->z_range_avl);
154 mutex_destroy(&zp->z_range_lock);
156 ASSERT(zp->z_dbuf == NULL);
157 ASSERT(zp->z_dirlocks == NULL);
162 uint64_t zms_zfsvfs_invalid;
163 uint64_t zms_zfsvfs_recheck1;
164 uint64_t zms_zfsvfs_unmounted;
165 uint64_t zms_zfsvfs_recheck2;
166 uint64_t zms_obj_held;
167 uint64_t zms_vnode_locked;
168 uint64_t zms_not_only_dnlc;
170 #endif /* ZNODE_STATS */
173 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
178 nzp->z_zfsvfs = ozp->z_zfsvfs;
182 nzp->z_vnode = ozp->z_vnode;
183 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
184 ZTOV(ozp)->v_data = ozp;
185 ZTOV(nzp)->v_data = nzp;
187 nzp->z_id = ozp->z_id;
188 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
189 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
190 nzp->z_unlinked = ozp->z_unlinked;
191 nzp->z_atime_dirty = ozp->z_atime_dirty;
192 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
193 nzp->z_blksz = ozp->z_blksz;
194 nzp->z_seq = ozp->z_seq;
195 nzp->z_mapcnt = ozp->z_mapcnt;
196 nzp->z_last_itx = ozp->z_last_itx;
197 nzp->z_gen = ozp->z_gen;
198 nzp->z_sync_cnt = ozp->z_sync_cnt;
199 nzp->z_phys = ozp->z_phys;
200 nzp->z_dbuf = ozp->z_dbuf;
202 /* Update back pointers. */
203 (void) dmu_buf_update_user(nzp->z_dbuf, ozp, nzp, &nzp->z_phys,
207 * Invalidate the original znode by clearing fields that provide a
208 * pointer back to the znode. Set the low bit of the vfs pointer to
209 * ensure that zfs_znode_move() recognizes the znode as invalid in any
210 * subsequent callback.
213 POINTER_INVALIDATE(&ozp->z_zfsvfs);
218 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
220 znode_t *ozp = buf, *nzp = newbuf;
225 * The znode is on the file system's list of known znodes if the vfs
226 * pointer is valid. We set the low bit of the vfs pointer when freeing
227 * the znode to invalidate it, and the memory patterns written by kmem
228 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
229 * created znode sets the vfs pointer last of all to indicate that the
230 * znode is known and in a valid state to be moved by this function.
232 zfsvfs = ozp->z_zfsvfs;
233 if (!POINTER_IS_VALID(zfsvfs)) {
234 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
235 return (KMEM_CBRC_DONT_KNOW);
239 * Close a small window in which it's possible that the filesystem could
240 * be unmounted and freed, and zfsvfs, though valid in the previous
241 * statement, could point to unrelated memory by the time we try to
242 * prevent the filesystem from being unmounted.
244 rw_enter(&zfsvfs_lock, RW_WRITER);
245 if (zfsvfs != ozp->z_zfsvfs) {
246 rw_exit(&zfsvfs_lock);
247 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
248 return (KMEM_CBRC_DONT_KNOW);
252 * If the znode is still valid, then so is the file system. We know that
253 * no valid file system can be freed while we hold zfsvfs_lock, so we
254 * can safely ensure that the filesystem is not and will not be
255 * unmounted. The next statement is equivalent to ZFS_ENTER().
257 rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
258 if (zfsvfs->z_unmounted) {
260 rw_exit(&zfsvfs_lock);
261 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
262 return (KMEM_CBRC_DONT_KNOW);
264 rw_exit(&zfsvfs_lock);
266 mutex_enter(&zfsvfs->z_znodes_lock);
268 * Recheck the vfs pointer in case the znode was removed just before
269 * acquiring the lock.
271 if (zfsvfs != ozp->z_zfsvfs) {
272 mutex_exit(&zfsvfs->z_znodes_lock);
274 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
275 return (KMEM_CBRC_DONT_KNOW);
279 * At this point we know that as long as we hold z_znodes_lock, the
280 * znode cannot be freed and fields within the znode can be safely
281 * accessed. Now, prevent a race with zfs_zget().
283 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
284 mutex_exit(&zfsvfs->z_znodes_lock);
286 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
287 return (KMEM_CBRC_LATER);
291 if (mutex_tryenter(&vp->v_lock) == 0) {
292 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
293 mutex_exit(&zfsvfs->z_znodes_lock);
295 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
296 return (KMEM_CBRC_LATER);
299 /* Only move znodes that are referenced _only_ by the DNLC. */
300 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
301 mutex_exit(&vp->v_lock);
302 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
303 mutex_exit(&zfsvfs->z_znodes_lock);
305 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
306 return (KMEM_CBRC_LATER);
310 * The znode is known and in a valid state to move. We're holding the
311 * locks needed to execute the critical section.
313 zfs_znode_move_impl(ozp, nzp);
314 mutex_exit(&vp->v_lock);
315 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
317 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
318 mutex_exit(&zfsvfs->z_znodes_lock);
321 return (KMEM_CBRC_YES);
330 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
331 ASSERT(znode_cache == NULL);
332 znode_cache = kmem_cache_create("zfs_znode_cache",
333 sizeof (znode_t), 0, zfs_znode_cache_constructor,
334 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
335 kmem_cache_set_move(znode_cache, zfs_znode_move);
342 * Cleanup vfs & vnode ops
344 zfs_remove_op_tables();
350 kmem_cache_destroy(znode_cache);
352 rw_destroy(&zfsvfs_lock);
355 struct vnodeops *zfs_dvnodeops;
356 struct vnodeops *zfs_fvnodeops;
357 struct vnodeops *zfs_symvnodeops;
358 struct vnodeops *zfs_xdvnodeops;
359 struct vnodeops *zfs_evnodeops;
360 struct vnodeops *zfs_sharevnodeops;
363 zfs_remove_op_tables()
369 (void) vfs_freevfsops_by_type(zfsfstype);
376 vn_freevnodeops(zfs_dvnodeops);
378 vn_freevnodeops(zfs_fvnodeops);
380 vn_freevnodeops(zfs_symvnodeops);
382 vn_freevnodeops(zfs_xdvnodeops);
384 vn_freevnodeops(zfs_evnodeops);
385 if (zfs_sharevnodeops)
386 vn_freevnodeops(zfs_sharevnodeops);
388 zfs_dvnodeops = NULL;
389 zfs_fvnodeops = NULL;
390 zfs_symvnodeops = NULL;
391 zfs_xdvnodeops = NULL;
392 zfs_evnodeops = NULL;
393 zfs_sharevnodeops = NULL;
396 extern const fs_operation_def_t zfs_dvnodeops_template[];
397 extern const fs_operation_def_t zfs_fvnodeops_template[];
398 extern const fs_operation_def_t zfs_xdvnodeops_template[];
399 extern const fs_operation_def_t zfs_symvnodeops_template[];
400 extern const fs_operation_def_t zfs_evnodeops_template[];
401 extern const fs_operation_def_t zfs_sharevnodeops_template[];
404 zfs_create_op_tables()
409 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
410 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
411 * In this case we just return as the ops vectors are already set up.
416 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
421 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
426 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
431 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
436 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
441 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
448 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
450 zfs_acl_ids_t acl_ids;
457 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
458 vattr.va_type = VDIR;
459 vattr.va_mode = S_IFDIR|0555;
460 vattr.va_uid = crgetuid(kcred);
461 vattr.va_gid = crgetgid(kcred);
463 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
464 sharezp->z_unlinked = 0;
465 sharezp->z_atime_dirty = 0;
466 sharezp->z_zfsvfs = zfsvfs;
472 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
473 kcred, NULL, &acl_ids));
474 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE,
476 ASSERT3P(zp, ==, sharezp);
477 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
478 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
479 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
480 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
481 zfsvfs->z_shares_dir = sharezp->z_id;
483 zfs_acl_ids_free(&acl_ids);
484 ZTOV(sharezp)->v_count = 0;
485 dmu_buf_rele(sharezp->z_dbuf, NULL);
486 sharezp->z_dbuf = NULL;
487 kmem_cache_free(znode_cache, sharezp);
493 * define a couple of values we need available
494 * for both 64 and 32 bit environments.
497 #define NBITSMINOR64 32
500 #define MAXMAJ64 0xffffffffUL
503 #define MAXMIN64 0xffffffffUL
507 * Create special expldev for ZFS private use.
508 * Can't use standard expldev since it doesn't do
509 * what we want. The standard expldev() takes a
510 * dev32_t in LP64 and expands it to a long dev_t.
511 * We need an interface that takes a dev32_t in ILP32
512 * and expands it to a long dev_t.
515 zfs_expldev(dev_t dev)
518 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
519 return (((uint64_t)major << NBITSMINOR64) |
520 ((minor_t)dev & MAXMIN32));
527 * Special cmpldev for ZFS private use.
528 * Can't use standard cmpldev since it takes
529 * a long dev_t and compresses it to dev32_t in
530 * LP64. We need to do a compaction of a long dev_t
531 * to a dev32_t in ILP32.
534 zfs_cmpldev(uint64_t dev)
537 minor_t minor = (minor_t)dev & MAXMIN64;
538 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
540 if (major > MAXMAJ32 || minor > MAXMIN32)
543 return (((dev32_t)major << NBITSMINOR32) | minor);
550 zfs_znode_dmu_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db)
554 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
555 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
557 mutex_enter(&zp->z_lock);
559 ASSERT(zp->z_dbuf == NULL);
561 nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error);
565 * concurrent zgets on this object.
568 panic("existing znode %p for dbuf %p", (void *)nzp, (void *)db);
571 * Slap on VROOT if we are the root znode
573 if (zp->z_id == zfsvfs->z_root)
574 ZTOV(zp)->v_flag |= VROOT;
576 mutex_exit(&zp->z_lock);
581 zfs_znode_dmu_fini(znode_t *zp)
583 dmu_buf_t *db = zp->z_dbuf;
584 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
586 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
587 ASSERT(zp->z_dbuf != NULL);
589 VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL));
590 dmu_buf_rele(db, NULL);
594 * Construct a new znode/vnode and intialize.
596 * This does not do a call to dmu_set_user() that is
597 * up to the caller to do, in case you don't want to
601 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz)
606 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
608 ASSERT(zp->z_dirlocks == NULL);
609 ASSERT(zp->z_dbuf == NULL);
610 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
613 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
614 * the zfs_znode_move() callback.
618 zp->z_atime_dirty = 0;
621 zp->z_id = db->db_object;
623 zp->z_seq = 0x7A4653;
629 zfs_znode_dmu_init(zfsvfs, zp, db);
631 zp->z_gen = zp->z_phys->zp_gen;
633 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
634 vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode);
636 switch (vp->v_type) {
638 if (zp->z_phys->zp_flags & ZFS_XATTR) {
639 vn_setops(vp, zfs_xdvnodeops);
640 vp->v_flag |= V_XATTRDIR;
642 vn_setops(vp, zfs_dvnodeops);
644 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
648 vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev);
653 vn_setops(vp, zfs_fvnodeops);
656 vp->v_flag |= VMODSORT;
657 if (zp->z_phys->zp_parent == zfsvfs->z_shares_dir)
658 vn_setops(vp, zfs_sharevnodeops);
660 vn_setops(vp, zfs_fvnodeops);
663 vn_setops(vp, zfs_symvnodeops);
666 vn_setops(vp, zfs_evnodeops);
670 mutex_enter(&zfsvfs->z_znodes_lock);
671 list_insert_tail(&zfsvfs->z_all_znodes, zp);
674 * Everything else must be valid before assigning z_zfsvfs makes the
675 * znode eligible for zfs_znode_move().
677 zp->z_zfsvfs = zfsvfs;
678 mutex_exit(&zfsvfs->z_znodes_lock);
680 VFS_HOLD(zfsvfs->z_vfs);
685 * Create a new DMU object to hold a zfs znode.
687 * IN: dzp - parent directory for new znode
688 * vap - file attributes for new znode
689 * tx - dmu transaction id for zap operations
690 * cr - credentials of caller
692 * IS_ROOT_NODE - new object will be root
693 * IS_XATTR - new object is an attribute
694 * IS_REPLAY - intent log replay
695 * bonuslen - length of bonus buffer
696 * setaclp - File/Dir initial ACL
697 * fuidp - Tracks fuid allocation.
699 * OUT: zpp - allocated znode
703 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
704 uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_ids_t *acl_ids)
708 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
713 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
715 if (zfsvfs->z_replay) {
716 obj = vap->va_nodeid;
718 now = vap->va_ctime; /* see zfs_replay_create() */
719 gen = vap->va_nblocks; /* ditto */
723 gen = dmu_tx_get_txg(tx);
727 * Create a new DMU object.
730 * There's currently no mechanism for pre-reading the blocks that will
731 * be to needed allocate a new object, so we accept the small chance
732 * that there will be an i/o error and we will fail one of the
735 if (vap->va_type == VDIR) {
736 if (flag & IS_REPLAY) {
737 err = zap_create_claim_norm(zfsvfs->z_os, obj,
738 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
739 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
740 ASSERT3U(err, ==, 0);
742 obj = zap_create_norm(zfsvfs->z_os,
743 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
744 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
747 if (flag & IS_REPLAY) {
748 err = dmu_object_claim(zfsvfs->z_os, obj,
749 DMU_OT_PLAIN_FILE_CONTENTS, 0,
750 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
751 ASSERT3U(err, ==, 0);
753 obj = dmu_object_alloc(zfsvfs->z_os,
754 DMU_OT_PLAIN_FILE_CONTENTS, 0,
755 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
758 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db));
759 dmu_buf_will_dirty(db, tx);
762 * Initialize the znode physical data to zero.
764 ASSERT(db->db_size >= sizeof (znode_phys_t));
765 bzero(db->db_data, db->db_size);
769 * If this is the root, fix up the half-initialized parent pointer
770 * to reference the just-allocated physical data area.
772 if (flag & IS_ROOT_NODE) {
779 * If parent is an xattr, so am I.
781 if (dzp->z_phys->zp_flags & ZFS_XATTR)
784 if (vap->va_type == VBLK || vap->va_type == VCHR) {
785 pzp->zp_rdev = zfs_expldev(vap->va_rdev);
788 if (zfsvfs->z_use_fuids)
789 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
791 if (vap->va_type == VDIR) {
792 pzp->zp_size = 2; /* contents ("." and "..") */
793 pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
796 pzp->zp_parent = dzp->z_id;
798 pzp->zp_flags |= ZFS_XATTR;
802 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
803 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
805 if (vap->va_mask & AT_ATIME) {
806 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
808 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
811 if (vap->va_mask & AT_MTIME) {
812 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
814 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
817 pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode);
818 if (!(flag & IS_ROOT_NODE)) {
819 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
820 *zpp = zfs_znode_alloc(zfsvfs, db, 0);
821 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
824 * If we are creating the root node, the "parent" we
825 * passed in is the znode for the root.
829 pzp->zp_uid = acl_ids->z_fuid;
830 pzp->zp_gid = acl_ids->z_fgid;
831 pzp->zp_mode = acl_ids->z_mode;
832 VERIFY(0 == zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
833 if (vap->va_mask & AT_XVATTR)
834 zfs_xvattr_set(*zpp, (xvattr_t *)vap);
838 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap)
842 xoap = xva_getxoptattr(xvap);
845 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
846 ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime);
847 XVA_SET_RTN(xvap, XAT_CREATETIME);
849 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
850 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly);
851 XVA_SET_RTN(xvap, XAT_READONLY);
853 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
854 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden);
855 XVA_SET_RTN(xvap, XAT_HIDDEN);
857 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
858 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system);
859 XVA_SET_RTN(xvap, XAT_SYSTEM);
861 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
862 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive);
863 XVA_SET_RTN(xvap, XAT_ARCHIVE);
865 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
866 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable);
867 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
869 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
870 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink);
871 XVA_SET_RTN(xvap, XAT_NOUNLINK);
873 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
874 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly);
875 XVA_SET_RTN(xvap, XAT_APPENDONLY);
877 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
878 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump);
879 XVA_SET_RTN(xvap, XAT_NODUMP);
881 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
882 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque);
883 XVA_SET_RTN(xvap, XAT_OPAQUE);
885 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
886 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
887 xoap->xoa_av_quarantined);
888 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
890 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
891 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified);
892 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
894 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
895 (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp,
896 sizeof (xoap->xoa_av_scanstamp));
897 zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP;
898 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
903 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
905 dmu_object_info_t doi;
912 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
914 err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
916 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
920 dmu_object_info_from_db(db, &doi);
921 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
922 doi.doi_bonus_size < sizeof (znode_phys_t)) {
923 dmu_buf_rele(db, NULL);
924 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
928 zp = dmu_buf_get_user(db);
930 mutex_enter(&zp->z_lock);
933 * Since we do immediate eviction of the z_dbuf, we
934 * should never find a dbuf with a znode that doesn't
935 * know about the dbuf.
937 ASSERT3P(zp->z_dbuf, ==, db);
938 ASSERT3U(zp->z_id, ==, obj_num);
939 if (zp->z_unlinked) {
946 dmu_buf_rele(db, NULL);
947 mutex_exit(&zp->z_lock);
948 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
953 * Not found create new znode/vnode
955 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size);
956 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
962 zfs_rezget(znode_t *zp)
964 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
965 dmu_object_info_t doi;
967 uint64_t obj_num = zp->z_id;
970 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
972 err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
974 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
978 dmu_object_info_from_db(db, &doi);
979 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
980 doi.doi_bonus_size < sizeof (znode_phys_t)) {
981 dmu_buf_rele(db, NULL);
982 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
986 if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) {
987 dmu_buf_rele(db, NULL);
988 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
992 zfs_znode_dmu_init(zfsvfs, zp, db);
993 zp->z_unlinked = (zp->z_phys->zp_links == 0);
994 zp->z_blksz = doi.doi_data_block_size;
996 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1002 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1004 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1005 objset_t *os = zfsvfs->z_os;
1006 uint64_t obj = zp->z_id;
1007 uint64_t acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
1009 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1011 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1012 VERIFY(0 == dmu_object_free(os, obj, tx));
1013 zfs_znode_dmu_fini(zp);
1014 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1019 zfs_zinactive(znode_t *zp)
1021 vnode_t *vp = ZTOV(zp);
1022 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1023 uint64_t z_id = zp->z_id;
1025 ASSERT(zp->z_dbuf && zp->z_phys);
1028 * Don't allow a zfs_zget() while were trying to release this znode
1030 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1032 mutex_enter(&zp->z_lock);
1033 mutex_enter(&vp->v_lock);
1035 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1037 * If the hold count is greater than zero, somebody has
1038 * obtained a new reference on this znode while we were
1039 * processing it here, so we are done. If we still have
1040 * mapped pages then we are also done, since we don't
1041 * want to inactivate the znode until the pages get pushed.
1043 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1044 * this seems like it would leave the znode hanging with
1045 * no chance to go inactive...
1047 mutex_exit(&vp->v_lock);
1048 mutex_exit(&zp->z_lock);
1049 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1052 mutex_exit(&vp->v_lock);
1055 * If this was the last reference to a file with no links,
1056 * remove the file from the file system.
1058 if (zp->z_unlinked) {
1059 mutex_exit(&zp->z_lock);
1060 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1064 mutex_exit(&zp->z_lock);
1065 zfs_znode_dmu_fini(zp);
1066 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1071 zfs_znode_free(znode_t *zp)
1073 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1075 vn_invalid(ZTOV(zp));
1077 ASSERT(ZTOV(zp)->v_count == 0);
1079 mutex_enter(&zfsvfs->z_znodes_lock);
1080 POINTER_INVALIDATE(&zp->z_zfsvfs);
1081 list_remove(&zfsvfs->z_all_znodes, zp);
1082 mutex_exit(&zfsvfs->z_znodes_lock);
1084 kmem_cache_free(znode_cache, zp);
1086 VFS_RELE(zfsvfs->z_vfs);
1090 zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx)
1094 ASSERT(MUTEX_HELD(&zp->z_lock));
1099 dmu_buf_will_dirty(zp->z_dbuf, tx);
1100 zp->z_atime_dirty = 0;
1103 zp->z_atime_dirty = 1;
1106 if (flag & AT_ATIME)
1107 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime);
1109 if (flag & AT_MTIME) {
1110 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime);
1111 if (zp->z_zfsvfs->z_use_fuids)
1112 zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED);
1115 if (flag & AT_CTIME) {
1116 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime);
1117 if (zp->z_zfsvfs->z_use_fuids)
1118 zp->z_phys->zp_flags |= ZFS_ARCHIVE;
1123 * Update the requested znode timestamps with the current time.
1124 * If we are in a transaction, then go ahead and mark the znode
1125 * dirty in the transaction so the timestamps will go to disk.
1126 * Otherwise, we will get pushed next time the znode is updated
1127 * in a transaction, or when this znode eventually goes inactive.
1130 * 1 - Only the ACCESS time is ever updated outside of a transaction.
1131 * 2 - Multiple consecutive updates will be collapsed into a single
1132 * znode update by the transaction grouping semantics of the DMU.
1135 zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx)
1137 mutex_enter(&zp->z_lock);
1138 zfs_time_stamper_locked(zp, flag, tx);
1139 mutex_exit(&zp->z_lock);
1143 * Grow the block size for a file.
1145 * IN: zp - znode of file to free data in.
1146 * size - requested block size
1147 * tx - open transaction.
1149 * NOTE: this function assumes that the znode is write locked.
1152 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1157 if (size <= zp->z_blksz)
1160 * If the file size is already greater than the current blocksize,
1161 * we will not grow. If there is more than one block in a file,
1162 * the blocksize cannot change.
1164 if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz)
1167 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1169 if (error == ENOTSUP)
1171 ASSERT3U(error, ==, 0);
1173 /* What blocksize did we actually get? */
1174 dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy);
1178 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1179 * be calling back into the fs for a putpage(). E.g.: when truncating
1180 * a file, the pages being "thrown away* don't need to be written out.
1184 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1185 int flags, cred_t *cr)
1192 * Increase the file length
1194 * IN: zp - znode of file to free data in.
1195 * end - new end-of-file
1197 * RETURN: 0 if success
1198 * error code if failure
1201 zfs_extend(znode_t *zp, uint64_t end)
1203 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1210 * We will change zp_size, lock the whole file.
1212 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1215 * Nothing to do if file already at desired length.
1217 if (end <= zp->z_phys->zp_size) {
1218 zfs_range_unlock(rl);
1222 tx = dmu_tx_create(zfsvfs->z_os);
1223 dmu_tx_hold_bonus(tx, zp->z_id);
1224 if (end > zp->z_blksz &&
1225 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1227 * We are growing the file past the current block size.
1229 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1230 ASSERT(!ISP2(zp->z_blksz));
1231 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1233 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1235 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1240 error = dmu_tx_assign(tx, TXG_NOWAIT);
1242 if (error == ERESTART) {
1248 zfs_range_unlock(rl);
1251 dmu_buf_will_dirty(zp->z_dbuf, tx);
1254 zfs_grow_blocksize(zp, newblksz, tx);
1256 zp->z_phys->zp_size = end;
1258 zfs_range_unlock(rl);
1266 * Free space in a file.
1268 * IN: zp - znode of file to free data in.
1269 * off - start of section to free.
1270 * len - length of section to free.
1272 * RETURN: 0 if success
1273 * error code if failure
1276 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1278 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1283 * Lock the range being freed.
1285 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1288 * Nothing to do if file already at desired length.
1290 if (off >= zp->z_phys->zp_size) {
1291 zfs_range_unlock(rl);
1295 if (off + len > zp->z_phys->zp_size)
1296 len = zp->z_phys->zp_size - off;
1298 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1300 zfs_range_unlock(rl);
1308 * IN: zp - znode of file to free data in.
1309 * end - new end-of-file.
1311 * RETURN: 0 if success
1312 * error code if failure
1315 zfs_trunc(znode_t *zp, uint64_t end)
1317 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1318 vnode_t *vp = ZTOV(zp);
1324 * We will change zp_size, lock the whole file.
1326 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1329 * Nothing to do if file already at desired length.
1331 if (end >= zp->z_phys->zp_size) {
1332 zfs_range_unlock(rl);
1336 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1338 zfs_range_unlock(rl);
1342 tx = dmu_tx_create(zfsvfs->z_os);
1343 dmu_tx_hold_bonus(tx, zp->z_id);
1344 error = dmu_tx_assign(tx, TXG_NOWAIT);
1346 if (error == ERESTART) {
1352 zfs_range_unlock(rl);
1355 dmu_buf_will_dirty(zp->z_dbuf, tx);
1357 zp->z_phys->zp_size = end;
1362 * Clear any mapped pages in the truncated region. This has to
1363 * happen outside of the transaction to avoid the possibility of
1364 * a deadlock with someone trying to push a page that we are
1365 * about to invalidate.
1367 if (vn_has_cached_data(vp)) {
1369 uint64_t start = end & PAGEMASK;
1370 int poff = end & PAGEOFFSET;
1372 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1374 * We need to zero a partial page.
1376 pagezero(pp, poff, PAGESIZE - poff);
1380 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1381 B_INVAL | B_TRUNC, NULL);
1385 zfs_range_unlock(rl);
1391 * Free space in a file
1393 * IN: zp - znode of file to free data in.
1394 * off - start of range
1395 * len - end of range (0 => EOF)
1396 * flag - current file open mode flags.
1397 * log - TRUE if this action should be logged
1399 * RETURN: 0 if success
1400 * error code if failure
1403 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1405 vnode_t *vp = ZTOV(zp);
1407 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1408 zilog_t *zilog = zfsvfs->z_log;
1411 if (off > zp->z_phys->zp_size) {
1412 error = zfs_extend(zp, off+len);
1413 if (error == 0 && log)
1420 * Check for any locks in the region to be freed.
1422 if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) {
1423 uint64_t length = (len ? len : zp->z_phys->zp_size - off);
1424 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1429 error = zfs_trunc(zp, off);
1431 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1432 off + len > zp->z_phys->zp_size)
1433 error = zfs_extend(zp, off+len);
1438 tx = dmu_tx_create(zfsvfs->z_os);
1439 dmu_tx_hold_bonus(tx, zp->z_id);
1440 error = dmu_tx_assign(tx, TXG_NOWAIT);
1442 if (error == ERESTART) {
1451 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
1452 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1459 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1462 uint64_t moid, obj, version;
1463 uint64_t sense = ZFS_CASE_SENSITIVE;
1467 znode_t *rootzp = NULL;
1471 zfs_acl_ids_t acl_ids;
1474 * First attempt to create master node.
1477 * In an empty objset, there are no blocks to read and thus
1478 * there can be no i/o errors (which we assert below).
1480 moid = MASTER_NODE_OBJ;
1481 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1482 DMU_OT_NONE, 0, tx);
1486 * Set starting attributes.
1488 if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_USERSPACE)
1489 version = ZPL_VERSION;
1490 else if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
1491 version = ZPL_VERSION_USERSPACE - 1;
1493 version = ZPL_VERSION_FUID - 1;
1495 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1496 /* For the moment we expect all zpl props to be uint64_ts */
1500 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1501 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1502 name = nvpair_name(elem);
1503 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1507 error = zap_update(os, moid, name, 8, 1, &val, tx);
1510 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1512 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1515 ASSERT(version != 0);
1516 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1519 * Create a delete queue.
1521 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1523 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1527 * Create root znode. Create minimal znode/vnode/zfsvfs
1528 * to allow zfs_mknode to work.
1530 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1531 vattr.va_type = VDIR;
1532 vattr.va_mode = S_IFDIR|0755;
1533 vattr.va_uid = crgetuid(cr);
1534 vattr.va_gid = crgetgid(cr);
1536 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1537 rootzp->z_unlinked = 0;
1538 rootzp->z_atime_dirty = 0;
1544 bzero(&zfsvfs, sizeof (zfsvfs_t));
1547 zfsvfs.z_parent = &zfsvfs;
1548 zfsvfs.z_version = version;
1549 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1550 zfsvfs.z_norm = norm;
1552 * Fold case on file systems that are always or sometimes case
1555 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1556 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1558 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1559 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1560 offsetof(znode_t, z_link_node));
1562 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1563 rootzp->z_zfsvfs = &zfsvfs;
1564 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1565 cr, NULL, &acl_ids));
1566 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, &acl_ids);
1567 ASSERT3P(zp, ==, rootzp);
1568 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1569 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1571 zfs_acl_ids_free(&acl_ids);
1572 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1574 ZTOV(rootzp)->v_count = 0;
1575 dmu_buf_rele(rootzp->z_dbuf, NULL);
1576 rootzp->z_dbuf = NULL;
1577 kmem_cache_free(znode_cache, rootzp);
1580 * Create shares directory
1583 error = zfs_create_share_dir(&zfsvfs, tx);
1588 #endif /* _KERNEL */
1590 * Given an object number, return its parent object number and whether
1591 * or not the object is an extended attribute directory.
1594 zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir)
1597 dmu_object_info_t doi;
1601 if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0)
1604 dmu_object_info_from_db(db, &doi);
1605 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
1606 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1607 dmu_buf_rele(db, FTAG);
1612 *pobjp = zp->zp_parent;
1613 *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) &&
1614 S_ISDIR(zp->zp_mode);
1615 dmu_buf_rele(db, FTAG);
1621 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1623 char *path = buf + len - 1;
1630 char component[MAXNAMELEN + 2];
1634 if ((error = zfs_obj_to_pobj(osp, obj, &pobj,
1635 &is_xattrdir)) != 0)
1646 (void) sprintf(component + 1, "<xattrdir>");
1648 error = zap_value_search(osp, pobj, obj,
1649 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1654 complen = strlen(component);
1656 ASSERT(path >= buf);
1657 bcopy(component, path, complen);
1662 (void) memmove(buf, path, buf + len - path);