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;
301 * rdev is only stored in a SA only for device files.
305 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
306 &rdev, sizeof (rdev)) == 0);
310 init_special_inode(ip, ip->i_mode, rdev);
311 ip->i_op = &zpl_special_inode_operations;
315 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
321 * Construct a znode+inode and initialize.
323 * This does not do a call to dmu_set_user() that is
324 * up to the caller to do, in case you don't want to
328 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
329 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
330 struct dentry *dentry)
335 sa_bulk_attr_t bulk[9];
340 ip = new_inode(zsb->z_sb);
345 ASSERT(zp->z_dirlocks == NULL);
349 zp->z_atime_dirty = 0;
351 zp->z_id = db->db_object;
353 zp->z_seq = 0x7A4653;
357 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
359 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
360 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
361 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
362 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
363 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
365 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
367 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
369 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
370 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
372 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
374 sa_handle_destroy(zp->z_sa_hdl);
380 zfs_inode_update(zp);
381 zfs_inode_set_ops(zsb, ip);
383 if (insert_inode_locked(ip))
387 d_instantiate(dentry, ip);
389 mutex_enter(&zsb->z_znodes_lock);
390 list_insert_tail(&zsb->z_all_znodes, zp);
392 mutex_exit(&zsb->z_znodes_lock);
394 unlock_new_inode(ip);
398 unlock_new_inode(ip);
404 * Update the embedded inode given the znode. We should work toward
405 * eliminating this function as soon as possible by removing values
406 * which are duplicated between the znode and inode. If the generic
407 * inode has the correct field it should be used, and the ZFS code
408 * updated to access the inode. This can be done incrementally.
411 zfs_inode_update(znode_t *zp)
416 uint64_t atime[2], mtime[2], ctime[2];
422 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
423 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
424 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
426 spin_lock(&ip->i_lock);
427 ip->i_generation = zp->z_gen;
428 ip->i_uid = zp->z_uid;
429 ip->i_gid = zp->z_gid;
430 ip->i_nlink = zp->z_links;
431 ip->i_mode = zp->z_mode;
432 ip->i_blkbits = SPA_MINBLOCKSHIFT;
433 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
434 (u_longlong_t *)&ip->i_blocks);
436 ZFS_TIME_DECODE(&ip->i_atime, atime);
437 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
438 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
440 i_size_write(ip, zp->z_size);
441 spin_unlock(&ip->i_lock);
444 static uint64_t empty_xattr;
445 static uint64_t pad[4];
446 static zfs_acl_phys_t acl_phys;
448 * Create a new DMU object to hold a zfs znode.
450 * IN: dzp - parent directory for new znode
451 * vap - file attributes for new znode
452 * tx - dmu transaction id for zap operations
453 * cr - credentials of caller
455 * IS_ROOT_NODE - new object will be root
456 * IS_XATTR - new object is an attribute
457 * bonuslen - length of bonus buffer
458 * setaclp - File/Dir initial ACL
459 * fuidp - Tracks fuid allocation.
461 * OUT: zpp - allocated znode
465 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
466 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
468 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
469 uint64_t mode, size, links, parent, pflags;
470 uint64_t dzp_pflags = 0;
472 zfs_sb_t *zsb = ZTOZSB(dzp);
479 dmu_object_type_t obj_type;
480 sa_bulk_attr_t *sa_attrs;
482 zfs_acl_locator_cb_t locate = { 0 };
485 obj = vap->va_nodeid;
486 now = vap->va_ctime; /* see zfs_replay_create() */
487 gen = vap->va_nblocks; /* ditto */
491 gen = dmu_tx_get_txg(tx);
494 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
495 bonuslen = (obj_type == DMU_OT_SA) ?
496 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
499 * Create a new DMU object.
502 * There's currently no mechanism for pre-reading the blocks that will
503 * be needed to allocate a new object, so we accept the small chance
504 * that there will be an i/o error and we will fail one of the
507 if (S_ISDIR(vap->va_mode)) {
509 err = zap_create_claim_norm(zsb->z_os, obj,
510 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
511 obj_type, bonuslen, tx);
512 ASSERT3U(err, ==, 0);
514 obj = zap_create_norm(zsb->z_os,
515 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
516 obj_type, bonuslen, tx);
520 err = dmu_object_claim(zsb->z_os, obj,
521 DMU_OT_PLAIN_FILE_CONTENTS, 0,
522 obj_type, bonuslen, tx);
523 ASSERT3U(err, ==, 0);
525 obj = dmu_object_alloc(zsb->z_os,
526 DMU_OT_PLAIN_FILE_CONTENTS, 0,
527 obj_type, bonuslen, tx);
531 ZFS_OBJ_HOLD_ENTER(zsb, obj);
532 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
535 * If this is the root, fix up the half-initialized parent pointer
536 * to reference the just-allocated physical data area.
538 if (flag & IS_ROOT_NODE) {
541 dzp_pflags = dzp->z_pflags;
545 * If parent is an xattr, so am I.
547 if (dzp_pflags & ZFS_XATTR) {
551 if (zsb->z_use_fuids)
552 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
556 if (S_ISDIR(vap->va_mode)) {
557 size = 2; /* contents ("." and "..") */
558 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
563 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))
567 mode = acl_ids->z_mode;
572 * No execs denied will be deterimed when zfs_mode_compute() is called.
574 pflags |= acl_ids->z_aclp->z_hints &
575 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
576 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
578 ZFS_TIME_ENCODE(&now, crtime);
579 ZFS_TIME_ENCODE(&now, ctime);
581 if (vap->va_mask & ATTR_ATIME) {
582 ZFS_TIME_ENCODE(&vap->va_atime, atime);
584 ZFS_TIME_ENCODE(&now, atime);
587 if (vap->va_mask & ATTR_MTIME) {
588 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
590 ZFS_TIME_ENCODE(&now, mtime);
593 /* Now add in all of the "SA" attributes */
594 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
598 * Setup the array of attributes to be replaced/set on the new file
600 * order for DMU_OT_ZNODE is critical since it needs to be constructed
601 * in the old znode_phys_t format. Don't change this ordering
603 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
605 if (obj_type == DMU_OT_ZNODE) {
606 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
608 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
610 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
612 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
614 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
616 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
618 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
620 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
623 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
625 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
627 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
629 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
630 NULL, &acl_ids->z_fuid, 8);
631 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
632 NULL, &acl_ids->z_fgid, 8);
633 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
635 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
637 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
639 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
641 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
643 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
647 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
649 if (obj_type == DMU_OT_ZNODE) {
650 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
653 if (obj_type == DMU_OT_ZNODE ||
654 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) {
655 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
658 if (obj_type == DMU_OT_ZNODE) {
659 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
661 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
662 &acl_ids->z_fuid, 8);
663 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
664 &acl_ids->z_fgid, 8);
665 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
666 sizeof (uint64_t) * 4);
667 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
668 &acl_phys, sizeof (zfs_acl_phys_t));
669 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
670 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
671 &acl_ids->z_aclp->z_acl_count, 8);
672 locate.cb_aclp = acl_ids->z_aclp;
673 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
674 zfs_acl_data_locator, &locate,
675 acl_ids->z_aclp->z_acl_bytes);
676 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
677 acl_ids->z_fuid, acl_ids->z_fgid);
680 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
682 if (!(flag & IS_ROOT_NODE)) {
683 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
685 ASSERT(*zpp != NULL);
687 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
688 ASSERT3S(err, ==, 0);
691 * If we are creating the root node, the "parent" we
692 * passed in is the znode for the root.
696 (*zpp)->z_sa_hdl = sa_hdl;
699 (*zpp)->z_pflags = pflags;
700 (*zpp)->z_mode = mode;
702 if (obj_type == DMU_OT_ZNODE ||
703 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
704 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
705 ASSERT3S(err, ==, 0);
707 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
708 ZFS_OBJ_HOLD_EXIT(zsb, obj);
712 * zfs_xvattr_set only updates the in-core attributes
713 * it is assumed the caller will be doing an sa_bulk_update
714 * to push the changes out
717 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
721 xoap = xva_getxoptattr(xvap);
724 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
726 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
727 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
728 ×, sizeof (times), tx);
729 XVA_SET_RTN(xvap, XAT_CREATETIME);
731 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
732 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
734 XVA_SET_RTN(xvap, XAT_READONLY);
736 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
737 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
739 XVA_SET_RTN(xvap, XAT_HIDDEN);
741 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
742 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
744 XVA_SET_RTN(xvap, XAT_SYSTEM);
746 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
747 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
749 XVA_SET_RTN(xvap, XAT_ARCHIVE);
751 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
752 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
754 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
756 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
757 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
759 XVA_SET_RTN(xvap, XAT_NOUNLINK);
761 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
762 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
764 XVA_SET_RTN(xvap, XAT_APPENDONLY);
766 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
767 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
769 XVA_SET_RTN(xvap, XAT_NODUMP);
771 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
772 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
774 XVA_SET_RTN(xvap, XAT_OPAQUE);
776 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
777 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
778 xoap->xoa_av_quarantined, zp->z_pflags, tx);
779 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
781 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
782 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
784 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
786 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
787 zfs_sa_set_scanstamp(zp, xvap, tx);
788 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
790 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
791 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
793 XVA_SET_RTN(xvap, XAT_REPARSE);
795 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
796 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
798 XVA_SET_RTN(xvap, XAT_OFFLINE);
800 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
801 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
803 XVA_SET_RTN(xvap, XAT_SPARSE);
808 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
810 dmu_object_info_t doi;
820 ip = ilookup(zsb->z_sb, obj_num);
822 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
824 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
826 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
831 dmu_object_info_from_db(db, &doi);
832 if (doi.doi_bonus_type != DMU_OT_SA &&
833 (doi.doi_bonus_type != DMU_OT_ZNODE ||
834 (doi.doi_bonus_type == DMU_OT_ZNODE &&
835 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
836 sa_buf_rele(db, NULL);
837 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
842 hdl = dmu_buf_get_user(db);
846 * ilookup returned NULL, which means
847 * the znode is dying - but the SA handle isn't
848 * quite dead yet, we need to drop any locks
849 * we're holding, re-schedule the task and try again.
851 sa_buf_rele(db, NULL);
852 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
858 zp = sa_get_userdata(hdl);
861 * Since "SA" does immediate eviction we
862 * should never find a sa handle that doesn't
863 * know about the znode.
866 ASSERT3P(zp, !=, NULL);
868 mutex_enter(&zp->z_lock);
869 ASSERT3U(zp->z_id, ==, obj_num);
870 if (zp->z_unlinked) {
877 sa_buf_rele(db, NULL);
878 mutex_exit(&zp->z_lock);
879 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
884 ASSERT3P(ip, ==, NULL);
887 * Not found create new znode/vnode but only if file exists.
889 * There is a small window where zfs_vget() could
890 * find this object while a file create is still in
891 * progress. This is checked for in zfs_znode_alloc()
893 * if zfs_znode_alloc() fails it will drop the hold on the
896 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
897 doi.doi_bonus_type, obj_num, NULL, NULL);
903 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
908 zfs_rezget(znode_t *zp)
910 zfs_sb_t *zsb = ZTOZSB(zp);
911 dmu_object_info_t doi;
913 uint64_t obj_num = zp->z_id;
915 sa_bulk_attr_t bulk[8];
920 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
922 mutex_enter(&zp->z_acl_lock);
923 if (zp->z_acl_cached) {
924 zfs_acl_free(zp->z_acl_cached);
925 zp->z_acl_cached = NULL;
928 mutex_exit(&zp->z_acl_lock);
929 ASSERT(zp->z_sa_hdl == NULL);
930 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
932 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
936 dmu_object_info_from_db(db, &doi);
937 if (doi.doi_bonus_type != DMU_OT_SA &&
938 (doi.doi_bonus_type != DMU_OT_ZNODE ||
939 (doi.doi_bonus_type == DMU_OT_ZNODE &&
940 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
941 sa_buf_rele(db, NULL);
942 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
946 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
948 /* reload cached values */
949 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
951 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
952 &zp->z_size, sizeof (zp->z_size));
953 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
954 &zp->z_links, sizeof (zp->z_links));
955 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
956 &zp->z_pflags, sizeof (zp->z_pflags));
957 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
958 &zp->z_atime, sizeof (zp->z_atime));
959 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
960 &zp->z_uid, sizeof (zp->z_uid));
961 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
962 &zp->z_gid, sizeof (zp->z_gid));
963 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
964 &mode, sizeof (mode));
966 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
967 zfs_znode_dmu_fini(zp);
968 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
974 if (gen != zp->z_gen) {
975 zfs_znode_dmu_fini(zp);
976 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
980 zp->z_unlinked = (zp->z_links == 0);
981 zp->z_blksz = doi.doi_data_block_size;
983 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
989 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
991 zfs_sb_t *zsb = ZTOZSB(zp);
992 objset_t *os = zsb->z_os;
993 uint64_t obj = zp->z_id;
994 uint64_t acl_obj = zfs_external_acl(zp);
996 ZFS_OBJ_HOLD_ENTER(zsb, obj);
998 VERIFY(!zp->z_is_sa);
999 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1001 VERIFY(0 == dmu_object_free(os, obj, tx));
1002 zfs_znode_dmu_fini(zp);
1003 ZFS_OBJ_HOLD_EXIT(zsb, obj);
1007 zfs_zinactive(znode_t *zp)
1009 zfs_sb_t *zsb = ZTOZSB(zp);
1010 uint64_t z_id = zp->z_id;
1011 boolean_t drop_mutex = 0;
1013 ASSERT(zp->z_sa_hdl);
1016 * Don't allow a zfs_zget() while were trying to release this znode.
1018 * Linux allows direct memory reclaim which means that any KM_SLEEP
1019 * allocation may trigger inode eviction. This can lead to a deadlock
1020 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
1021 * zfs_zinactive() call path. To avoid this deadlock the process
1022 * must not reacquire the mutex when it is already holding it.
1024 if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) {
1025 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
1029 mutex_enter(&zp->z_lock);
1032 * If this was the last reference to a file with no links,
1033 * remove the file from the file system.
1035 if (zp->z_unlinked) {
1036 mutex_exit(&zp->z_lock);
1039 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1045 mutex_exit(&zp->z_lock);
1046 zfs_znode_dmu_fini(zp);
1049 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1053 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1054 uint64_t ctime[2], boolean_t have_tx)
1060 if (have_tx) { /* will sa_bulk_update happen really soon? */
1061 zp->z_atime_dirty = 0;
1064 zp->z_atime_dirty = 1;
1067 if (flag & ATTR_ATIME) {
1068 ZFS_TIME_ENCODE(&now, zp->z_atime);
1071 if (flag & ATTR_MTIME) {
1072 ZFS_TIME_ENCODE(&now, mtime);
1073 if (ZTOZSB(zp)->z_use_fuids) {
1074 zp->z_pflags |= (ZFS_ARCHIVE |
1079 if (flag & ATTR_CTIME) {
1080 ZFS_TIME_ENCODE(&now, ctime);
1081 if (ZTOZSB(zp)->z_use_fuids)
1082 zp->z_pflags |= ZFS_ARCHIVE;
1087 * Grow the block size for a file.
1089 * IN: zp - znode of file to free data in.
1090 * size - requested block size
1091 * tx - open transaction.
1093 * NOTE: this function assumes that the znode is write locked.
1096 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1101 if (size <= zp->z_blksz)
1104 * If the file size is already greater than the current blocksize,
1105 * we will not grow. If there is more than one block in a file,
1106 * the blocksize cannot change.
1108 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1111 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1114 if (error == ENOTSUP)
1116 ASSERT3U(error, ==, 0);
1118 /* What blocksize did we actually get? */
1119 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1124 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1125 * be calling back into the fs for a putpage(). E.g.: when truncating
1126 * a file, the pages being "thrown away* don't need to be written out.
1130 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1131 int flags, cred_t *cr)
1136 #endif /* HAVE_MMAP */
1139 * Increase the file length
1141 * IN: zp - znode of file to free data in.
1142 * end - new end-of-file
1144 * RETURN: 0 if success
1145 * error code if failure
1148 zfs_extend(znode_t *zp, uint64_t end)
1150 zfs_sb_t *zsb = ZTOZSB(zp);
1157 * We will change zp_size, lock the whole file.
1159 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1162 * Nothing to do if file already at desired length.
1164 if (end <= zp->z_size) {
1165 zfs_range_unlock(rl);
1169 tx = dmu_tx_create(zsb->z_os);
1170 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1171 zfs_sa_upgrade_txholds(tx, zp);
1172 if (end > zp->z_blksz &&
1173 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1175 * We are growing the file past the current block size.
1177 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1178 ASSERT(!ISP2(zp->z_blksz));
1179 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1181 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1183 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1188 error = dmu_tx_assign(tx, TXG_NOWAIT);
1190 if (error == ERESTART) {
1196 zfs_range_unlock(rl);
1201 zfs_grow_blocksize(zp, newblksz, tx);
1205 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1206 &zp->z_size, sizeof (zp->z_size), tx));
1208 zfs_range_unlock(rl);
1216 * Free space in a file.
1218 * IN: zp - znode of file to free data in.
1219 * off - start of section to free.
1220 * len - length of section to free.
1222 * RETURN: 0 if success
1223 * error code if failure
1226 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1228 zfs_sb_t *zsb = ZTOZSB(zp);
1233 * Lock the range being freed.
1235 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1238 * Nothing to do if file already at desired length.
1240 if (off >= zp->z_size) {
1241 zfs_range_unlock(rl);
1245 if (off + len > zp->z_size)
1246 len = zp->z_size - off;
1248 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1250 zfs_range_unlock(rl);
1258 * IN: zp - znode of file to free data in.
1259 * end - new end-of-file.
1261 * RETURN: 0 if success
1262 * error code if failure
1265 zfs_trunc(znode_t *zp, uint64_t end)
1267 zfs_sb_t *zsb = ZTOZSB(zp);
1271 sa_bulk_attr_t bulk[2];
1275 * We will change zp_size, lock the whole file.
1277 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1280 * Nothing to do if file already at desired length.
1282 if (end >= zp->z_size) {
1283 zfs_range_unlock(rl);
1287 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1289 zfs_range_unlock(rl);
1293 tx = dmu_tx_create(zsb->z_os);
1294 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1295 zfs_sa_upgrade_txholds(tx, zp);
1296 error = dmu_tx_assign(tx, TXG_NOWAIT);
1298 if (error == ERESTART) {
1304 zfs_range_unlock(rl);
1309 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1310 NULL, &zp->z_size, sizeof (zp->z_size));
1313 zp->z_pflags &= ~ZFS_SPARSE;
1314 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1315 NULL, &zp->z_pflags, 8);
1317 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1321 zfs_range_unlock(rl);
1327 * Free space in a file
1329 * IN: zp - znode of file to free data in.
1330 * off - start of range
1331 * len - end of range (0 => EOF)
1332 * flag - current file open mode flags.
1333 * log - TRUE if this action should be logged
1335 * RETURN: 0 if success
1336 * error code if failure
1339 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1341 #ifdef HAVE_MANDLOCKS
1342 struct inode *ip = ZTOI(zp);
1343 #endif /* HAVE_MANDLOCKS */
1345 zfs_sb_t *zsb = ZTOZSB(zp);
1346 zilog_t *zilog = zsb->z_log;
1348 uint64_t mtime[2], ctime[2];
1349 sa_bulk_attr_t bulk[3];
1353 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1354 sizeof (mode))) != 0)
1357 if (off > zp->z_size) {
1358 error = zfs_extend(zp, off+len);
1359 if (error == 0 && log)
1365 #ifdef HAVE_MANDLOCKS
1367 * Check for any locks in the region to be freed.
1370 if (MANDLOCK(ip, (mode_t)mode)) {
1371 uint64_t length = (len ? len : zp->z_size - off);
1372 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1375 #endif /* HAVE_MANDLOCKS */
1378 error = zfs_trunc(zp, off);
1380 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1381 off + len > zp->z_size)
1382 error = zfs_extend(zp, off+len);
1387 tx = dmu_tx_create(zsb->z_os);
1388 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1389 zfs_sa_upgrade_txholds(tx, zp);
1390 error = dmu_tx_assign(tx, TXG_NOWAIT);
1392 if (error == ERESTART) {
1401 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1402 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1403 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1404 NULL, &zp->z_pflags, 8);
1405 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1406 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1409 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1412 zfs_inode_update(zp);
1417 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1419 uint64_t moid, obj, sa_obj, version;
1428 * First attempt to create master node.
1431 * In an empty objset, there are no blocks to read and thus
1432 * there can be no i/o errors (which we assert below).
1434 moid = MASTER_NODE_OBJ;
1435 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1436 DMU_OT_NONE, 0, tx);
1440 * Set starting attributes.
1442 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1444 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1445 /* For the moment we expect all zpl props to be uint64_ts */
1449 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1450 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1451 name = nvpair_name(elem);
1452 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1456 error = zap_update(os, moid, name, 8, 1, &val, tx);
1459 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1462 ASSERT(version != 0);
1463 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1466 * Create zap object used for SA attribute registration
1469 if (version >= ZPL_VERSION_SA) {
1470 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1471 DMU_OT_NONE, 0, tx);
1472 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1478 * Create a delete queue.
1480 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1482 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1486 * Create root znode with code free of VFS dependencies. This
1487 * is important because without a registered filesystem and super
1488 * block all the required VFS hooks will be missing. The critical
1489 * thing is to just crete the required root znode.
1491 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1492 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1494 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1495 dmu_buf_will_dirty(db, tx);
1498 * Initialize the znode physical data to zero.
1500 ASSERT(db->db_size >= sizeof (znode_phys_t));
1501 bzero(db->db_data, db->db_size);
1504 if (USE_FUIDS(version, os))
1505 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1507 pzp->zp_size = 2; /* "." and ".." */
1509 pzp->zp_parent = obj;
1510 pzp->zp_gen = dmu_tx_get_txg(tx);
1511 pzp->zp_mode = S_IFDIR | 0755;
1512 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1516 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1517 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1518 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1519 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1521 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1524 dmu_buf_rele(db, FTAG);
1527 #endif /* _KERNEL */
1530 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1532 uint64_t sa_obj = 0;
1535 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1536 if (error != 0 && error != ENOENT)
1539 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1544 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1547 dmu_object_info_t doi;
1550 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1553 dmu_object_info_from_db(*db, &doi);
1554 if ((doi.doi_bonus_type != DMU_OT_SA &&
1555 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1556 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1557 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1558 sa_buf_rele(*db, FTAG);
1562 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1564 sa_buf_rele(*db, FTAG);
1572 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1574 sa_handle_destroy(hdl);
1575 sa_buf_rele(db, FTAG);
1579 * Given an object number, return its parent object number and whether
1580 * or not the object is an extended attribute directory.
1583 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1589 sa_bulk_attr_t bulk[3];
1593 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1594 &parent, sizeof (parent));
1595 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1596 &pflags, sizeof (pflags));
1597 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1598 &mode, sizeof (mode));
1600 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1604 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1610 * Given an object number, return some zpl level statistics
1613 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1616 sa_bulk_attr_t bulk[4];
1619 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1620 &sb->zs_mode, sizeof (sb->zs_mode));
1621 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1622 &sb->zs_gen, sizeof (sb->zs_gen));
1623 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1624 &sb->zs_links, sizeof (sb->zs_links));
1625 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1626 &sb->zs_ctime, sizeof (sb->zs_ctime));
1628 return (sa_bulk_lookup(hdl, bulk, count));
1632 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1633 sa_attr_type_t *sa_table, char *buf, int len)
1635 sa_handle_t *sa_hdl;
1636 sa_handle_t *prevhdl = NULL;
1637 dmu_buf_t *prevdb = NULL;
1638 dmu_buf_t *sa_db = NULL;
1639 char *path = buf + len - 1;
1647 char component[MAXNAMELEN + 2];
1652 zfs_release_sa_handle(prevhdl, prevdb);
1654 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1655 &is_xattrdir)) != 0)
1666 (void) sprintf(component + 1, "<xattrdir>");
1668 error = zap_value_search(osp, pobj, obj,
1669 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1674 complen = strlen(component);
1676 ASSERT(path >= buf);
1677 bcopy(component, path, complen);
1680 if (sa_hdl != hdl) {
1684 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1692 if (sa_hdl != NULL && sa_hdl != hdl) {
1693 ASSERT(sa_db != NULL);
1694 zfs_release_sa_handle(sa_hdl, sa_db);
1698 (void) memmove(buf, path, buf + len - path);
1704 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1706 sa_attr_type_t *sa_table;
1711 error = zfs_sa_setup(osp, &sa_table);
1715 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1719 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1721 zfs_release_sa_handle(hdl, db);
1726 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1729 char *path = buf + len - 1;
1730 sa_attr_type_t *sa_table;
1737 error = zfs_sa_setup(osp, &sa_table);
1741 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1745 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1747 zfs_release_sa_handle(hdl, db);
1751 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1753 zfs_release_sa_handle(hdl, db);
1757 #if defined(_KERNEL) && defined(HAVE_SPL)
1758 EXPORT_SYMBOL(zfs_create_fs);
1759 EXPORT_SYMBOL(zfs_obj_to_path);