4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
24 * Copyright (c) 2014 Integros [integros.com]
27 /* Portions Copyright 2007 Jeremy Teo */
28 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
31 #include <sys/types.h>
32 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/sysmacros.h>
36 #include <sys/resource.h>
37 #include <sys/mntent.h>
38 #include <sys/u8_textprep.h>
39 #include <sys/dsl_dataset.h>
41 #include <sys/vnode.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/atomic.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/zfs_ioctl.h>
50 #include <sys/zfs_rlock.h>
51 #include <sys/zfs_fuid.h>
52 #include <sys/dnode.h>
53 #include <sys/fs/zfs.h>
54 #include <sys/kidmap.h>
58 #include <sys/dmu_objset.h>
59 #include <sys/dmu_tx.h>
60 #include <sys/refcount.h>
63 #include <sys/zfs_znode.h>
65 #include <sys/zfs_sa.h>
66 #include <sys/zfs_stat.h>
67 #include <sys/refcount.h>
70 #include "zfs_comutil.h"
72 /* Used by fstat(1). */
73 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
74 SYSCTL_NULL_INT_PTR, sizeof(znode_t), "sizeof(znode_t)");
77 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
78 * turned on when DEBUG is also defined.
85 #define ZNODE_STAT_ADD(stat) ((stat)++)
87 #define ZNODE_STAT_ADD(stat) /* nothing */
88 #endif /* ZNODE_STATS */
91 * Functions needed for userland (ie: libzpool) are not put under
92 * #ifdef_KERNEL; the rest of the functions have dependencies
93 * (such as VFS logic) that will not compile easily in userland.
97 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
98 * be freed before it can be safely accessed.
100 krwlock_t zfsvfs_lock;
102 #if defined(_KERNEL) && !defined(KMEM_DEBUG)
104 static uma_zone_t znode_uma_zone;
106 static kmem_cache_t *znode_cache = NULL;
111 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
114 * We should never drop all dbuf refs without first clearing
115 * the eviction callback.
117 panic("evicting znode %p\n", user_ptr);
120 extern struct vop_vector zfs_vnodeops;
121 extern struct vop_vector zfs_fifoops;
122 extern struct vop_vector zfs_shareops;
125 * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on
126 * z_rangelock. It will modify the offset and length of the lock to reflect
127 * znode-specific information, and convert RL_APPEND to RL_WRITER. This is
128 * called with the rangelock_t's rl_lock held, which avoids races.
131 zfs_rangelock_cb(locked_range_t *new, void *arg)
136 * If in append mode, convert to writer and lock starting at the
137 * current end of file.
139 if (new->lr_type == RL_APPEND) {
140 new->lr_offset = zp->z_size;
141 new->lr_type = RL_WRITER;
145 * If we need to grow the block size then lock the whole file range.
147 uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length);
148 if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
149 zp->z_blksz < zp->z_zfsvfs->z_max_blksz)) {
151 new->lr_length = UINT64_MAX;
157 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
161 POINTER_INVALIDATE(&zp->z_zfsvfs);
163 list_link_init(&zp->z_link_node);
165 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
167 rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp);
169 zp->z_acl_cached = NULL;
177 zfs_znode_cache_destructor(void *buf, void *arg)
181 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
182 ASSERT3P(zp->z_vnode, ==, NULL);
183 ASSERT(!list_link_active(&zp->z_link_node));
184 mutex_destroy(&zp->z_acl_lock);
185 rangelock_fini(&zp->z_rangelock);
187 ASSERT(zp->z_acl_cached == NULL);
192 uint64_t zms_zfsvfs_invalid;
193 uint64_t zms_zfsvfs_recheck1;
194 uint64_t zms_zfsvfs_unmounted;
195 uint64_t zms_zfsvfs_recheck2;
196 uint64_t zms_obj_held;
197 uint64_t zms_vnode_locked;
198 uint64_t zms_not_only_dnlc;
200 #endif /* ZNODE_STATS */
204 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
209 nzp->z_zfsvfs = ozp->z_zfsvfs;
213 nzp->z_vnode = ozp->z_vnode;
214 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
215 ZTOV(ozp)->v_data = ozp;
216 ZTOV(nzp)->v_data = nzp;
218 nzp->z_id = ozp->z_id;
219 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
220 nzp->z_unlinked = ozp->z_unlinked;
221 nzp->z_atime_dirty = ozp->z_atime_dirty;
222 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
223 nzp->z_blksz = ozp->z_blksz;
224 nzp->z_seq = ozp->z_seq;
225 nzp->z_mapcnt = ozp->z_mapcnt;
226 nzp->z_gen = ozp->z_gen;
227 nzp->z_sync_cnt = ozp->z_sync_cnt;
228 nzp->z_is_sa = ozp->z_is_sa;
229 nzp->z_sa_hdl = ozp->z_sa_hdl;
230 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
231 nzp->z_links = ozp->z_links;
232 nzp->z_size = ozp->z_size;
233 nzp->z_pflags = ozp->z_pflags;
234 nzp->z_uid = ozp->z_uid;
235 nzp->z_gid = ozp->z_gid;
236 nzp->z_mode = ozp->z_mode;
239 * Since this is just an idle znode and kmem is already dealing with
240 * memory pressure, release any cached ACL.
242 if (ozp->z_acl_cached) {
243 zfs_acl_free(ozp->z_acl_cached);
244 ozp->z_acl_cached = NULL;
247 sa_set_userp(nzp->z_sa_hdl, nzp);
250 * Invalidate the original znode by clearing fields that provide a
251 * pointer back to the znode. Set the low bit of the vfs pointer to
252 * ensure that zfs_znode_move() recognizes the znode as invalid in any
253 * subsequent callback.
255 ozp->z_sa_hdl = NULL;
256 POINTER_INVALIDATE(&ozp->z_zfsvfs);
262 ozp->z_moved = (uint8_t)-1;
267 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
269 znode_t *ozp = buf, *nzp = newbuf;
274 * The znode is on the file system's list of known znodes if the vfs
275 * pointer is valid. We set the low bit of the vfs pointer when freeing
276 * the znode to invalidate it, and the memory patterns written by kmem
277 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
278 * created znode sets the vfs pointer last of all to indicate that the
279 * znode is known and in a valid state to be moved by this function.
281 zfsvfs = ozp->z_zfsvfs;
282 if (!POINTER_IS_VALID(zfsvfs)) {
283 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
284 return (KMEM_CBRC_DONT_KNOW);
288 * Close a small window in which it's possible that the filesystem could
289 * be unmounted and freed, and zfsvfs, though valid in the previous
290 * statement, could point to unrelated memory by the time we try to
291 * prevent the filesystem from being unmounted.
293 rw_enter(&zfsvfs_lock, RW_WRITER);
294 if (zfsvfs != ozp->z_zfsvfs) {
295 rw_exit(&zfsvfs_lock);
296 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
297 return (KMEM_CBRC_DONT_KNOW);
301 * If the znode is still valid, then so is the file system. We know that
302 * no valid file system can be freed while we hold zfsvfs_lock, so we
303 * can safely ensure that the filesystem is not and will not be
304 * unmounted. The next statement is equivalent to ZFS_ENTER().
306 rrm_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
307 if (zfsvfs->z_unmounted) {
309 rw_exit(&zfsvfs_lock);
310 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
311 return (KMEM_CBRC_DONT_KNOW);
313 rw_exit(&zfsvfs_lock);
315 mutex_enter(&zfsvfs->z_znodes_lock);
317 * Recheck the vfs pointer in case the znode was removed just before
318 * acquiring the lock.
320 if (zfsvfs != ozp->z_zfsvfs) {
321 mutex_exit(&zfsvfs->z_znodes_lock);
323 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
324 return (KMEM_CBRC_DONT_KNOW);
328 * At this point we know that as long as we hold z_znodes_lock, the
329 * znode cannot be freed and fields within the znode can be safely
330 * accessed. Now, prevent a race with zfs_zget().
332 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
333 mutex_exit(&zfsvfs->z_znodes_lock);
335 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
336 return (KMEM_CBRC_LATER);
340 if (mutex_tryenter(&vp->v_lock) == 0) {
341 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
342 mutex_exit(&zfsvfs->z_znodes_lock);
344 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
345 return (KMEM_CBRC_LATER);
348 /* Only move znodes that are referenced _only_ by the DNLC. */
349 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
350 mutex_exit(&vp->v_lock);
351 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
352 mutex_exit(&zfsvfs->z_znodes_lock);
354 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
355 return (KMEM_CBRC_LATER);
359 * The znode is known and in a valid state to move. We're holding the
360 * locks needed to execute the critical section.
362 zfs_znode_move_impl(ozp, nzp);
363 mutex_exit(&vp->v_lock);
364 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
366 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
367 mutex_exit(&zfsvfs->z_znodes_lock);
370 return (KMEM_CBRC_YES);
378 zfs_znode_cache_constructor_smr(void *mem, int size __unused, void *private, int flags)
381 return (zfs_znode_cache_constructor(mem, private, flags));
385 zfs_znode_cache_destructor_smr(void *mem, int size __unused, void *private)
388 zfs_znode_cache_destructor(mem, private);
397 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
398 ASSERT(znode_uma_zone == NULL);
399 znode_uma_zone = uma_zcreate("zfs_znode_cache",
400 sizeof (znode_t), zfs_znode_cache_constructor_smr,
401 zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0);
402 VFS_SMR_ZONE_SET(znode_uma_zone);
406 zfs_znode_alloc_kmem(int flags)
409 return (uma_zalloc_smr(znode_uma_zone, flags));
413 zfs_znode_free_kmem(znode_t *zp)
416 uma_zfree_smr(znode_uma_zone, zp);
425 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
426 ASSERT(znode_cache == NULL);
427 znode_cache = kmem_cache_create("zfs_znode_cache",
428 sizeof (znode_t), 0, zfs_znode_cache_constructor,
429 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
430 kmem_cache_set_move(znode_cache, zfs_znode_move);
434 zfs_znode_alloc_kmem(int flags)
437 return (kmem_cache_alloc(znode_cache, flags));
441 zfs_znode_free_kmem(znode_t *zp)
444 kmem_cache_free(znode_cache, zp);
453 * Cleanup vfs & vnode ops
455 zfs_remove_op_tables();
462 if (znode_uma_zone) {
463 uma_zdestroy(znode_uma_zone);
464 znode_uma_zone = NULL;
468 kmem_cache_destroy(znode_cache);
472 rw_destroy(&zfsvfs_lock);
476 struct vnodeops *zfs_dvnodeops;
477 struct vnodeops *zfs_fvnodeops;
478 struct vnodeops *zfs_symvnodeops;
479 struct vnodeops *zfs_xdvnodeops;
480 struct vnodeops *zfs_evnodeops;
481 struct vnodeops *zfs_sharevnodeops;
484 zfs_remove_op_tables()
490 (void) vfs_freevfsops_by_type(zfsfstype);
497 vn_freevnodeops(zfs_dvnodeops);
499 vn_freevnodeops(zfs_fvnodeops);
501 vn_freevnodeops(zfs_symvnodeops);
503 vn_freevnodeops(zfs_xdvnodeops);
505 vn_freevnodeops(zfs_evnodeops);
506 if (zfs_sharevnodeops)
507 vn_freevnodeops(zfs_sharevnodeops);
509 zfs_dvnodeops = NULL;
510 zfs_fvnodeops = NULL;
511 zfs_symvnodeops = NULL;
512 zfs_xdvnodeops = NULL;
513 zfs_evnodeops = NULL;
514 zfs_sharevnodeops = NULL;
517 extern const fs_operation_def_t zfs_dvnodeops_template[];
518 extern const fs_operation_def_t zfs_fvnodeops_template[];
519 extern const fs_operation_def_t zfs_xdvnodeops_template[];
520 extern const fs_operation_def_t zfs_symvnodeops_template[];
521 extern const fs_operation_def_t zfs_evnodeops_template[];
522 extern const fs_operation_def_t zfs_sharevnodeops_template[];
525 zfs_create_op_tables()
530 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
531 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
532 * In this case we just return as the ops vectors are already set up.
537 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
542 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
547 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
552 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
557 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
562 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
570 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
572 zfs_acl_ids_t acl_ids;
578 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
579 vattr.va_type = VDIR;
580 vattr.va_mode = S_IFDIR|0555;
581 vattr.va_uid = crgetuid(kcred);
582 vattr.va_gid = crgetgid(kcred);
584 sharezp = zfs_znode_alloc_kmem(KM_SLEEP);
585 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
586 sharezp->z_moved = 0;
587 sharezp->z_unlinked = 0;
588 sharezp->z_atime_dirty = 0;
589 sharezp->z_zfsvfs = zfsvfs;
590 sharezp->z_is_sa = zfsvfs->z_use_sa;
592 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
593 kcred, NULL, &acl_ids));
594 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
595 ASSERT3P(zp, ==, sharezp);
596 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
597 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
598 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
599 zfsvfs->z_shares_dir = sharezp->z_id;
601 zfs_acl_ids_free(&acl_ids);
602 sa_handle_destroy(sharezp->z_sa_hdl);
603 zfs_znode_free_kmem(sharezp);
609 * define a couple of values we need available
610 * for both 64 and 32 bit environments.
613 #define NBITSMINOR64 32
616 #define MAXMAJ64 0xffffffffUL
619 #define MAXMIN64 0xffffffffUL
623 * Create special expldev for ZFS private use.
624 * Can't use standard expldev since it doesn't do
625 * what we want. The standard expldev() takes a
626 * dev32_t in LP64 and expands it to a long dev_t.
627 * We need an interface that takes a dev32_t in ILP32
628 * and expands it to a long dev_t.
631 zfs_expldev(dev_t dev)
633 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
636 * Special cmpldev for ZFS private use.
637 * Can't use standard cmpldev since it takes
638 * a long dev_t and compresses it to dev32_t in
639 * LP64. We need to do a compaction of a long dev_t
640 * to a dev32_t in ILP32.
643 zfs_cmpldev(uint64_t dev)
645 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
649 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
650 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
652 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
653 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
655 ASSERT(zp->z_sa_hdl == NULL);
656 ASSERT(zp->z_acl_cached == NULL);
657 if (sa_hdl == NULL) {
658 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
659 SA_HDL_SHARED, &zp->z_sa_hdl));
661 zp->z_sa_hdl = sa_hdl;
662 sa_set_userp(sa_hdl, zp);
665 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
668 * Slap on VROOT if we are the root znode unless we are the root
669 * node of a snapshot mounted under .zfs.
671 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
672 ZTOV(zp)->v_flag |= VROOT;
678 zfs_znode_dmu_fini(znode_t *zp)
680 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
682 ZFS_TEARDOWN_INACTIVE_WLOCKED(zp->z_zfsvfs));
684 sa_handle_destroy(zp->z_sa_hdl);
689 zfs_vnode_forget(vnode_t *vp)
692 /* copied from insmntque_stddtr */
694 vp->v_op = &dead_vnodeops;
700 * Construct a new znode/vnode and intialize.
702 * This does not do a call to dmu_set_user() that is
703 * up to the caller to do, in case you don't want to
707 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
708 dmu_object_type_t obj_type, sa_handle_t *hdl)
714 sa_bulk_attr_t bulk[9];
718 zp = zfs_znode_alloc_kmem(KM_SLEEP);
721 KASSERT((zfsvfs->z_parent->z_vfs->mnt_kern_flag & MNTK_FPLOOKUP) == 0,
722 ("%s: fast path lookup enabled without smr", __func__));
725 KASSERT(curthread->td_vp_reserved != NULL,
726 ("zfs_znode_alloc: getnewvnode without preallocated vnode"));
727 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
729 zfs_znode_free_kmem(zp);
735 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
739 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
740 * the zfs_znode_move() callback.
744 zp->z_atime_dirty = 0;
746 zp->z_id = db->db_object;
748 zp->z_seq = 0x7A4653;
753 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
755 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
756 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
757 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
759 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
761 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
763 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
764 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
766 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
768 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
771 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
773 sa_handle_destroy(zp->z_sa_hdl);
774 zfs_vnode_forget(vp);
776 zfs_znode_free_kmem(zp);
782 vp->v_type = IFTOVT((mode_t)mode);
784 switch (vp->v_type) {
786 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
793 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
794 &rdev, sizeof (rdev)) == 0);
796 vp->v_rdev = zfs_cmpldev(rdev);
805 vp->v_op = &zfs_fifoops;
808 if (parent == zfsvfs->z_shares_dir) {
809 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
810 vp->v_op = &zfs_shareops;
815 vn_setops(vp, zfs_symvnodeops);
818 vn_setops(vp, zfs_evnodeops);
823 mutex_enter(&zfsvfs->z_znodes_lock);
824 list_insert_tail(&zfsvfs->z_all_znodes, zp);
827 * Everything else must be valid before assigning z_zfsvfs makes the
828 * znode eligible for zfs_znode_move().
830 zp->z_zfsvfs = zfsvfs;
831 mutex_exit(&zfsvfs->z_znodes_lock);
834 * Acquire vnode lock before making it available to the world.
836 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
838 if (vp->v_type != VFIFO)
842 VFS_HOLD(zfsvfs->z_vfs);
847 static uint64_t empty_xattr;
848 static uint64_t pad[4];
849 static zfs_acl_phys_t acl_phys;
851 * Create a new DMU object to hold a zfs znode.
853 * IN: dzp - parent directory for new znode
854 * vap - file attributes for new znode
855 * tx - dmu transaction id for zap operations
856 * cr - credentials of caller
858 * IS_ROOT_NODE - new object will be root
859 * IS_XATTR - new object is an attribute
860 * bonuslen - length of bonus buffer
861 * setaclp - File/Dir initial ACL
862 * fuidp - Tracks fuid allocation.
864 * OUT: zpp - allocated znode
868 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
869 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
871 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
872 uint64_t mode, size, links, parent, pflags;
873 uint64_t dzp_pflags = 0;
875 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
883 dmu_object_type_t obj_type;
884 sa_bulk_attr_t *sa_attrs;
886 zfs_acl_locator_cb_t locate = { 0 };
888 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
890 if (zfsvfs->z_replay) {
891 obj = vap->va_nodeid;
892 now = vap->va_ctime; /* see zfs_replay_create() */
893 gen = vap->va_nblocks; /* ditto */
894 dnodesize = vap->va_fsid; /* ditto */
898 gen = dmu_tx_get_txg(tx);
899 dnodesize = dmu_objset_dnodesize(zfsvfs->z_os);
903 dnodesize = DNODE_MIN_SIZE;
905 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
906 bonuslen = (obj_type == DMU_OT_SA) ?
907 DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE;
910 * Create a new DMU object.
913 * There's currently no mechanism for pre-reading the blocks that will
914 * be needed to allocate a new object, so we accept the small chance
915 * that there will be an i/o error and we will fail one of the
918 if (vap->va_type == VDIR) {
919 if (zfsvfs->z_replay) {
920 VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj,
921 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
922 obj_type, bonuslen, dnodesize, tx));
924 obj = zap_create_norm_dnsize(zfsvfs->z_os,
925 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
926 obj_type, bonuslen, dnodesize, tx);
929 if (zfsvfs->z_replay) {
930 VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj,
931 DMU_OT_PLAIN_FILE_CONTENTS, 0,
932 obj_type, bonuslen, dnodesize, tx));
934 obj = dmu_object_alloc_dnsize(zfsvfs->z_os,
935 DMU_OT_PLAIN_FILE_CONTENTS, 0,
936 obj_type, bonuslen, dnodesize, tx);
940 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
941 VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
944 * If this is the root, fix up the half-initialized parent pointer
945 * to reference the just-allocated physical data area.
947 if (flag & IS_ROOT_NODE) {
950 dzp_pflags = dzp->z_pflags;
954 * If parent is an xattr, so am I.
956 if (dzp_pflags & ZFS_XATTR) {
960 if (zfsvfs->z_use_fuids)
961 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
965 if (vap->va_type == VDIR) {
966 size = 2; /* contents ("." and "..") */
967 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
972 if (vap->va_type == VBLK || vap->va_type == VCHR) {
973 rdev = zfs_expldev(vap->va_rdev);
977 mode = acl_ids->z_mode;
982 * No execs denied will be deterimed when zfs_mode_compute() is called.
984 pflags |= acl_ids->z_aclp->z_hints &
985 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
986 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
988 ZFS_TIME_ENCODE(&now, crtime);
989 ZFS_TIME_ENCODE(&now, ctime);
991 if (vap->va_mask & AT_ATIME) {
992 ZFS_TIME_ENCODE(&vap->va_atime, atime);
994 ZFS_TIME_ENCODE(&now, atime);
997 if (vap->va_mask & AT_MTIME) {
998 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
1000 ZFS_TIME_ENCODE(&now, mtime);
1003 /* Now add in all of the "SA" attributes */
1004 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
1008 * Setup the array of attributes to be replaced/set on the new file
1010 * order for DMU_OT_ZNODE is critical since it needs to be constructed
1011 * in the old znode_phys_t format. Don't change this ordering
1013 sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
1015 if (obj_type == DMU_OT_ZNODE) {
1016 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
1018 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
1020 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
1022 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
1024 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
1026 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
1028 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
1030 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
1033 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
1035 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
1037 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
1039 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs),
1040 NULL, &acl_ids->z_fuid, 8);
1041 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs),
1042 NULL, &acl_ids->z_fgid, 8);
1043 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
1045 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
1047 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
1049 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
1051 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
1053 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
1057 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
1059 if (obj_type == DMU_OT_ZNODE) {
1060 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
1063 if (obj_type == DMU_OT_ZNODE ||
1064 (vap->va_type == VBLK || vap->va_type == VCHR)) {
1065 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
1069 if (obj_type == DMU_OT_ZNODE) {
1070 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
1072 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
1073 &acl_ids->z_fuid, 8);
1074 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
1075 &acl_ids->z_fgid, 8);
1076 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
1077 sizeof (uint64_t) * 4);
1078 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1079 &acl_phys, sizeof (zfs_acl_phys_t));
1080 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
1081 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
1082 &acl_ids->z_aclp->z_acl_count, 8);
1083 locate.cb_aclp = acl_ids->z_aclp;
1084 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
1085 zfs_acl_data_locator, &locate,
1086 acl_ids->z_aclp->z_acl_bytes);
1087 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
1088 acl_ids->z_fuid, acl_ids->z_fgid);
1091 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
1093 if (!(flag & IS_ROOT_NODE)) {
1094 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
1095 ASSERT(*zpp != NULL);
1098 * If we are creating the root node, the "parent" we
1099 * passed in is the znode for the root.
1103 (*zpp)->z_sa_hdl = sa_hdl;
1106 (*zpp)->z_pflags = pflags;
1107 (*zpp)->z_mode = mode;
1108 (*zpp)->z_dnodesize = dnodesize;
1110 if (vap->va_mask & AT_XVATTR)
1111 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1113 if (obj_type == DMU_OT_ZNODE ||
1114 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1115 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1117 if (!(flag & IS_ROOT_NODE)) {
1121 vp->v_vflag |= VV_FORCEINSMQ;
1122 err = insmntque(vp, zfsvfs->z_vfs);
1123 vp->v_vflag &= ~VV_FORCEINSMQ;
1124 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1126 kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
1127 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1131 * Update in-core attributes. It is assumed the caller will be doing an
1132 * sa_bulk_update to push the changes out.
1135 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1139 xoap = xva_getxoptattr(xvap);
1142 ASSERT_VOP_IN_SEQC(ZTOV(zp));
1144 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1146 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1147 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1148 ×, sizeof (times), tx);
1149 XVA_SET_RTN(xvap, XAT_CREATETIME);
1151 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1152 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1154 XVA_SET_RTN(xvap, XAT_READONLY);
1156 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1157 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1159 XVA_SET_RTN(xvap, XAT_HIDDEN);
1161 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1162 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1164 XVA_SET_RTN(xvap, XAT_SYSTEM);
1166 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1167 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1169 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1171 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1172 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1174 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1176 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1177 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1179 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1181 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1182 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1184 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1186 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1187 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1189 XVA_SET_RTN(xvap, XAT_NODUMP);
1191 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1192 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1194 XVA_SET_RTN(xvap, XAT_OPAQUE);
1196 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1197 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1198 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1199 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1201 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1202 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1204 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1206 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1207 zfs_sa_set_scanstamp(zp, xvap, tx);
1208 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1210 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1211 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1213 XVA_SET_RTN(xvap, XAT_REPARSE);
1215 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1216 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1218 XVA_SET_RTN(xvap, XAT_OFFLINE);
1220 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1221 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1223 XVA_SET_RTN(xvap, XAT_SPARSE);
1228 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1230 dmu_object_info_t doi;
1240 getnewvnode_reserve();
1243 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1245 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1247 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1248 getnewvnode_drop_reserve();
1252 dmu_object_info_from_db(db, &doi);
1253 if (doi.doi_bonus_type != DMU_OT_SA &&
1254 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1255 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1256 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1257 sa_buf_rele(db, NULL);
1258 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1260 getnewvnode_drop_reserve();
1262 return (SET_ERROR(EINVAL));
1265 hdl = dmu_buf_get_user(db);
1267 zp = sa_get_userdata(hdl);
1270 * Since "SA" does immediate eviction we
1271 * should never find a sa handle that doesn't
1272 * know about the znode.
1274 ASSERT3P(zp, !=, NULL);
1275 ASSERT3U(zp->z_id, ==, obj_num);
1276 if (zp->z_unlinked) {
1277 err = SET_ERROR(ENOENT);
1281 * Don't let the vnode disappear after
1282 * ZFS_OBJ_HOLD_EXIT.
1289 sa_buf_rele(db, NULL);
1290 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1293 getnewvnode_drop_reserve();
1297 locked = VOP_ISLOCKED(vp);
1299 if (VN_IS_DOOMED(vp) && locked != LK_EXCLUSIVE) {
1301 * The vnode is doomed and this thread doesn't
1302 * hold the exclusive lock on it, so the vnode
1303 * must be being reclaimed by another thread.
1304 * Otherwise the doomed vnode is being reclaimed
1305 * by this thread and zfs_zget is called from
1311 * XXX vrele() locks the vnode when the last reference
1312 * is dropped. Although in this case the vnode is
1313 * doomed / dead and so no inactivation is required,
1314 * the vnode lock is still acquired. That could result
1315 * in a LOR with z_teardown_lock if another thread holds
1316 * the vnode's lock and tries to take z_teardown_lock.
1317 * But that is only possible if the other thread peforms
1318 * a ZFS vnode operation on the vnode. That either
1319 * should not happen if the vnode is dead or the thread
1320 * should also have a refrence to the vnode and thus
1321 * our reference is not last.
1327 getnewvnode_drop_reserve();
1332 * Not found create new znode/vnode
1333 * but only if file exists.
1335 * There is a small window where zfs_vget() could
1336 * find this object while a file create is still in
1337 * progress. This is checked for in zfs_znode_alloc()
1339 * if zfs_znode_alloc() fails it will drop the hold on the
1342 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1343 doi.doi_bonus_type, NULL);
1345 err = SET_ERROR(ENOENT);
1350 vnode_t *vp = ZTOV(zp);
1352 err = insmntque(vp, zfsvfs->z_vfs);
1354 vp->v_hash = obj_num;
1358 zfs_znode_dmu_fini(zp);
1363 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1364 getnewvnode_drop_reserve();
1369 zfs_rezget(znode_t *zp)
1371 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1372 dmu_object_info_t doi;
1375 uint64_t obj_num = zp->z_id;
1376 uint64_t mode, size;
1377 sa_bulk_attr_t bulk[8];
1383 * Remove cached pages before reloading the znode, so that they are not
1384 * lingering after we run into any error. Ideally, we should vgone()
1385 * the vnode in case of error, but currently we cannot do that
1386 * because of the LOR between the vnode lock and z_teardown_lock.
1387 * So, instead, we have to "doom" the znode in the illumos style.
1390 vn_pages_remove(vp, 0, 0);
1392 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1394 mutex_enter(&zp->z_acl_lock);
1395 if (zp->z_acl_cached) {
1396 zfs_acl_free(zp->z_acl_cached);
1397 zp->z_acl_cached = NULL;
1400 mutex_exit(&zp->z_acl_lock);
1401 ASSERT(zp->z_sa_hdl == NULL);
1402 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1404 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1408 dmu_object_info_from_db(db, &doi);
1409 if (doi.doi_bonus_type != DMU_OT_SA &&
1410 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1411 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1412 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1413 sa_buf_rele(db, NULL);
1414 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1415 return (SET_ERROR(EINVAL));
1418 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1421 /* reload cached values */
1422 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1423 &gen, sizeof (gen));
1424 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1425 &zp->z_size, sizeof (zp->z_size));
1426 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1427 &zp->z_links, sizeof (zp->z_links));
1428 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1429 &zp->z_pflags, sizeof (zp->z_pflags));
1430 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1431 &zp->z_atime, sizeof (zp->z_atime));
1432 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1433 &zp->z_uid, sizeof (zp->z_uid));
1434 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1435 &zp->z_gid, sizeof (zp->z_gid));
1436 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1437 &mode, sizeof (mode));
1439 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1440 zfs_znode_dmu_fini(zp);
1441 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1442 return (SET_ERROR(EIO));
1447 if (gen != zp->z_gen) {
1448 zfs_znode_dmu_fini(zp);
1449 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1450 return (SET_ERROR(EIO));
1454 * It is highly improbable but still quite possible that two
1455 * objects in different datasets are created with the same
1456 * object numbers and in transaction groups with the same
1457 * numbers. znodes corresponding to those objects would
1458 * have the same z_id and z_gen, but their other attributes
1460 * zfs recv -F may replace one of such objects with the other.
1461 * As a result file properties recorded in the replaced
1462 * object's vnode may no longer match the received object's
1463 * properties. At present the only cached property is the
1464 * files type recorded in v_type.
1465 * So, handle this case by leaving the old vnode and znode
1466 * disassociated from the actual object. A new vnode and a
1467 * znode will be created if the object is accessed
1468 * (e.g. via a look-up). The old vnode and znode will be
1469 * recycled when the last vnode reference is dropped.
1471 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1472 zfs_znode_dmu_fini(zp);
1473 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1474 return (SET_ERROR(EIO));
1478 * If the file has zero links, then it has been unlinked on the send
1479 * side and it must be in the received unlinked set.
1480 * We call zfs_znode_dmu_fini() now to prevent any accesses to the
1481 * stale data and to prevent automatical removal of the file in
1482 * zfs_zinactive(). The file will be removed either when it is removed
1483 * on the send side and the next incremental stream is received or
1484 * when the unlinked set gets processed.
1486 zp->z_unlinked = (zp->z_links == 0);
1487 if (zp->z_unlinked) {
1488 zfs_znode_dmu_fini(zp);
1489 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1493 zp->z_blksz = doi.doi_data_block_size;
1494 if (zp->z_size != size)
1495 vnode_pager_setsize(vp, zp->z_size);
1497 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1503 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1505 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1506 objset_t *os = zfsvfs->z_os;
1507 uint64_t obj = zp->z_id;
1508 uint64_t acl_obj = zfs_external_acl(zp);
1510 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1512 VERIFY(!zp->z_is_sa);
1513 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1515 VERIFY(0 == dmu_object_free(os, obj, tx));
1516 zfs_znode_dmu_fini(zp);
1517 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1522 zfs_zinactive(znode_t *zp)
1524 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1525 uint64_t z_id = zp->z_id;
1527 ASSERT(zp->z_sa_hdl);
1530 * Don't allow a zfs_zget() while were trying to release this znode
1532 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1535 * If this was the last reference to a file with no links, remove
1536 * the file from the file system unless the file system is mounted
1537 * read-only. That can happen, for example, if the file system was
1538 * originally read-write, the file was opened, then unlinked and
1539 * the file system was made read-only before the file was finally
1540 * closed. The file will remain in the unlinked set.
1542 if (zp->z_unlinked) {
1543 ASSERT(!zfsvfs->z_issnap);
1544 if ((zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) == 0) {
1545 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1551 zfs_znode_dmu_fini(zp);
1552 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1557 zfs_znode_free(znode_t *zp)
1559 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1561 ASSERT(zp->z_sa_hdl == NULL);
1563 mutex_enter(&zfsvfs->z_znodes_lock);
1564 POINTER_INVALIDATE(&zp->z_zfsvfs);
1565 list_remove(&zfsvfs->z_all_znodes, zp);
1566 mutex_exit(&zfsvfs->z_znodes_lock);
1568 if (zp->z_acl_cached) {
1569 zfs_acl_free(zp->z_acl_cached);
1570 zp->z_acl_cached = NULL;
1573 zfs_znode_free_kmem(zp);
1576 VFS_RELE(zfsvfs->z_vfs);
1581 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1582 uint64_t ctime[2], boolean_t have_tx)
1586 vfs_timestamp(&now);
1588 if (have_tx) { /* will sa_bulk_update happen really soon? */
1589 zp->z_atime_dirty = 0;
1592 zp->z_atime_dirty = 1;
1595 if (flag & AT_ATIME) {
1596 ZFS_TIME_ENCODE(&now, zp->z_atime);
1599 if (flag & AT_MTIME) {
1600 ZFS_TIME_ENCODE(&now, mtime);
1601 if (zp->z_zfsvfs->z_use_fuids) {
1602 zp->z_pflags |= (ZFS_ARCHIVE |
1607 if (flag & AT_CTIME) {
1608 ZFS_TIME_ENCODE(&now, ctime);
1609 if (zp->z_zfsvfs->z_use_fuids)
1610 zp->z_pflags |= ZFS_ARCHIVE;
1615 * Grow the block size for a file.
1617 * IN: zp - znode of file to free data in.
1618 * size - requested block size
1619 * tx - open transaction.
1621 * NOTE: this function assumes that the znode is write locked.
1624 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1629 if (size <= zp->z_blksz)
1632 * If the file size is already greater than the current blocksize,
1633 * we will not grow. If there is more than one block in a file,
1634 * the blocksize cannot change.
1636 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1639 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1642 if (error == ENOTSUP)
1646 /* What blocksize did we actually get? */
1647 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1652 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1653 * be calling back into the fs for a putpage(). E.g.: when truncating
1654 * a file, the pages being "thrown away* don't need to be written out.
1658 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1659 int flags, cred_t *cr)
1667 * Increase the file length
1669 * IN: zp - znode of file to free data in.
1670 * end - new end-of-file
1672 * RETURN: 0 on success, error code on failure
1675 zfs_extend(znode_t *zp, uint64_t end)
1677 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1684 * We will change zp_size, lock the whole file.
1686 lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
1689 * Nothing to do if file already at desired length.
1691 if (end <= zp->z_size) {
1695 tx = dmu_tx_create(zfsvfs->z_os);
1696 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1697 zfs_sa_upgrade_txholds(tx, zp);
1698 if (end > zp->z_blksz &&
1699 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1701 * We are growing the file past the current block size.
1703 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1705 * File's blocksize is already larger than the
1706 * "recordsize" property. Only let it grow to
1707 * the next power of 2.
1709 ASSERT(!ISP2(zp->z_blksz));
1710 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1712 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1714 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1719 error = dmu_tx_assign(tx, TXG_WAIT);
1727 zfs_grow_blocksize(zp, newblksz, tx);
1731 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1732 &zp->z_size, sizeof (zp->z_size), tx));
1734 vnode_pager_setsize(ZTOV(zp), end);
1744 * Free space in a file.
1746 * IN: zp - znode of file to free data in.
1747 * off - start of section to free.
1748 * len - length of section to free.
1750 * RETURN: 0 on success, error code on failure
1753 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1755 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1760 * Lock the range being freed.
1762 lr = rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER);
1765 * Nothing to do if file already at desired length.
1767 if (off >= zp->z_size) {
1772 if (off + len > zp->z_size)
1773 len = zp->z_size - off;
1775 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1779 * In FreeBSD we cannot free block in the middle of a file,
1780 * but only at the end of a file, so this code path should
1783 vnode_pager_setsize(ZTOV(zp), off);
1794 * IN: zp - znode of file to free data in.
1795 * end - new end-of-file.
1797 * RETURN: 0 on success, error code on failure
1800 zfs_trunc(znode_t *zp, uint64_t end)
1802 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1803 vnode_t *vp = ZTOV(zp);
1807 sa_bulk_attr_t bulk[2];
1811 * We will change zp_size, lock the whole file.
1813 lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
1816 * Nothing to do if file already at desired length.
1818 if (end >= zp->z_size) {
1823 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,
1829 tx = dmu_tx_create(zfsvfs->z_os);
1830 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1831 zfs_sa_upgrade_txholds(tx, zp);
1832 dmu_tx_mark_netfree(tx);
1833 error = dmu_tx_assign(tx, TXG_WAIT);
1841 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1842 NULL, &zp->z_size, sizeof (zp->z_size));
1845 zp->z_pflags &= ~ZFS_SPARSE;
1846 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1847 NULL, &zp->z_pflags, 8);
1849 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1854 * Clear any mapped pages in the truncated region. This has to
1855 * happen outside of the transaction to avoid the possibility of
1856 * a deadlock with someone trying to push a page that we are
1857 * about to invalidate.
1859 vnode_pager_setsize(vp, end);
1867 * Free space in a file
1869 * IN: zp - znode of file to free data in.
1870 * off - start of range
1871 * len - end of range (0 => EOF)
1872 * flag - current file open mode flags.
1873 * log - TRUE if this action should be logged
1875 * RETURN: 0 on success, error code on failure
1878 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1880 vnode_t *vp = ZTOV(zp);
1882 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1883 zilog_t *zilog = zfsvfs->z_log;
1885 uint64_t mtime[2], ctime[2];
1886 sa_bulk_attr_t bulk[3];
1890 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1891 sizeof (mode))) != 0)
1894 if (off > zp->z_size) {
1895 error = zfs_extend(zp, off+len);
1896 if (error == 0 && log)
1903 * Check for any locks in the region to be freed.
1906 if (MANDLOCK(vp, (mode_t)mode)) {
1907 uint64_t length = (len ? len : zp->z_size - off);
1908 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1913 error = zfs_trunc(zp, off);
1915 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1916 off + len > zp->z_size)
1917 error = zfs_extend(zp, off+len);
1922 tx = dmu_tx_create(zfsvfs->z_os);
1923 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1924 zfs_sa_upgrade_txholds(tx, zp);
1925 error = dmu_tx_assign(tx, TXG_WAIT);
1931 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1932 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1933 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1934 NULL, &zp->z_pflags, 8);
1935 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1936 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1939 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1946 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1948 uint64_t moid, obj, sa_obj, version;
1949 uint64_t sense = ZFS_CASE_SENSITIVE;
1954 znode_t *rootzp = NULL;
1958 zfs_acl_ids_t acl_ids;
1961 * First attempt to create master node.
1964 * In an empty objset, there are no blocks to read and thus
1965 * there can be no i/o errors (which we assert below).
1967 moid = MASTER_NODE_OBJ;
1968 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1969 DMU_OT_NONE, 0, tx);
1973 * Set starting attributes.
1975 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1977 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1978 /* For the moment we expect all zpl props to be uint64_ts */
1982 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1983 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1984 name = nvpair_name(elem);
1985 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1989 error = zap_update(os, moid, name, 8, 1, &val, tx);
1992 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1994 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1997 ASSERT(version != 0);
1998 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
2001 * Create zap object used for SA attribute registration
2004 if (version >= ZPL_VERSION_SA) {
2005 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
2006 DMU_OT_NONE, 0, tx);
2007 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
2013 * Create a delete queue.
2015 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
2017 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
2021 * Create root znode. Create minimal znode/vnode/zfsvfs
2022 * to allow zfs_mknode to work.
2025 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
2026 vattr.va_type = VDIR;
2027 vattr.va_mode = S_IFDIR|0755;
2028 vattr.va_uid = crgetuid(cr);
2029 vattr.va_gid = crgetgid(cr);
2031 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
2033 rootzp = zfs_znode_alloc_kmem(KM_SLEEP);
2034 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
2035 rootzp->z_moved = 0;
2036 rootzp->z_unlinked = 0;
2037 rootzp->z_atime_dirty = 0;
2038 rootzp->z_is_sa = USE_SA(version, os);
2041 zfsvfs->z_parent = zfsvfs;
2042 zfsvfs->z_version = version;
2043 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
2044 zfsvfs->z_use_sa = USE_SA(version, os);
2045 zfsvfs->z_norm = norm;
2047 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
2048 &zfsvfs->z_attr_table);
2053 * Fold case on file systems that are always or sometimes case
2056 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
2057 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
2059 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
2060 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
2061 offsetof(znode_t, z_link_node));
2063 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
2064 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
2066 rootzp->z_zfsvfs = zfsvfs;
2067 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
2068 cr, NULL, &acl_ids));
2069 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
2070 ASSERT3P(zp, ==, rootzp);
2071 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
2073 zfs_acl_ids_free(&acl_ids);
2074 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
2076 sa_handle_destroy(rootzp->z_sa_hdl);
2077 zfs_znode_free_kmem(rootzp);
2080 * Create shares directory
2083 error = zfs_create_share_dir(zfsvfs, tx);
2087 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
2088 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
2089 kmem_free(zfsvfs, sizeof (zfsvfs_t));
2091 #endif /* _KERNEL */
2094 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
2096 uint64_t sa_obj = 0;
2099 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
2100 if (error != 0 && error != ENOENT)
2103 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
2108 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
2109 dmu_buf_t **db, void *tag)
2111 dmu_object_info_t doi;
2114 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
2117 dmu_object_info_from_db(*db, &doi);
2118 if ((doi.doi_bonus_type != DMU_OT_SA &&
2119 doi.doi_bonus_type != DMU_OT_ZNODE) ||
2120 doi.doi_bonus_type == DMU_OT_ZNODE &&
2121 doi.doi_bonus_size < sizeof (znode_phys_t)) {
2122 sa_buf_rele(*db, tag);
2123 return (SET_ERROR(ENOTSUP));
2126 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
2128 sa_buf_rele(*db, tag);
2136 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
2138 sa_handle_destroy(hdl);
2139 sa_buf_rele(db, tag);
2143 * Given an object number, return its parent object number and whether
2144 * or not the object is an extended attribute directory.
2147 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
2148 uint64_t *pobjp, int *is_xattrdir)
2153 uint64_t parent_mode;
2154 sa_bulk_attr_t bulk[3];
2155 sa_handle_t *sa_hdl;
2160 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2161 &parent, sizeof (parent));
2162 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2163 &pflags, sizeof (pflags));
2164 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2165 &mode, sizeof (mode));
2167 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2171 * When a link is removed its parent pointer is not changed and will
2172 * be invalid. There are two cases where a link is removed but the
2173 * file stays around, when it goes to the delete queue and when there
2174 * are additional links.
2176 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2180 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2181 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2185 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2188 * Extended attributes can be applied to files, directories, etc.
2189 * Otherwise the parent must be a directory.
2191 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2192 return (SET_ERROR(EINVAL));
2200 * Given an object number, return some zpl level statistics
2203 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2206 sa_bulk_attr_t bulk[4];
2209 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2210 &sb->zs_mode, sizeof (sb->zs_mode));
2211 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2212 &sb->zs_gen, sizeof (sb->zs_gen));
2213 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2214 &sb->zs_links, sizeof (sb->zs_links));
2215 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2216 &sb->zs_ctime, sizeof (sb->zs_ctime));
2218 return (sa_bulk_lookup(hdl, bulk, count));
2222 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2223 sa_attr_type_t *sa_table, char *buf, int len)
2225 sa_handle_t *sa_hdl;
2226 sa_handle_t *prevhdl = NULL;
2227 dmu_buf_t *prevdb = NULL;
2228 dmu_buf_t *sa_db = NULL;
2229 char *path = buf + len - 1;
2235 uint64_t deleteq_obj;
2236 VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ,
2237 ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj));
2238 error = zap_lookup_int(osp, deleteq_obj, obj);
2241 } else if (error != ENOENT) {
2248 char component[MAXNAMELEN + 2];
2253 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2255 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2256 &is_xattrdir)) != 0)
2267 (void) sprintf(component + 1, "<xattrdir>");
2269 error = zap_value_search(osp, pobj, obj,
2270 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2275 complen = strlen(component);
2277 ASSERT(path >= buf);
2278 bcopy(component, path, complen);
2281 if (sa_hdl != hdl) {
2285 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2293 if (sa_hdl != NULL && sa_hdl != hdl) {
2294 ASSERT(sa_db != NULL);
2295 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2299 (void) memmove(buf, path, buf + len - path);
2305 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2307 sa_attr_type_t *sa_table;
2312 error = zfs_sa_setup(osp, &sa_table);
2316 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2320 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2322 zfs_release_sa_handle(hdl, db, FTAG);
2327 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2330 char *path = buf + len - 1;
2331 sa_attr_type_t *sa_table;
2338 error = zfs_sa_setup(osp, &sa_table);
2342 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2346 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2348 zfs_release_sa_handle(hdl, db, FTAG);
2352 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2354 zfs_release_sa_handle(hdl, db, FTAG);
2360 zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf)
2362 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2367 /* Extended attributes should not be visible as regular files. */
2368 if ((zp->z_pflags & ZFS_XATTR) != 0)
2369 return (SET_ERROR(EINVAL));
2371 err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table,
2372 &parent, &is_xattrdir);
2375 ASSERT0(is_xattrdir);
2377 /* No name as this is a root object. */
2378 if (parent == zp->z_id)
2379 return (SET_ERROR(EINVAL));
2381 err = zap_value_search(zfsvfs->z_os, parent, zp->z_id,
2382 ZFS_DIRENT_OBJ(-1ULL), buf);
2385 err = zfs_zget(zfsvfs, parent, dzpp);
2388 #endif /* _KERNEL */