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.
25 /* Portions Copyright 2007 Jeremy Teo */
28 #include <sys/types.h>
29 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
34 #include <sys/mntent.h>
35 #include <sys/mkdev.h>
36 #include <sys/u8_textprep.h>
37 #include <sys/dsl_dataset.h>
39 #include <sys/vfs_opreg.h>
40 #include <sys/vnode.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
46 #include <sys/atomic.h>
48 #include "fs/fs_subr.h"
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_acl.h>
51 #include <sys/zfs_ioctl.h>
52 #include <sys/zfs_rlock.h>
53 #include <sys/zfs_fuid.h>
54 #include <sys/zfs_vnops.h>
55 #include <sys/dnode.h>
56 #include <sys/fs/zfs.h>
57 #include <sys/kidmap.h>
62 #include <sys/refcount.h>
65 #include <sys/zfs_znode.h>
67 #include <sys/zfs_sa.h>
68 #include <sys/zfs_stat.h>
71 #include "zfs_comutil.h"
74 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
75 * turned on when DEBUG is also defined.
82 #define ZNODE_STAT_ADD(stat) ((stat)++)
84 #define ZNODE_STAT_ADD(stat) /* nothing */
85 #endif /* ZNODE_STATS */
88 * Functions needed for userland (ie: libzpool) are not put under
89 * #ifdef_KERNEL; the rest of the functions have dependencies
90 * (such as VFS logic) that will not compile easily in userland.
94 static kmem_cache_t *znode_cache = NULL;
98 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
102 inode_init_once(ZTOI(zp));
103 list_link_init(&zp->z_link_node);
105 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
106 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
107 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
108 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
110 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
111 avl_create(&zp->z_range_avl, zfs_range_compare,
112 sizeof (rl_t), offsetof(rl_t, r_node));
114 zp->z_dirlocks = NULL;
115 zp->z_acl_cached = NULL;
122 zfs_znode_cache_destructor(void *buf, void *arg)
126 ASSERT(!list_link_active(&zp->z_link_node));
127 mutex_destroy(&zp->z_lock);
128 rw_destroy(&zp->z_parent_lock);
129 rw_destroy(&zp->z_name_lock);
130 mutex_destroy(&zp->z_acl_lock);
131 avl_destroy(&zp->z_range_avl);
132 mutex_destroy(&zp->z_range_lock);
134 ASSERT(zp->z_dirlocks == NULL);
135 ASSERT(zp->z_acl_cached == NULL);
144 ASSERT(znode_cache == NULL);
145 znode_cache = kmem_cache_create("zfs_znode_cache",
146 sizeof (znode_t), 0, zfs_znode_cache_constructor,
147 zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_KMEM);
157 kmem_cache_destroy(znode_cache);
162 zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx)
165 zfs_acl_ids_t acl_ids;
172 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
173 vattr.va_mode = S_IFDIR | 0555;
174 vattr.va_uid = crgetuid(kcred);
175 vattr.va_gid = crgetgid(kcred);
177 sharezp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
178 sharezp->z_moved = 0;
179 sharezp->z_unlinked = 0;
180 sharezp->z_atime_dirty = 0;
181 sharezp->z_zfsvfs = zfsvfs;
182 sharezp->z_is_sa = zfsvfs->z_use_sa;
188 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
189 kcred, NULL, &acl_ids));
190 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
191 ASSERT3P(zp, ==, sharezp);
192 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
193 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
194 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
195 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
196 zfsvfs->z_shares_dir = sharezp->z_id;
198 zfs_acl_ids_free(&acl_ids);
199 // ZTOV(sharezp)->v_count = 0;
200 sa_handle_destroy(sharezp->z_sa_hdl);
201 kmem_cache_free(znode_cache, sharezp);
206 #endif /* HAVE_SHARE */
210 zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp,
211 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
213 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id)));
215 mutex_enter(&zp->z_lock);
217 ASSERT(zp->z_sa_hdl == NULL);
218 ASSERT(zp->z_acl_cached == NULL);
219 if (sa_hdl == NULL) {
220 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp,
221 SA_HDL_SHARED, &zp->z_sa_hdl));
223 zp->z_sa_hdl = sa_hdl;
224 sa_set_userp(sa_hdl, zp);
227 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
229 mutex_exit(&zp->z_lock);
233 zfs_znode_dmu_fini(znode_t *zp)
235 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) ||
237 RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
239 sa_handle_destroy(zp->z_sa_hdl);
244 * Called by new_inode() to allocate a new inode.
247 zfs_inode_alloc(struct super_block *sb, struct inode **ip)
251 zp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
258 * Called in multiple places when an inode should be destroyed.
261 zfs_inode_destroy(struct inode *ip)
263 znode_t *zp = ITOZ(ip);
264 zfs_sb_t *zsb = ZTOZSB(zp);
266 mutex_enter(&zsb->z_znodes_lock);
267 list_remove(&zsb->z_all_znodes, zp);
268 mutex_exit(&zsb->z_znodes_lock);
270 if (zp->z_acl_cached) {
271 zfs_acl_free(zp->z_acl_cached);
272 zp->z_acl_cached = NULL;
275 kmem_cache_free(znode_cache, zp);
279 zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
283 switch (ip->i_mode & S_IFMT) {
285 ip->i_op = &zpl_inode_operations;
286 ip->i_fop = &zpl_file_operations;
287 ip->i_mapping->a_ops = &zpl_address_space_operations;
291 ip->i_op = &zpl_dir_inode_operations;
292 ip->i_fop = &zpl_dir_file_operations;
293 ITOZ(ip)->z_zn_prefetch = B_TRUE;
297 ip->i_op = &zpl_symlink_inode_operations;
304 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
305 &rdev, sizeof (rdev)) == 0);
306 init_special_inode(ip, ip->i_mode, rdev);
307 ip->i_op = &zpl_special_inode_operations;
311 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
317 * Construct a znode+inode and initialize.
319 * This does not do a call to dmu_set_user() that is
320 * up to the caller to do, in case you don't want to
324 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
325 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
326 struct dentry *dentry)
331 sa_bulk_attr_t bulk[9];
336 ip = new_inode(zsb->z_sb);
341 ASSERT(zp->z_dirlocks == NULL);
345 zp->z_atime_dirty = 0;
347 zp->z_id = db->db_object;
349 zp->z_seq = 0x7A4653;
353 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
358 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
359 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
361 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
363 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
365 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
366 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
368 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
370 sa_handle_destroy(zp->z_sa_hdl);
376 zfs_inode_update(zp);
377 zfs_inode_set_ops(zsb, ip);
379 if (insert_inode_locked(ip))
383 d_instantiate(dentry, ip);
385 mutex_enter(&zsb->z_znodes_lock);
386 list_insert_tail(&zsb->z_all_znodes, zp);
388 mutex_exit(&zsb->z_znodes_lock);
390 unlock_new_inode(ip);
394 unlock_new_inode(ip);
400 * Update the embedded inode given the znode. We should work toward
401 * eliminating this function as soon as possible by removing values
402 * which are duplicated between the znode and inode. If the generic
403 * inode has the correct field it should be used, and the ZFS code
404 * updated to access the inode. This can be done incrementally.
407 zfs_inode_update(znode_t *zp)
412 uint64_t atime[2], mtime[2], ctime[2];
418 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
419 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
420 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
422 spin_lock(&ip->i_lock);
423 ip->i_generation = zp->z_gen;
424 ip->i_uid = zp->z_uid;
425 ip->i_gid = zp->z_gid;
426 ip->i_nlink = zp->z_links;
427 ip->i_mode = zp->z_mode;
428 ip->i_blkbits = SPA_MINBLOCKSHIFT;
429 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
430 (u_longlong_t *)&ip->i_blocks);
432 ZFS_TIME_DECODE(&ip->i_atime, atime);
433 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
434 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
436 i_size_write(ip, zp->z_size);
437 spin_unlock(&ip->i_lock);
440 static uint64_t empty_xattr;
441 static uint64_t pad[4];
442 static zfs_acl_phys_t acl_phys;
444 * Create a new DMU object to hold a zfs znode.
446 * IN: dzp - parent directory for new znode
447 * vap - file attributes for new znode
448 * tx - dmu transaction id for zap operations
449 * cr - credentials of caller
451 * IS_ROOT_NODE - new object will be root
452 * IS_XATTR - new object is an attribute
453 * bonuslen - length of bonus buffer
454 * setaclp - File/Dir initial ACL
455 * fuidp - Tracks fuid allocation.
457 * OUT: zpp - allocated znode
461 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
462 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
464 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
465 uint64_t mode, size, links, parent, pflags;
466 uint64_t dzp_pflags = 0;
468 zfs_sb_t *zsb = ZTOZSB(dzp);
475 dmu_object_type_t obj_type;
476 sa_bulk_attr_t *sa_attrs;
478 zfs_acl_locator_cb_t locate = { 0 };
481 obj = vap->va_nodeid;
482 now = vap->va_ctime; /* see zfs_replay_create() */
483 gen = vap->va_nblocks; /* ditto */
487 gen = dmu_tx_get_txg(tx);
490 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
491 bonuslen = (obj_type == DMU_OT_SA) ?
492 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
495 * Create a new DMU object.
498 * There's currently no mechanism for pre-reading the blocks that will
499 * be needed to allocate a new object, so we accept the small chance
500 * that there will be an i/o error and we will fail one of the
503 if (S_ISDIR(vap->va_mode)) {
505 err = zap_create_claim_norm(zsb->z_os, obj,
506 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
507 obj_type, bonuslen, tx);
508 ASSERT3U(err, ==, 0);
510 obj = zap_create_norm(zsb->z_os,
511 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
512 obj_type, bonuslen, tx);
516 err = dmu_object_claim(zsb->z_os, obj,
517 DMU_OT_PLAIN_FILE_CONTENTS, 0,
518 obj_type, bonuslen, tx);
519 ASSERT3U(err, ==, 0);
521 obj = dmu_object_alloc(zsb->z_os,
522 DMU_OT_PLAIN_FILE_CONTENTS, 0,
523 obj_type, bonuslen, tx);
527 ZFS_OBJ_HOLD_ENTER(zsb, obj);
528 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
531 * If this is the root, fix up the half-initialized parent pointer
532 * to reference the just-allocated physical data area.
534 if (flag & IS_ROOT_NODE) {
537 dzp_pflags = dzp->z_pflags;
541 * If parent is an xattr, so am I.
543 if (dzp_pflags & ZFS_XATTR) {
547 if (zsb->z_use_fuids)
548 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
552 if (S_ISDIR(vap->va_mode)) {
553 size = 2; /* contents ("." and "..") */
554 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
559 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
560 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))
564 mode = acl_ids->z_mode;
569 * No execs denied will be deterimed when zfs_mode_compute() is called.
571 pflags |= acl_ids->z_aclp->z_hints &
572 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
573 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
575 ZFS_TIME_ENCODE(&now, crtime);
576 ZFS_TIME_ENCODE(&now, ctime);
578 if (vap->va_mask & ATTR_ATIME) {
579 ZFS_TIME_ENCODE(&vap->va_atime, atime);
581 ZFS_TIME_ENCODE(&now, atime);
584 if (vap->va_mask & ATTR_MTIME) {
585 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
587 ZFS_TIME_ENCODE(&now, mtime);
590 /* Now add in all of the "SA" attributes */
591 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
595 * Setup the array of attributes to be replaced/set on the new file
597 * order for DMU_OT_ZNODE is critical since it needs to be constructed
598 * in the old znode_phys_t format. Don't change this ordering
600 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
602 if (obj_type == DMU_OT_ZNODE) {
603 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
605 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
607 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
609 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
611 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
613 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
615 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
617 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
620 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
622 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
624 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
626 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
627 NULL, &acl_ids->z_fuid, 8);
628 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
629 NULL, &acl_ids->z_fgid, 8);
630 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
632 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
634 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
636 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
638 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
640 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
644 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
646 if (obj_type == DMU_OT_ZNODE) {
647 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
650 if (obj_type == DMU_OT_ZNODE ||
651 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
652 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))) {
653 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
656 if (obj_type == DMU_OT_ZNODE) {
657 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
659 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
660 &acl_ids->z_fuid, 8);
661 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
662 &acl_ids->z_fgid, 8);
663 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
664 sizeof (uint64_t) * 4);
665 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
666 &acl_phys, sizeof (zfs_acl_phys_t));
667 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
668 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
669 &acl_ids->z_aclp->z_acl_count, 8);
670 locate.cb_aclp = acl_ids->z_aclp;
671 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
672 zfs_acl_data_locator, &locate,
673 acl_ids->z_aclp->z_acl_bytes);
674 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
675 acl_ids->z_fuid, acl_ids->z_fgid);
678 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
680 if (!(flag & IS_ROOT_NODE)) {
681 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
683 ASSERT(*zpp != NULL);
685 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
686 ASSERT3S(err, ==, 0);
689 * If we are creating the root node, the "parent" we
690 * passed in is the znode for the root.
694 (*zpp)->z_sa_hdl = sa_hdl;
697 (*zpp)->z_pflags = pflags;
698 (*zpp)->z_mode = mode;
700 if (obj_type == DMU_OT_ZNODE ||
701 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
702 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
703 ASSERT3S(err, ==, 0);
705 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
706 ZFS_OBJ_HOLD_EXIT(zsb, obj);
710 * zfs_xvattr_set only updates the in-core attributes
711 * it is assumed the caller will be doing an sa_bulk_update
712 * to push the changes out
715 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
719 xoap = xva_getxoptattr(xvap);
722 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
724 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
725 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
726 ×, sizeof (times), tx);
727 XVA_SET_RTN(xvap, XAT_CREATETIME);
729 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
730 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
732 XVA_SET_RTN(xvap, XAT_READONLY);
734 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
735 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
737 XVA_SET_RTN(xvap, XAT_HIDDEN);
739 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
740 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
742 XVA_SET_RTN(xvap, XAT_SYSTEM);
744 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
745 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
747 XVA_SET_RTN(xvap, XAT_ARCHIVE);
749 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
750 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
752 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
754 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
755 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
757 XVA_SET_RTN(xvap, XAT_NOUNLINK);
759 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
760 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
762 XVA_SET_RTN(xvap, XAT_APPENDONLY);
764 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
765 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
767 XVA_SET_RTN(xvap, XAT_NODUMP);
769 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
770 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
772 XVA_SET_RTN(xvap, XAT_OPAQUE);
774 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
775 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
776 xoap->xoa_av_quarantined, zp->z_pflags, tx);
777 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
779 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
780 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
782 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
784 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
785 zfs_sa_set_scanstamp(zp, xvap, tx);
786 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
788 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
789 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
791 XVA_SET_RTN(xvap, XAT_REPARSE);
793 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
794 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
796 XVA_SET_RTN(xvap, XAT_OFFLINE);
798 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
799 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
801 XVA_SET_RTN(xvap, XAT_SPARSE);
806 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
808 dmu_object_info_t doi;
818 ip = ilookup(zsb->z_sb, obj_num);
820 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
822 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
824 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
829 dmu_object_info_from_db(db, &doi);
830 if (doi.doi_bonus_type != DMU_OT_SA &&
831 (doi.doi_bonus_type != DMU_OT_ZNODE ||
832 (doi.doi_bonus_type == DMU_OT_ZNODE &&
833 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
834 sa_buf_rele(db, NULL);
835 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
840 hdl = dmu_buf_get_user(db);
844 * ilookup returned NULL, which means
845 * the znode is dying - but the SA handle isn't
846 * quite dead yet, we need to drop any locks
847 * we're holding, re-schedule the task and try again.
849 sa_buf_rele(db, NULL);
850 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
856 zp = sa_get_userdata(hdl);
859 * Since "SA" does immediate eviction we
860 * should never find a sa handle that doesn't
861 * know about the znode.
864 ASSERT3P(zp, !=, NULL);
866 mutex_enter(&zp->z_lock);
867 ASSERT3U(zp->z_id, ==, obj_num);
868 if (zp->z_unlinked) {
875 sa_buf_rele(db, NULL);
876 mutex_exit(&zp->z_lock);
877 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
882 ASSERT3P(ip, ==, NULL);
885 * Not found create new znode/vnode but only if file exists.
887 * There is a small window where zfs_vget() could
888 * find this object while a file create is still in
889 * progress. This is checked for in zfs_znode_alloc()
891 * if zfs_znode_alloc() fails it will drop the hold on the
894 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
895 doi.doi_bonus_type, obj_num, NULL, NULL);
901 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
906 zfs_rezget(znode_t *zp)
908 zfs_sb_t *zsb = ZTOZSB(zp);
909 dmu_object_info_t doi;
911 uint64_t obj_num = zp->z_id;
913 sa_bulk_attr_t bulk[8];
918 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
920 mutex_enter(&zp->z_acl_lock);
921 if (zp->z_acl_cached) {
922 zfs_acl_free(zp->z_acl_cached);
923 zp->z_acl_cached = NULL;
926 mutex_exit(&zp->z_acl_lock);
927 ASSERT(zp->z_sa_hdl == NULL);
928 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
930 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
934 dmu_object_info_from_db(db, &doi);
935 if (doi.doi_bonus_type != DMU_OT_SA &&
936 (doi.doi_bonus_type != DMU_OT_ZNODE ||
937 (doi.doi_bonus_type == DMU_OT_ZNODE &&
938 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
939 sa_buf_rele(db, NULL);
940 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
944 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
946 /* reload cached values */
947 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
949 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
950 &zp->z_size, sizeof (zp->z_size));
951 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
952 &zp->z_links, sizeof (zp->z_links));
953 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
954 &zp->z_pflags, sizeof (zp->z_pflags));
955 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
956 &zp->z_atime, sizeof (zp->z_atime));
957 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
958 &zp->z_uid, sizeof (zp->z_uid));
959 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
960 &zp->z_gid, sizeof (zp->z_gid));
961 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
962 &mode, sizeof (mode));
964 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
965 zfs_znode_dmu_fini(zp);
966 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
972 if (gen != zp->z_gen) {
973 zfs_znode_dmu_fini(zp);
974 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
978 zp->z_unlinked = (zp->z_links == 0);
979 zp->z_blksz = doi.doi_data_block_size;
981 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
987 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
989 zfs_sb_t *zsb = ZTOZSB(zp);
990 objset_t *os = zsb->z_os;
991 uint64_t obj = zp->z_id;
992 uint64_t acl_obj = zfs_external_acl(zp);
994 ZFS_OBJ_HOLD_ENTER(zsb, obj);
996 VERIFY(!zp->z_is_sa);
997 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
999 VERIFY(0 == dmu_object_free(os, obj, tx));
1000 zfs_znode_dmu_fini(zp);
1001 ZFS_OBJ_HOLD_EXIT(zsb, obj);
1005 zfs_zinactive(znode_t *zp)
1007 zfs_sb_t *zsb = ZTOZSB(zp);
1008 uint64_t z_id = zp->z_id;
1009 boolean_t drop_mutex = 0;
1011 ASSERT(zp->z_sa_hdl);
1014 * Don't allow a zfs_zget() while were trying to release this znode.
1016 * Linux allows direct memory reclaim which means that any KM_SLEEP
1017 * allocation may trigger inode eviction. This can lead to a deadlock
1018 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
1019 * zfs_zinactive() call path. To avoid this deadlock the process
1020 * must not reacquire the mutex when it is already holding it.
1022 if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) {
1023 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
1027 mutex_enter(&zp->z_lock);
1030 * If this was the last reference to a file with no links,
1031 * remove the file from the file system.
1033 if (zp->z_unlinked) {
1034 mutex_exit(&zp->z_lock);
1037 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1043 mutex_exit(&zp->z_lock);
1044 zfs_znode_dmu_fini(zp);
1047 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1051 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1052 uint64_t ctime[2], boolean_t have_tx)
1058 if (have_tx) { /* will sa_bulk_update happen really soon? */
1059 zp->z_atime_dirty = 0;
1062 zp->z_atime_dirty = 1;
1065 if (flag & ATTR_ATIME) {
1066 ZFS_TIME_ENCODE(&now, zp->z_atime);
1069 if (flag & ATTR_MTIME) {
1070 ZFS_TIME_ENCODE(&now, mtime);
1071 if (ZTOZSB(zp)->z_use_fuids) {
1072 zp->z_pflags |= (ZFS_ARCHIVE |
1077 if (flag & ATTR_CTIME) {
1078 ZFS_TIME_ENCODE(&now, ctime);
1079 if (ZTOZSB(zp)->z_use_fuids)
1080 zp->z_pflags |= ZFS_ARCHIVE;
1085 * Grow the block size for a file.
1087 * IN: zp - znode of file to free data in.
1088 * size - requested block size
1089 * tx - open transaction.
1091 * NOTE: this function assumes that the znode is write locked.
1094 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1099 if (size <= zp->z_blksz)
1102 * If the file size is already greater than the current blocksize,
1103 * we will not grow. If there is more than one block in a file,
1104 * the blocksize cannot change.
1106 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1109 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1112 if (error == ENOTSUP)
1114 ASSERT3U(error, ==, 0);
1116 /* What blocksize did we actually get? */
1117 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1122 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1123 * be calling back into the fs for a putpage(). E.g.: when truncating
1124 * a file, the pages being "thrown away* don't need to be written out.
1128 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1129 int flags, cred_t *cr)
1134 #endif /* HAVE_MMAP */
1137 * Increase the file length
1139 * IN: zp - znode of file to free data in.
1140 * end - new end-of-file
1142 * RETURN: 0 if success
1143 * error code if failure
1146 zfs_extend(znode_t *zp, uint64_t end)
1148 zfs_sb_t *zsb = ZTOZSB(zp);
1155 * We will change zp_size, lock the whole file.
1157 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1160 * Nothing to do if file already at desired length.
1162 if (end <= zp->z_size) {
1163 zfs_range_unlock(rl);
1167 tx = dmu_tx_create(zsb->z_os);
1168 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1169 zfs_sa_upgrade_txholds(tx, zp);
1170 if (end > zp->z_blksz &&
1171 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1173 * We are growing the file past the current block size.
1175 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1176 ASSERT(!ISP2(zp->z_blksz));
1177 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1179 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1181 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1186 error = dmu_tx_assign(tx, TXG_NOWAIT);
1188 if (error == ERESTART) {
1194 zfs_range_unlock(rl);
1199 zfs_grow_blocksize(zp, newblksz, tx);
1203 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1204 &zp->z_size, sizeof (zp->z_size), tx));
1206 zfs_range_unlock(rl);
1214 * Free space in a file.
1216 * IN: zp - znode of file to free data in.
1217 * off - start of section to free.
1218 * len - length of section to free.
1220 * RETURN: 0 if success
1221 * error code if failure
1224 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1226 zfs_sb_t *zsb = ZTOZSB(zp);
1231 * Lock the range being freed.
1233 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1236 * Nothing to do if file already at desired length.
1238 if (off >= zp->z_size) {
1239 zfs_range_unlock(rl);
1243 if (off + len > zp->z_size)
1244 len = zp->z_size - off;
1246 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1248 zfs_range_unlock(rl);
1256 * IN: zp - znode of file to free data in.
1257 * end - new end-of-file.
1259 * RETURN: 0 if success
1260 * error code if failure
1263 zfs_trunc(znode_t *zp, uint64_t end)
1265 zfs_sb_t *zsb = ZTOZSB(zp);
1269 sa_bulk_attr_t bulk[2];
1273 * We will change zp_size, lock the whole file.
1275 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1278 * Nothing to do if file already at desired length.
1280 if (end >= zp->z_size) {
1281 zfs_range_unlock(rl);
1285 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1287 zfs_range_unlock(rl);
1291 tx = dmu_tx_create(zsb->z_os);
1292 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1293 zfs_sa_upgrade_txholds(tx, zp);
1294 error = dmu_tx_assign(tx, TXG_NOWAIT);
1296 if (error == ERESTART) {
1302 zfs_range_unlock(rl);
1307 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1308 NULL, &zp->z_size, sizeof (zp->z_size));
1311 zp->z_pflags &= ~ZFS_SPARSE;
1312 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1313 NULL, &zp->z_pflags, 8);
1315 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1319 zfs_range_unlock(rl);
1325 * Free space in a file
1327 * IN: zp - znode of file to free data in.
1328 * off - start of range
1329 * len - end of range (0 => EOF)
1330 * flag - current file open mode flags.
1331 * log - TRUE if this action should be logged
1333 * RETURN: 0 if success
1334 * error code if failure
1337 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1339 #ifdef HAVE_MANDLOCKS
1340 struct inode *ip = ZTOI(zp);
1341 #endif /* HAVE_MANDLOCKS */
1343 zfs_sb_t *zsb = ZTOZSB(zp);
1344 zilog_t *zilog = zsb->z_log;
1346 uint64_t mtime[2], ctime[2];
1347 sa_bulk_attr_t bulk[3];
1351 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1352 sizeof (mode))) != 0)
1355 if (off > zp->z_size) {
1356 error = zfs_extend(zp, off+len);
1357 if (error == 0 && log)
1363 #ifdef HAVE_MANDLOCKS
1365 * Check for any locks in the region to be freed.
1368 if (MANDLOCK(ip, (mode_t)mode)) {
1369 uint64_t length = (len ? len : zp->z_size - off);
1370 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1373 #endif /* HAVE_MANDLOCKS */
1376 error = zfs_trunc(zp, off);
1378 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1379 off + len > zp->z_size)
1380 error = zfs_extend(zp, off+len);
1385 tx = dmu_tx_create(zsb->z_os);
1386 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1387 zfs_sa_upgrade_txholds(tx, zp);
1388 error = dmu_tx_assign(tx, TXG_NOWAIT);
1390 if (error == ERESTART) {
1399 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1400 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1401 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1402 NULL, &zp->z_pflags, 8);
1403 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1404 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1407 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1410 zfs_inode_update(zp);
1415 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1417 uint64_t moid, obj, sa_obj, version;
1426 * First attempt to create master node.
1429 * In an empty objset, there are no blocks to read and thus
1430 * there can be no i/o errors (which we assert below).
1432 moid = MASTER_NODE_OBJ;
1433 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1434 DMU_OT_NONE, 0, tx);
1438 * Set starting attributes.
1440 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1442 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1443 /* For the moment we expect all zpl props to be uint64_ts */
1447 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1448 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1449 name = nvpair_name(elem);
1450 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1454 error = zap_update(os, moid, name, 8, 1, &val, tx);
1457 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1460 ASSERT(version != 0);
1461 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1464 * Create zap object used for SA attribute registration
1467 if (version >= ZPL_VERSION_SA) {
1468 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1469 DMU_OT_NONE, 0, tx);
1470 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1476 * Create a delete queue.
1478 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1480 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1484 * Create root znode with code free of VFS dependencies. This
1485 * is important because without a registered filesystem and super
1486 * block all the required VFS hooks will be missing. The critical
1487 * thing is to just crete the required root znode.
1489 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1490 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1492 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1493 dmu_buf_will_dirty(db, tx);
1496 * Initialize the znode physical data to zero.
1498 ASSERT(db->db_size >= sizeof (znode_phys_t));
1499 bzero(db->db_data, db->db_size);
1502 if (USE_FUIDS(version, os))
1503 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1505 pzp->zp_size = 2; /* "." and ".." */
1507 pzp->zp_parent = obj;
1508 pzp->zp_gen = dmu_tx_get_txg(tx);
1509 pzp->zp_mode = S_IFDIR | 0755;
1510 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1514 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1515 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1516 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1517 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1519 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1522 dmu_buf_rele(db, FTAG);
1525 #endif /* _KERNEL */
1528 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1530 uint64_t sa_obj = 0;
1533 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1534 if (error != 0 && error != ENOENT)
1537 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1542 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1545 dmu_object_info_t doi;
1548 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1551 dmu_object_info_from_db(*db, &doi);
1552 if ((doi.doi_bonus_type != DMU_OT_SA &&
1553 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1554 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1555 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1556 sa_buf_rele(*db, FTAG);
1560 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1562 sa_buf_rele(*db, FTAG);
1570 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1572 sa_handle_destroy(hdl);
1573 sa_buf_rele(db, FTAG);
1577 * Given an object number, return its parent object number and whether
1578 * or not the object is an extended attribute directory.
1581 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1587 sa_bulk_attr_t bulk[3];
1591 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1592 &parent, sizeof (parent));
1593 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1594 &pflags, sizeof (pflags));
1595 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1596 &mode, sizeof (mode));
1598 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1602 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1608 * Given an object number, return some zpl level statistics
1611 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1614 sa_bulk_attr_t bulk[4];
1617 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1618 &sb->zs_mode, sizeof (sb->zs_mode));
1619 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1620 &sb->zs_gen, sizeof (sb->zs_gen));
1621 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1622 &sb->zs_links, sizeof (sb->zs_links));
1623 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1624 &sb->zs_ctime, sizeof (sb->zs_ctime));
1626 return (sa_bulk_lookup(hdl, bulk, count));
1630 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1631 sa_attr_type_t *sa_table, char *buf, int len)
1633 sa_handle_t *sa_hdl;
1634 sa_handle_t *prevhdl = NULL;
1635 dmu_buf_t *prevdb = NULL;
1636 dmu_buf_t *sa_db = NULL;
1637 char *path = buf + len - 1;
1645 char component[MAXNAMELEN + 2];
1650 zfs_release_sa_handle(prevhdl, prevdb);
1652 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1653 &is_xattrdir)) != 0)
1664 (void) sprintf(component + 1, "<xattrdir>");
1666 error = zap_value_search(osp, pobj, obj,
1667 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1672 complen = strlen(component);
1674 ASSERT(path >= buf);
1675 bcopy(component, path, complen);
1678 if (sa_hdl != hdl) {
1682 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1690 if (sa_hdl != NULL && sa_hdl != hdl) {
1691 ASSERT(sa_db != NULL);
1692 zfs_release_sa_handle(sa_hdl, sa_db);
1696 (void) memmove(buf, path, buf + len - path);
1702 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1704 sa_attr_type_t *sa_table;
1709 error = zfs_sa_setup(osp, &sa_table);
1713 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1717 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1719 zfs_release_sa_handle(hdl, db);
1724 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1727 char *path = buf + len - 1;
1728 sa_attr_type_t *sa_table;
1735 error = zfs_sa_setup(osp, &sa_table);
1739 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1743 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1745 zfs_release_sa_handle(hdl, db);
1749 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1751 zfs_release_sa_handle(hdl, db);
1755 #if defined(_KERNEL) && defined(HAVE_SPL)
1756 EXPORT_SYMBOL(zfs_create_fs);
1757 EXPORT_SYMBOL(zfs_obj_to_path);