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/zfs_ctldir.h>
56 #include <sys/dnode.h>
57 #include <sys/fs/zfs.h>
58 #include <sys/kidmap.h>
63 #include <sys/refcount.h>
66 #include <sys/zfs_znode.h>
68 #include <sys/zfs_sa.h>
69 #include <sys/zfs_stat.h>
72 #include "zfs_comutil.h"
75 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
76 * turned on when DEBUG is also defined.
83 #define ZNODE_STAT_ADD(stat) ((stat)++)
85 #define ZNODE_STAT_ADD(stat) /* nothing */
86 #endif /* ZNODE_STATS */
89 * Functions needed for userland (ie: libzpool) are not put under
90 * #ifdef_KERNEL; the rest of the functions have dependencies
91 * (such as VFS logic) that will not compile easily in userland.
95 static kmem_cache_t *znode_cache = NULL;
99 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
103 inode_init_once(ZTOI(zp));
104 list_link_init(&zp->z_link_node);
106 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
107 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
108 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
109 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
110 rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
112 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
113 avl_create(&zp->z_range_avl, zfs_range_compare,
114 sizeof (rl_t), offsetof(rl_t, r_node));
116 zp->z_dirlocks = NULL;
117 zp->z_acl_cached = NULL;
118 zp->z_xattr_cached = NULL;
119 zp->z_xattr_parent = NULL;
126 zfs_znode_cache_destructor(void *buf, void *arg)
130 ASSERT(!list_link_active(&zp->z_link_node));
131 mutex_destroy(&zp->z_lock);
132 rw_destroy(&zp->z_parent_lock);
133 rw_destroy(&zp->z_name_lock);
134 mutex_destroy(&zp->z_acl_lock);
135 rw_destroy(&zp->z_xattr_lock);
136 avl_destroy(&zp->z_range_avl);
137 mutex_destroy(&zp->z_range_lock);
139 ASSERT(zp->z_dirlocks == NULL);
140 ASSERT(zp->z_acl_cached == NULL);
141 ASSERT(zp->z_xattr_cached == NULL);
142 ASSERT(zp->z_xattr_parent == NULL);
151 ASSERT(znode_cache == NULL);
152 znode_cache = kmem_cache_create("zfs_znode_cache",
153 sizeof (znode_t), 0, zfs_znode_cache_constructor,
154 zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_KMEM);
164 kmem_cache_destroy(znode_cache);
169 zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx)
171 #ifdef HAVE_SMB_SHARE
172 zfs_acl_ids_t acl_ids;
179 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
180 vattr.va_mode = S_IFDIR | 0555;
181 vattr.va_uid = crgetuid(kcred);
182 vattr.va_gid = crgetgid(kcred);
184 sharezp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
185 sharezp->z_moved = 0;
186 sharezp->z_unlinked = 0;
187 sharezp->z_atime_dirty = 0;
188 sharezp->z_zfsvfs = zfsvfs;
189 sharezp->z_is_sa = zfsvfs->z_use_sa;
195 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
196 kcred, NULL, &acl_ids));
197 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
198 ASSERT3P(zp, ==, sharezp);
199 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
200 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
201 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
202 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
203 zfsvfs->z_shares_dir = sharezp->z_id;
205 zfs_acl_ids_free(&acl_ids);
206 // ZTOV(sharezp)->v_count = 0;
207 sa_handle_destroy(sharezp->z_sa_hdl);
208 kmem_cache_free(znode_cache, sharezp);
213 #endif /* HAVE_SMB_SHARE */
217 zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp,
218 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
220 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id)));
222 mutex_enter(&zp->z_lock);
224 ASSERT(zp->z_sa_hdl == NULL);
225 ASSERT(zp->z_acl_cached == NULL);
226 if (sa_hdl == NULL) {
227 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp,
228 SA_HDL_SHARED, &zp->z_sa_hdl));
230 zp->z_sa_hdl = sa_hdl;
231 sa_set_userp(sa_hdl, zp);
234 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
236 mutex_exit(&zp->z_lock);
240 zfs_znode_dmu_fini(znode_t *zp)
242 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) ||
244 RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
246 sa_handle_destroy(zp->z_sa_hdl);
251 * Called by new_inode() to allocate a new inode.
254 zfs_inode_alloc(struct super_block *sb, struct inode **ip)
258 zp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
265 * Called in multiple places when an inode should be destroyed.
268 zfs_inode_destroy(struct inode *ip)
270 znode_t *zp = ITOZ(ip);
271 zfs_sb_t *zsb = ZTOZSB(zp);
273 if (zfsctl_is_node(ip))
274 zfsctl_inode_destroy(ip);
276 mutex_enter(&zsb->z_znodes_lock);
277 if (list_link_active(&zp->z_link_node)) {
278 list_remove(&zsb->z_all_znodes, zp);
281 mutex_exit(&zsb->z_znodes_lock);
283 if (zp->z_acl_cached) {
284 zfs_acl_free(zp->z_acl_cached);
285 zp->z_acl_cached = NULL;
288 if (zp->z_xattr_cached) {
289 nvlist_free(zp->z_xattr_cached);
290 zp->z_xattr_cached = NULL;
293 if (zp->z_xattr_parent) {
294 iput(ZTOI(zp->z_xattr_parent));
295 zp->z_xattr_parent = NULL;
298 kmem_cache_free(znode_cache, zp);
302 zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
306 switch (ip->i_mode & S_IFMT) {
308 ip->i_op = &zpl_inode_operations;
309 ip->i_fop = &zpl_file_operations;
310 ip->i_mapping->a_ops = &zpl_address_space_operations;
314 ip->i_op = &zpl_dir_inode_operations;
315 ip->i_fop = &zpl_dir_file_operations;
316 ITOZ(ip)->z_zn_prefetch = B_TRUE;
320 ip->i_op = &zpl_symlink_inode_operations;
324 * rdev is only stored in a SA only for device files.
328 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
329 &rdev, sizeof (rdev)) == 0);
333 init_special_inode(ip, ip->i_mode, rdev);
334 ip->i_op = &zpl_special_inode_operations;
338 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
344 * Construct a znode+inode and initialize.
346 * This does not do a call to dmu_set_user() that is
347 * up to the caller to do, in case you don't want to
351 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
352 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
358 sa_bulk_attr_t bulk[9];
363 ip = new_inode(zsb->z_sb);
368 ASSERT(zp->z_dirlocks == NULL);
369 ASSERT3P(zp->z_acl_cached, ==, NULL);
370 ASSERT3P(zp->z_xattr_cached, ==, NULL);
371 ASSERT3P(zp->z_xattr_parent, ==, NULL);
375 zp->z_atime_dirty = 0;
377 zp->z_id = db->db_object;
379 zp->z_seq = 0x7A4653;
381 zp->z_is_zvol = B_FALSE;
382 zp->z_is_mapped = B_FALSE;
383 zp->z_is_ctldir = B_FALSE;
384 zp->z_is_stale = B_FALSE;
386 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
388 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
389 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
390 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
391 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
392 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
394 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
396 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
398 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
399 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
401 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
403 sa_handle_destroy(zp->z_sa_hdl);
409 * xattr znodes hold a reference on their unique parent
411 if (dip && zp->z_pflags & ZFS_XATTR) {
413 zp->z_xattr_parent = ITOZ(dip);
417 zfs_inode_update(zp);
418 zfs_inode_set_ops(zsb, ip);
421 * The only way insert_inode_locked() can fail is if the ip->i_ino
422 * number is already hashed for this super block. This can never
423 * happen because the inode numbers map 1:1 with the object numbers.
425 * The one exception is rolling back a mounted file system, but in
426 * this case all the active inode are unhashed during the rollback.
428 VERIFY3S(insert_inode_locked(ip), ==, 0);
430 mutex_enter(&zsb->z_znodes_lock);
431 list_insert_tail(&zsb->z_all_znodes, zp);
434 mutex_exit(&zsb->z_znodes_lock);
436 unlock_new_inode(ip);
440 unlock_new_inode(ip);
446 * Update the embedded inode given the znode. We should work toward
447 * eliminating this function as soon as possible by removing values
448 * which are duplicated between the znode and inode. If the generic
449 * inode has the correct field it should be used, and the ZFS code
450 * updated to access the inode. This can be done incrementally.
453 zfs_inode_update(znode_t *zp)
458 uint64_t atime[2], mtime[2], ctime[2];
464 /* Skip .zfs control nodes which do not exist on disk. */
465 if (zfsctl_is_node(ip))
468 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
469 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
470 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
472 spin_lock(&ip->i_lock);
473 ip->i_generation = zp->z_gen;
474 ip->i_uid = zp->z_uid;
475 ip->i_gid = zp->z_gid;
476 set_nlink(ip, zp->z_links);
477 ip->i_mode = zp->z_mode;
478 ip->i_blkbits = SPA_MINBLOCKSHIFT;
479 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
480 (u_longlong_t *)&ip->i_blocks);
482 ZFS_TIME_DECODE(&ip->i_atime, atime);
483 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
484 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
486 i_size_write(ip, zp->z_size);
487 spin_unlock(&ip->i_lock);
490 static uint64_t empty_xattr;
491 static uint64_t pad[4];
492 static zfs_acl_phys_t acl_phys;
494 * Create a new DMU object to hold a zfs znode.
496 * IN: dzp - parent directory for new znode
497 * vap - file attributes for new znode
498 * tx - dmu transaction id for zap operations
499 * cr - credentials of caller
501 * IS_ROOT_NODE - new object will be root
502 * IS_XATTR - new object is an attribute
503 * bonuslen - length of bonus buffer
504 * setaclp - File/Dir initial ACL
505 * fuidp - Tracks fuid allocation.
507 * OUT: zpp - allocated znode
511 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
512 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
514 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
515 uint64_t mode, size, links, parent, pflags;
516 uint64_t dzp_pflags = 0;
518 zfs_sb_t *zsb = ZTOZSB(dzp);
525 dmu_object_type_t obj_type;
526 sa_bulk_attr_t *sa_attrs;
528 zfs_acl_locator_cb_t locate = { 0 };
531 obj = vap->va_nodeid;
532 now = vap->va_ctime; /* see zfs_replay_create() */
533 gen = vap->va_nblocks; /* ditto */
537 gen = dmu_tx_get_txg(tx);
540 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
541 bonuslen = (obj_type == DMU_OT_SA) ?
542 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
545 * Create a new DMU object.
548 * There's currently no mechanism for pre-reading the blocks that will
549 * be needed to allocate a new object, so we accept the small chance
550 * that there will be an i/o error and we will fail one of the
553 if (S_ISDIR(vap->va_mode)) {
555 err = zap_create_claim_norm(zsb->z_os, obj,
556 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
557 obj_type, bonuslen, tx);
558 ASSERT3U(err, ==, 0);
560 obj = zap_create_norm(zsb->z_os,
561 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
562 obj_type, bonuslen, tx);
566 err = dmu_object_claim(zsb->z_os, obj,
567 DMU_OT_PLAIN_FILE_CONTENTS, 0,
568 obj_type, bonuslen, tx);
569 ASSERT3U(err, ==, 0);
571 obj = dmu_object_alloc(zsb->z_os,
572 DMU_OT_PLAIN_FILE_CONTENTS, 0,
573 obj_type, bonuslen, tx);
577 ZFS_OBJ_HOLD_ENTER(zsb, obj);
578 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
581 * If this is the root, fix up the half-initialized parent pointer
582 * to reference the just-allocated physical data area.
584 if (flag & IS_ROOT_NODE) {
587 dzp_pflags = dzp->z_pflags;
591 * If parent is an xattr, so am I.
593 if (dzp_pflags & ZFS_XATTR) {
597 if (zsb->z_use_fuids)
598 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
602 if (S_ISDIR(vap->va_mode)) {
603 size = 2; /* contents ("." and "..") */
604 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
609 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))
613 mode = acl_ids->z_mode;
618 * No execs denied will be deterimed when zfs_mode_compute() is called.
620 pflags |= acl_ids->z_aclp->z_hints &
621 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
622 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
624 ZFS_TIME_ENCODE(&now, crtime);
625 ZFS_TIME_ENCODE(&now, ctime);
627 if (vap->va_mask & ATTR_ATIME) {
628 ZFS_TIME_ENCODE(&vap->va_atime, atime);
630 ZFS_TIME_ENCODE(&now, atime);
633 if (vap->va_mask & ATTR_MTIME) {
634 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
636 ZFS_TIME_ENCODE(&now, mtime);
639 /* Now add in all of the "SA" attributes */
640 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
644 * Setup the array of attributes to be replaced/set on the new file
646 * order for DMU_OT_ZNODE is critical since it needs to be constructed
647 * in the old znode_phys_t format. Don't change this ordering
649 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_PUSHPAGE);
651 if (obj_type == DMU_OT_ZNODE) {
652 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
654 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
656 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
658 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
660 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
662 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
664 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
666 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
669 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
671 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
673 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
675 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
676 NULL, &acl_ids->z_fuid, 8);
677 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
678 NULL, &acl_ids->z_fgid, 8);
679 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
681 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
683 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
685 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
687 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
689 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
693 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
695 if (obj_type == DMU_OT_ZNODE) {
696 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
699 if (obj_type == DMU_OT_ZNODE ||
700 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) {
701 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
704 if (obj_type == DMU_OT_ZNODE) {
705 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
707 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
708 &acl_ids->z_fuid, 8);
709 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
710 &acl_ids->z_fgid, 8);
711 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
712 sizeof (uint64_t) * 4);
713 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
714 &acl_phys, sizeof (zfs_acl_phys_t));
715 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
716 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
717 &acl_ids->z_aclp->z_acl_count, 8);
718 locate.cb_aclp = acl_ids->z_aclp;
719 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
720 zfs_acl_data_locator, &locate,
721 acl_ids->z_aclp->z_acl_bytes);
722 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
723 acl_ids->z_fuid, acl_ids->z_fgid);
726 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
728 if (!(flag & IS_ROOT_NODE)) {
729 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
731 VERIFY(*zpp != NULL);
735 * If we are creating the root node, the "parent" we
736 * passed in is the znode for the root.
740 (*zpp)->z_sa_hdl = sa_hdl;
743 (*zpp)->z_pflags = pflags;
744 (*zpp)->z_mode = mode;
746 if (obj_type == DMU_OT_ZNODE ||
747 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
748 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
749 ASSERT3S(err, ==, 0);
751 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
752 ZFS_OBJ_HOLD_EXIT(zsb, obj);
756 * zfs_xvattr_set only updates the in-core attributes
757 * it is assumed the caller will be doing an sa_bulk_update
758 * to push the changes out
761 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
765 xoap = xva_getxoptattr(xvap);
768 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
770 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
771 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
772 ×, sizeof (times), tx);
773 XVA_SET_RTN(xvap, XAT_CREATETIME);
775 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
776 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
778 XVA_SET_RTN(xvap, XAT_READONLY);
780 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
781 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
783 XVA_SET_RTN(xvap, XAT_HIDDEN);
785 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
786 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
788 XVA_SET_RTN(xvap, XAT_SYSTEM);
790 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
791 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
793 XVA_SET_RTN(xvap, XAT_ARCHIVE);
795 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
796 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
798 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
800 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
801 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
803 XVA_SET_RTN(xvap, XAT_NOUNLINK);
805 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
806 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
808 XVA_SET_RTN(xvap, XAT_APPENDONLY);
810 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
811 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
813 XVA_SET_RTN(xvap, XAT_NODUMP);
815 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
816 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
818 XVA_SET_RTN(xvap, XAT_OPAQUE);
820 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
821 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
822 xoap->xoa_av_quarantined, zp->z_pflags, tx);
823 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
825 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
826 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
828 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
830 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
831 zfs_sa_set_scanstamp(zp, xvap, tx);
832 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
834 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
835 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
837 XVA_SET_RTN(xvap, XAT_REPARSE);
839 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
840 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
842 XVA_SET_RTN(xvap, XAT_OFFLINE);
844 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
845 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
847 XVA_SET_RTN(xvap, XAT_SPARSE);
852 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
854 dmu_object_info_t doi;
864 ip = ilookup(zsb->z_sb, obj_num);
866 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
868 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
870 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
875 dmu_object_info_from_db(db, &doi);
876 if (doi.doi_bonus_type != DMU_OT_SA &&
877 (doi.doi_bonus_type != DMU_OT_ZNODE ||
878 (doi.doi_bonus_type == DMU_OT_ZNODE &&
879 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
880 sa_buf_rele(db, NULL);
881 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
886 hdl = dmu_buf_get_user(db);
890 * ilookup returned NULL, which means
891 * the znode is dying - but the SA handle isn't
892 * quite dead yet, we need to drop any locks
893 * we're holding, re-schedule the task and try again.
895 sa_buf_rele(db, NULL);
896 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
902 zp = sa_get_userdata(hdl);
905 * Since "SA" does immediate eviction we
906 * should never find a sa handle that doesn't
907 * know about the znode.
910 ASSERT3P(zp, !=, NULL);
912 mutex_enter(&zp->z_lock);
913 ASSERT3U(zp->z_id, ==, obj_num);
914 if (zp->z_unlinked) {
921 sa_buf_rele(db, NULL);
922 mutex_exit(&zp->z_lock);
923 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
928 ASSERT3P(ip, ==, NULL);
931 * Not found create new znode/vnode but only if file exists.
933 * There is a small window where zfs_vget() could
934 * find this object while a file create is still in
935 * progress. This is checked for in zfs_znode_alloc()
937 * if zfs_znode_alloc() fails it will drop the hold on the
940 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
941 doi.doi_bonus_type, obj_num, NULL, NULL);
947 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
952 zfs_rezget(znode_t *zp)
954 zfs_sb_t *zsb = ZTOZSB(zp);
955 dmu_object_info_t doi;
957 uint64_t obj_num = zp->z_id;
959 sa_bulk_attr_t bulk[8];
964 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
966 mutex_enter(&zp->z_acl_lock);
967 if (zp->z_acl_cached) {
968 zfs_acl_free(zp->z_acl_cached);
969 zp->z_acl_cached = NULL;
971 mutex_exit(&zp->z_acl_lock);
973 rw_enter(&zp->z_xattr_lock, RW_WRITER);
974 if (zp->z_xattr_cached) {
975 nvlist_free(zp->z_xattr_cached);
976 zp->z_xattr_cached = NULL;
979 if (zp->z_xattr_parent) {
980 iput(ZTOI(zp->z_xattr_parent));
981 zp->z_xattr_parent = NULL;
983 rw_exit(&zp->z_xattr_lock);
985 ASSERT(zp->z_sa_hdl == NULL);
986 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
988 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
992 dmu_object_info_from_db(db, &doi);
993 if (doi.doi_bonus_type != DMU_OT_SA &&
994 (doi.doi_bonus_type != DMU_OT_ZNODE ||
995 (doi.doi_bonus_type == DMU_OT_ZNODE &&
996 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
997 sa_buf_rele(db, NULL);
998 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
1002 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
1004 /* reload cached values */
1005 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
1006 &gen, sizeof (gen));
1007 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
1008 &zp->z_size, sizeof (zp->z_size));
1009 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
1010 &zp->z_links, sizeof (zp->z_links));
1011 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
1012 &zp->z_pflags, sizeof (zp->z_pflags));
1013 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
1014 &zp->z_atime, sizeof (zp->z_atime));
1015 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
1016 &zp->z_uid, sizeof (zp->z_uid));
1017 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
1018 &zp->z_gid, sizeof (zp->z_gid));
1019 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
1020 &mode, sizeof (mode));
1022 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1023 zfs_znode_dmu_fini(zp);
1024 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
1030 if (gen != zp->z_gen) {
1031 zfs_znode_dmu_fini(zp);
1032 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
1036 zp->z_unlinked = (zp->z_links == 0);
1037 zp->z_blksz = doi.doi_data_block_size;
1038 zfs_inode_update(zp);
1040 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
1046 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1048 zfs_sb_t *zsb = ZTOZSB(zp);
1049 objset_t *os = zsb->z_os;
1050 uint64_t obj = zp->z_id;
1051 uint64_t acl_obj = zfs_external_acl(zp);
1053 ZFS_OBJ_HOLD_ENTER(zsb, obj);
1055 VERIFY(!zp->z_is_sa);
1056 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1058 VERIFY(0 == dmu_object_free(os, obj, tx));
1059 zfs_znode_dmu_fini(zp);
1060 ZFS_OBJ_HOLD_EXIT(zsb, obj);
1064 zfs_zinactive(znode_t *zp)
1066 zfs_sb_t *zsb = ZTOZSB(zp);
1067 uint64_t z_id = zp->z_id;
1068 boolean_t drop_mutex = 0;
1070 ASSERT(zp->z_sa_hdl);
1073 * Don't allow a zfs_zget() while were trying to release this znode.
1075 * Linux allows direct memory reclaim which means that any KM_SLEEP
1076 * allocation may trigger inode eviction. This can lead to a deadlock
1077 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
1078 * zfs_zinactive() call path. To avoid this deadlock the process
1079 * must not reacquire the mutex when it is already holding it.
1081 if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) {
1082 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
1086 mutex_enter(&zp->z_lock);
1089 * If this was the last reference to a file with no links,
1090 * remove the file from the file system.
1092 if (zp->z_unlinked) {
1093 mutex_exit(&zp->z_lock);
1096 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1102 mutex_exit(&zp->z_lock);
1103 zfs_znode_dmu_fini(zp);
1106 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1110 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1111 uint64_t ctime[2], boolean_t have_tx)
1117 if (have_tx) { /* will sa_bulk_update happen really soon? */
1118 zp->z_atime_dirty = 0;
1121 zp->z_atime_dirty = 1;
1124 if (flag & ATTR_ATIME) {
1125 ZFS_TIME_ENCODE(&now, zp->z_atime);
1128 if (flag & ATTR_MTIME) {
1129 ZFS_TIME_ENCODE(&now, mtime);
1130 if (ZTOZSB(zp)->z_use_fuids) {
1131 zp->z_pflags |= (ZFS_ARCHIVE |
1136 if (flag & ATTR_CTIME) {
1137 ZFS_TIME_ENCODE(&now, ctime);
1138 if (ZTOZSB(zp)->z_use_fuids)
1139 zp->z_pflags |= ZFS_ARCHIVE;
1144 * Grow the block size for a file.
1146 * IN: zp - znode of file to free data in.
1147 * size - requested block size
1148 * tx - open transaction.
1150 * NOTE: this function assumes that the znode is write locked.
1153 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1158 if (size <= zp->z_blksz)
1161 * If the file size is already greater than the current blocksize,
1162 * we will not grow. If there is more than one block in a file,
1163 * the blocksize cannot change.
1165 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1168 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1171 if (error == ENOTSUP)
1173 ASSERT3U(error, ==, 0);
1175 /* What blocksize did we actually get? */
1176 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1180 * Increase the file length
1182 * IN: zp - znode of file to free data in.
1183 * end - new end-of-file
1185 * RETURN: 0 if success
1186 * error code if failure
1189 zfs_extend(znode_t *zp, uint64_t end)
1191 zfs_sb_t *zsb = ZTOZSB(zp);
1198 * We will change zp_size, lock the whole file.
1200 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1203 * Nothing to do if file already at desired length.
1205 if (end <= zp->z_size) {
1206 zfs_range_unlock(rl);
1210 tx = dmu_tx_create(zsb->z_os);
1211 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1212 zfs_sa_upgrade_txholds(tx, zp);
1213 if (end > zp->z_blksz &&
1214 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1216 * We are growing the file past the current block size.
1218 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1219 ASSERT(!ISP2(zp->z_blksz));
1220 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1222 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1224 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1229 error = dmu_tx_assign(tx, TXG_NOWAIT);
1231 if (error == ERESTART) {
1237 zfs_range_unlock(rl);
1242 zfs_grow_blocksize(zp, newblksz, tx);
1246 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1247 &zp->z_size, sizeof (zp->z_size), tx));
1249 zfs_range_unlock(rl);
1257 * Free space in a file.
1259 * IN: zp - znode of file to free data in.
1260 * off - start of section to free.
1261 * len - length of section to free.
1263 * RETURN: 0 if success
1264 * error code if failure
1267 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1269 zfs_sb_t *zsb = ZTOZSB(zp);
1274 * Lock the range being freed.
1276 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1279 * Nothing to do if file already at desired length.
1281 if (off >= zp->z_size) {
1282 zfs_range_unlock(rl);
1286 if (off + len > zp->z_size)
1287 len = zp->z_size - off;
1289 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1291 zfs_range_unlock(rl);
1299 * IN: zp - znode of file to free data in.
1300 * end - new end-of-file.
1302 * RETURN: 0 if success
1303 * error code if failure
1306 zfs_trunc(znode_t *zp, uint64_t end)
1308 zfs_sb_t *zsb = ZTOZSB(zp);
1312 sa_bulk_attr_t bulk[2];
1316 * We will change zp_size, lock the whole file.
1318 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1321 * Nothing to do if file already at desired length.
1323 if (end >= zp->z_size) {
1324 zfs_range_unlock(rl);
1328 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1330 zfs_range_unlock(rl);
1334 tx = dmu_tx_create(zsb->z_os);
1335 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1336 zfs_sa_upgrade_txholds(tx, zp);
1337 error = dmu_tx_assign(tx, TXG_NOWAIT);
1339 if (error == ERESTART) {
1345 zfs_range_unlock(rl);
1350 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1351 NULL, &zp->z_size, sizeof (zp->z_size));
1354 zp->z_pflags &= ~ZFS_SPARSE;
1355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1356 NULL, &zp->z_pflags, 8);
1358 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1362 zfs_range_unlock(rl);
1368 * Free space in a file
1370 * IN: zp - znode of file to free data in.
1371 * off - start of range
1372 * len - end of range (0 => EOF)
1373 * flag - current file open mode flags.
1374 * log - TRUE if this action should be logged
1376 * RETURN: 0 if success
1377 * error code if failure
1380 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1382 struct inode *ip = ZTOI(zp);
1384 zfs_sb_t *zsb = ZTOZSB(zp);
1385 zilog_t *zilog = zsb->z_log;
1387 uint64_t mtime[2], ctime[2];
1388 sa_bulk_attr_t bulk[3];
1392 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1393 sizeof (mode))) != 0)
1396 if (off > zp->z_size) {
1397 error = zfs_extend(zp, off+len);
1398 if (error == 0 && log)
1405 * Check for any locks in the region to be freed.
1407 if (ip->i_flock && mandatory_lock(ip)) {
1408 uint64_t length = (len ? len : zp->z_size - off);
1409 if (!lock_may_write(ip, off, length))
1414 error = zfs_trunc(zp, off);
1416 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1417 off + len > zp->z_size)
1418 error = zfs_extend(zp, off+len);
1423 tx = dmu_tx_create(zsb->z_os);
1424 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1425 zfs_sa_upgrade_txholds(tx, zp);
1426 error = dmu_tx_assign(tx, TXG_NOWAIT);
1428 if (error == ERESTART) {
1437 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1438 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1439 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1440 NULL, &zp->z_pflags, 8);
1441 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1442 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1445 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1448 zfs_inode_update(zp);
1453 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1455 struct super_block *sb;
1457 uint64_t moid, obj, sa_obj, version;
1458 uint64_t sense = ZFS_CASE_SENSITIVE;
1463 znode_t *rootzp = NULL;
1466 zfs_acl_ids_t acl_ids;
1469 * First attempt to create master node.
1472 * In an empty objset, there are no blocks to read and thus
1473 * there can be no i/o errors (which we assert below).
1475 moid = MASTER_NODE_OBJ;
1476 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1477 DMU_OT_NONE, 0, tx);
1481 * Set starting attributes.
1483 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1485 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1486 /* For the moment we expect all zpl props to be uint64_ts */
1490 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1491 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1492 name = nvpair_name(elem);
1493 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1497 error = zap_update(os, moid, name, 8, 1, &val, tx);
1500 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1502 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1505 ASSERT(version != 0);
1506 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1509 * Create zap object used for SA attribute registration
1512 if (version >= ZPL_VERSION_SA) {
1513 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1514 DMU_OT_NONE, 0, tx);
1515 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1521 * Create a delete queue.
1523 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1525 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1529 * Create root znode. Create minimal znode/inode/zsb/sb
1530 * to allow zfs_mknode to work.
1532 vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID;
1533 vattr.va_mode = S_IFDIR|0755;
1534 vattr.va_uid = crgetuid(cr);
1535 vattr.va_gid = crgetgid(cr);
1537 rootzp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
1538 rootzp->z_moved = 0;
1539 rootzp->z_unlinked = 0;
1540 rootzp->z_atime_dirty = 0;
1541 rootzp->z_is_sa = USE_SA(version, os);
1543 zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_PUSHPAGE | KM_NODEBUG);
1545 zsb->z_parent = zsb;
1546 zsb->z_version = version;
1547 zsb->z_use_fuids = USE_FUIDS(version, os);
1548 zsb->z_use_sa = USE_SA(version, os);
1551 sb = kmem_zalloc(sizeof (struct super_block), KM_PUSHPAGE);
1552 sb->s_fs_info = zsb;
1554 ZTOI(rootzp)->i_sb = sb;
1556 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1557 &zsb->z_attr_table);
1562 * Fold case on file systems that are always or sometimes case
1565 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1566 zsb->z_norm |= U8_TEXTPREP_TOUPPER;
1568 mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1569 list_create(&zsb->z_all_znodes, sizeof (znode_t),
1570 offsetof(znode_t, z_link_node));
1572 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1573 mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1575 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1576 cr, NULL, &acl_ids));
1577 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1578 ASSERT3P(zp, ==, rootzp);
1579 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1581 zfs_acl_ids_free(&acl_ids);
1583 atomic_set(&ZTOI(rootzp)->i_count, 0);
1584 sa_handle_destroy(rootzp->z_sa_hdl);
1585 kmem_cache_free(znode_cache, rootzp);
1588 * Create shares directory
1590 error = zfs_create_share_dir(zsb, tx);
1593 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1594 mutex_destroy(&zsb->z_hold_mtx[i]);
1596 kmem_free(sb, sizeof (struct super_block));
1597 kmem_free(zsb, sizeof (zfs_sb_t));
1599 #endif /* _KERNEL */
1602 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1604 uint64_t sa_obj = 0;
1607 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1608 if (error != 0 && error != ENOENT)
1611 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1616 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1617 dmu_buf_t **db, void *tag)
1619 dmu_object_info_t doi;
1622 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1625 dmu_object_info_from_db(*db, &doi);
1626 if ((doi.doi_bonus_type != DMU_OT_SA &&
1627 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1628 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1629 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1630 sa_buf_rele(*db, tag);
1634 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1636 sa_buf_rele(*db, tag);
1644 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1646 sa_handle_destroy(hdl);
1647 sa_buf_rele(db, tag);
1651 * Given an object number, return its parent object number and whether
1652 * or not the object is an extended attribute directory.
1655 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1661 sa_bulk_attr_t bulk[3];
1665 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1666 &parent, sizeof (parent));
1667 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1668 &pflags, sizeof (pflags));
1669 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1670 &mode, sizeof (mode));
1672 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1676 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1682 * Given an object number, return some zpl level statistics
1685 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1688 sa_bulk_attr_t bulk[4];
1691 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1692 &sb->zs_mode, sizeof (sb->zs_mode));
1693 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1694 &sb->zs_gen, sizeof (sb->zs_gen));
1695 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1696 &sb->zs_links, sizeof (sb->zs_links));
1697 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1698 &sb->zs_ctime, sizeof (sb->zs_ctime));
1700 return (sa_bulk_lookup(hdl, bulk, count));
1704 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1705 sa_attr_type_t *sa_table, char *buf, int len)
1707 sa_handle_t *sa_hdl;
1708 sa_handle_t *prevhdl = NULL;
1709 dmu_buf_t *prevdb = NULL;
1710 dmu_buf_t *sa_db = NULL;
1711 char *path = buf + len - 1;
1719 char component[MAXNAMELEN + 2];
1724 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
1726 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1727 &is_xattrdir)) != 0)
1738 (void) sprintf(component + 1, "<xattrdir>");
1740 error = zap_value_search(osp, pobj, obj,
1741 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1746 complen = strlen(component);
1748 ASSERT(path >= buf);
1749 bcopy(component, path, complen);
1752 if (sa_hdl != hdl) {
1756 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
1764 if (sa_hdl != NULL && sa_hdl != hdl) {
1765 ASSERT(sa_db != NULL);
1766 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
1770 (void) memmove(buf, path, buf + len - path);
1776 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1778 sa_attr_type_t *sa_table;
1783 error = zfs_sa_setup(osp, &sa_table);
1787 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
1791 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1793 zfs_release_sa_handle(hdl, db, FTAG);
1798 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1801 char *path = buf + len - 1;
1802 sa_attr_type_t *sa_table;
1809 error = zfs_sa_setup(osp, &sa_table);
1813 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
1817 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1819 zfs_release_sa_handle(hdl, db, FTAG);
1823 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1825 zfs_release_sa_handle(hdl, db, FTAG);
1829 #if defined(_KERNEL) && defined(HAVE_SPL)
1830 EXPORT_SYMBOL(zfs_create_fs);
1831 EXPORT_SYMBOL(zfs_obj_to_path);