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/dnode.h>
55 #include <sys/fs/zfs.h>
56 #include <sys/kidmap.h>
60 #include <sys/refcount.h>
63 #include <sys/zfs_znode.h>
65 #include <sys/zfs_sa.h>
66 #include <sys/zfs_stat.h>
69 #include "zfs_comutil.h"
72 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
73 * turned on when DEBUG is also defined.
80 #define ZNODE_STAT_ADD(stat) ((stat)++)
82 #define ZNODE_STAT_ADD(stat) /* nothing */
83 #endif /* ZNODE_STATS */
86 * Functions needed for userland (ie: libzpool) are not put under
87 * #ifdef_KERNEL; the rest of the functions have dependencies
88 * (such as VFS logic) that will not compile easily in userland.
92 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
93 * be freed before it can be safely accessed.
95 krwlock_t zfsvfs_lock;
97 static kmem_cache_t *znode_cache = NULL;
101 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
105 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
107 zp->z_vnode = vn_alloc(kmflags);
108 if (zp->z_vnode == NULL) {
111 ZTOV(zp)->v_data = zp;
113 list_link_init(&zp->z_link_node);
115 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
116 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
117 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
118 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
120 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
121 avl_create(&zp->z_range_avl, zfs_range_compare,
122 sizeof (rl_t), offsetof(rl_t, r_node));
124 zp->z_dirlocks = NULL;
125 zp->z_acl_cached = NULL;
132 zfs_znode_cache_destructor(void *buf, void *arg)
136 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
137 ASSERT(ZTOV(zp)->v_data == zp);
139 ASSERT(!list_link_active(&zp->z_link_node));
140 mutex_destroy(&zp->z_lock);
141 rw_destroy(&zp->z_parent_lock);
142 rw_destroy(&zp->z_name_lock);
143 mutex_destroy(&zp->z_acl_lock);
144 avl_destroy(&zp->z_range_avl);
145 mutex_destroy(&zp->z_range_lock);
147 ASSERT(zp->z_dirlocks == NULL);
148 ASSERT(zp->z_acl_cached == NULL);
157 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
158 ASSERT(znode_cache == NULL);
159 znode_cache = kmem_cache_create("zfs_znode_cache",
160 sizeof (znode_t), 0, zfs_znode_cache_constructor,
161 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
171 kmem_cache_destroy(znode_cache);
173 rw_destroy(&zfsvfs_lock);
178 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
181 zfs_acl_ids_t acl_ids;
188 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
189 vattr.va_type = VDIR;
190 vattr.va_mode = S_IFDIR|0555;
191 vattr.va_uid = crgetuid(kcred);
192 vattr.va_gid = crgetgid(kcred);
194 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
195 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
196 sharezp->z_moved = 0;
197 sharezp->z_unlinked = 0;
198 sharezp->z_atime_dirty = 0;
199 sharezp->z_zfsvfs = zfsvfs;
200 sharezp->z_is_sa = zfsvfs->z_use_sa;
206 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
207 kcred, NULL, &acl_ids));
208 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
209 ASSERT3P(zp, ==, sharezp);
210 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
211 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
212 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
213 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
214 zfsvfs->z_shares_dir = sharezp->z_id;
216 zfs_acl_ids_free(&acl_ids);
217 ZTOV(sharezp)->v_count = 0;
218 sa_handle_destroy(sharezp->z_sa_hdl);
219 kmem_cache_free(znode_cache, sharezp);
224 #endif /* HAVE_SHARE */
228 * define a couple of values we need available
229 * for both 64 and 32 bit environments.
232 #define NBITSMINOR64 32
235 #define MAXMAJ64 0xffffffffUL
238 #define MAXMIN64 0xffffffffUL
241 #endif /* HAVE_ZPL */
244 * Create special expldev for ZFS private use.
245 * Can't use standard expldev since it doesn't do
246 * what we want. The standard expldev() takes a
247 * dev32_t in LP64 and expands it to a long dev_t.
248 * We need an interface that takes a dev32_t in ILP32
249 * and expands it to a long dev_t.
252 zfs_expldev(dev_t dev)
255 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
256 return (((uint64_t)major << NBITSMINOR64) |
257 ((minor_t)dev & MAXMIN32));
264 * Special cmpldev for ZFS private use.
265 * Can't use standard cmpldev since it takes
266 * a long dev_t and compresses it to dev32_t in
267 * LP64. We need to do a compaction of a long dev_t
268 * to a dev32_t in ILP32.
271 zfs_cmpldev(uint64_t dev)
274 minor_t minor = (minor_t)dev & MAXMIN64;
275 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
277 if (major > MAXMAJ32 || minor > MAXMIN32)
280 return (((dev32_t)major << NBITSMINOR32) | minor);
287 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
288 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
290 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
291 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
293 mutex_enter(&zp->z_lock);
295 ASSERT(zp->z_sa_hdl == NULL);
296 ASSERT(zp->z_acl_cached == NULL);
297 if (sa_hdl == NULL) {
298 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
299 SA_HDL_SHARED, &zp->z_sa_hdl));
301 zp->z_sa_hdl = sa_hdl;
302 sa_set_userp(sa_hdl, zp);
305 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
308 * Slap on VROOT if we are the root znode
310 if (zp->z_id == zfsvfs->z_root)
311 ZTOV(zp)->v_flag |= VROOT;
313 mutex_exit(&zp->z_lock);
318 zfs_znode_dmu_fini(znode_t *zp)
320 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
322 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
324 sa_handle_destroy(zp->z_sa_hdl);
329 * Construct a new znode+inode and initialize.
331 * This does not do a call to dmu_set_user() that is
332 * up to the caller to do, in case you don't want to
336 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
337 dmu_object_type_t obj_type, sa_handle_t *hdl)
342 sa_bulk_attr_t bulk[9];
345 ASSERT(zfsvfs != NULL);
346 ASSERT(zfsvfs->z_vfs != NULL);
347 ASSERT(zfsvfs->z_vfs->mnt_sb != NULL);
349 inode = iget_locked(zfsvfs->z_vfs->mnt_sb, db->db_object);
352 ASSERT(inode->i_state & I_NEW);
353 ASSERT(zp->z_dirlocks == NULL);
354 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
358 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
359 * the zfs_znode_move() callback.
363 zp->z_atime_dirty = 0;
365 zp->z_id = db->db_object;
367 zp->z_seq = 0x7A4653;
370 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
372 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
374 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
376 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
378 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
380 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
382 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
384 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
386 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
388 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
391 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
393 sa_handle_destroy(zp->z_sa_hdl);
398 inode->i_mode = (umode_t)zp->z_mode;
399 if ((S_ISCHR(inode->i_mode)) || (S_ISBLK(inode->i_mode))) {
401 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
402 &rdev, sizeof (rdev)) == 0);
403 inode->i_rdev = zfs_cmpldev(rdev);
406 /* zp->z_set_ops_inode() must be set in sb->alloc_inode() */
407 ASSERT(zp->z_set_ops_inode != NULL);
408 zp->z_set_ops_inode(inode);
409 unlock_new_inode(inode);
411 mutex_enter(&zfsvfs->z_znodes_lock);
412 list_insert_tail(&zfsvfs->z_all_znodes, zp);
415 * Everything else must be valid before assigning z_zfsvfs makes the
416 * znode eligible for zfs_znode_move().
418 zp->z_zfsvfs = zfsvfs;
419 mutex_exit(&zfsvfs->z_znodes_lock);
421 VFS_HOLD(zfsvfs->z_vfs);
426 * Update the embedded inode given the znode. We should work toward
427 * eliminating this function as soon as possible by removing values
428 * which are duplicated between the znode and inode. If the generic
429 * inode has the correct field it should be used, and the ZFS code
430 * updated to access the inode. This can be done incrementally.
433 zfs_inode_update(znode_t *zp)
438 uint64_t atime[2], mtime[2], ctime[2];
441 zfsvfs = zp->z_zfsvfs;
444 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), &atime, 16);
445 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs), &mtime, 16);
446 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs), &ctime, 16);
448 spin_lock(&inode->i_lock);
449 inode->i_generation = zp->z_gen;
450 inode->i_uid = zp->z_uid;
451 inode->i_gid = zp->z_gid;
452 inode->i_nlink = zp->z_links;
453 inode->i_mode = zp->z_mode;
454 inode->i_blkbits = SPA_MINBLOCKSHIFT;
455 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
456 (u_longlong_t *)&inode->i_blocks);
458 ZFS_TIME_DECODE(&inode->i_atime, atime);
459 ZFS_TIME_DECODE(&inode->i_mtime, mtime);
460 ZFS_TIME_DECODE(&inode->i_ctime, ctime);
462 i_size_write(inode, zp->z_size);
463 spin_unlock(&inode->i_lock);
466 static uint64_t empty_xattr;
467 static uint64_t pad[4];
468 static zfs_acl_phys_t acl_phys;
470 * Create a new DMU object to hold a zfs znode.
472 * IN: dzp - parent directory for new znode
473 * vap - file attributes for new znode
474 * tx - dmu transaction id for zap operations
475 * cr - credentials of caller
477 * IS_ROOT_NODE - new object will be root
478 * IS_XATTR - new object is an attribute
479 * bonuslen - length of bonus buffer
480 * setaclp - File/Dir initial ACL
481 * fuidp - Tracks fuid allocation.
483 * OUT: zpp - allocated znode
487 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
488 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
490 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
491 uint64_t mode, size, links, parent, pflags;
492 uint64_t dzp_pflags = 0;
494 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
501 dmu_object_type_t obj_type;
502 sa_bulk_attr_t *sa_attrs;
504 zfs_acl_locator_cb_t locate = { 0 };
506 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
508 if (zfsvfs->z_replay) {
509 obj = vap->va_nodeid;
510 now = vap->va_ctime; /* see zfs_replay_create() */
511 gen = vap->va_nblocks; /* ditto */
515 gen = dmu_tx_get_txg(tx);
518 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
519 bonuslen = (obj_type == DMU_OT_SA) ?
520 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
523 * Create a new DMU object.
526 * There's currently no mechanism for pre-reading the blocks that will
527 * be needed to allocate a new object, so we accept the small chance
528 * that there will be an i/o error and we will fail one of the
531 if (vap->va_type == VDIR) {
532 if (zfsvfs->z_replay) {
533 err = zap_create_claim_norm(zfsvfs->z_os, obj,
534 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
535 obj_type, bonuslen, tx);
536 ASSERT3U(err, ==, 0);
538 obj = zap_create_norm(zfsvfs->z_os,
539 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
540 obj_type, bonuslen, tx);
543 if (zfsvfs->z_replay) {
544 err = dmu_object_claim(zfsvfs->z_os, obj,
545 DMU_OT_PLAIN_FILE_CONTENTS, 0,
546 obj_type, bonuslen, tx);
547 ASSERT3U(err, ==, 0);
549 obj = dmu_object_alloc(zfsvfs->z_os,
550 DMU_OT_PLAIN_FILE_CONTENTS, 0,
551 obj_type, bonuslen, tx);
555 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
556 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
559 * If this is the root, fix up the half-initialized parent pointer
560 * to reference the just-allocated physical data area.
562 if (flag & IS_ROOT_NODE) {
565 dzp_pflags = dzp->z_pflags;
569 * If parent is an xattr, so am I.
571 if (dzp_pflags & ZFS_XATTR) {
575 if (zfsvfs->z_use_fuids)
576 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
580 if (vap->va_type == VDIR) {
581 size = 2; /* contents ("." and "..") */
582 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
587 if (vap->va_type == VBLK || vap->va_type == VCHR) {
588 rdev = zfs_expldev(vap->va_rdev);
592 mode = acl_ids->z_mode;
597 * No execs denied will be deterimed when zfs_mode_compute() is called.
599 pflags |= acl_ids->z_aclp->z_hints &
600 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
601 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
603 ZFS_TIME_ENCODE(&now, crtime);
604 ZFS_TIME_ENCODE(&now, ctime);
606 if (vap->va_mask & AT_ATIME) {
607 ZFS_TIME_ENCODE(&vap->va_atime, atime);
609 ZFS_TIME_ENCODE(&now, atime);
612 if (vap->va_mask & AT_MTIME) {
613 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
615 ZFS_TIME_ENCODE(&now, mtime);
618 /* Now add in all of the "SA" attributes */
619 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
623 * Setup the array of attributes to be replaced/set on the new file
625 * order for DMU_OT_ZNODE is critical since it needs to be constructed
626 * in the old znode_phys_t format. Don't change this ordering
628 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
630 if (obj_type == DMU_OT_ZNODE) {
631 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
633 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
635 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
637 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
639 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
641 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
643 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
645 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
648 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
650 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
652 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
654 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
655 &acl_ids->z_fuid, 8);
656 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
657 &acl_ids->z_fgid, 8);
658 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
660 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
662 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
664 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
666 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
668 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
672 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
674 if (obj_type == DMU_OT_ZNODE) {
675 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
678 if (obj_type == DMU_OT_ZNODE ||
679 (vap->va_type == VBLK || vap->va_type == VCHR)) {
680 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
684 if (obj_type == DMU_OT_ZNODE) {
685 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
687 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
688 &acl_ids->z_fuid, 8);
689 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
690 &acl_ids->z_fgid, 8);
691 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
692 sizeof (uint64_t) * 4);
693 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
694 &acl_phys, sizeof (zfs_acl_phys_t));
695 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
696 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
697 &acl_ids->z_aclp->z_acl_count, 8);
698 locate.cb_aclp = acl_ids->z_aclp;
699 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
700 zfs_acl_data_locator, &locate,
701 acl_ids->z_aclp->z_acl_bytes);
702 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
703 acl_ids->z_fuid, acl_ids->z_fgid);
706 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
708 if (!(flag & IS_ROOT_NODE)) {
709 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
710 ASSERT(*zpp != NULL);
713 * If we are creating the root node, the "parent" we
714 * passed in is the znode for the root.
718 (*zpp)->z_sa_hdl = sa_hdl;
721 (*zpp)->z_pflags = pflags;
722 (*zpp)->z_mode = mode;
724 if (vap->va_mask & AT_XVATTR)
725 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
727 if (obj_type == DMU_OT_ZNODE ||
728 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
729 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
730 ASSERT3S(err, ==, 0);
732 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
733 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
737 * zfs_xvattr_set only updates the in-core attributes
738 * it is assumed the caller will be doing an sa_bulk_update
739 * to push the changes out
742 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
747 xoap = xva_getxoptattr(xvap);
750 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
752 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
753 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
754 ×, sizeof (times), tx);
755 XVA_SET_RTN(xvap, XAT_CREATETIME);
757 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
758 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
760 XVA_SET_RTN(xvap, XAT_READONLY);
762 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
763 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
765 XVA_SET_RTN(xvap, XAT_HIDDEN);
767 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
768 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
770 XVA_SET_RTN(xvap, XAT_SYSTEM);
772 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
773 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
775 XVA_SET_RTN(xvap, XAT_ARCHIVE);
777 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
778 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
780 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
782 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
783 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
785 XVA_SET_RTN(xvap, XAT_NOUNLINK);
787 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
788 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
790 XVA_SET_RTN(xvap, XAT_APPENDONLY);
792 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
793 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
795 XVA_SET_RTN(xvap, XAT_NODUMP);
797 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
798 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
800 XVA_SET_RTN(xvap, XAT_OPAQUE);
802 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
803 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
804 xoap->xoa_av_quarantined, zp->z_pflags, tx);
805 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
807 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
808 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
810 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
812 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
813 zfs_sa_set_scanstamp(zp, xvap, tx);
814 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
816 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
817 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
819 XVA_SET_RTN(xvap, XAT_REPARSE);
821 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
822 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
824 XVA_SET_RTN(xvap, XAT_OFFLINE);
826 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
827 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
829 XVA_SET_RTN(xvap, XAT_SPARSE);
831 #endif /* HAVE_XVATTR */
835 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
837 dmu_object_info_t doi;
845 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
847 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
849 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
853 dmu_object_info_from_db(db, &doi);
854 if (doi.doi_bonus_type != DMU_OT_SA &&
855 (doi.doi_bonus_type != DMU_OT_ZNODE ||
856 (doi.doi_bonus_type == DMU_OT_ZNODE &&
857 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
858 sa_buf_rele(db, NULL);
859 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
863 hdl = dmu_buf_get_user(db);
865 zp = sa_get_userdata(hdl);
869 * Since "SA" does immediate eviction we
870 * should never find a sa handle that doesn't
871 * know about the znode.
874 ASSERT3P(zp, !=, NULL);
876 mutex_enter(&zp->z_lock);
877 ASSERT3U(zp->z_id, ==, obj_num);
878 if (zp->z_unlinked) {
885 sa_buf_rele(db, NULL);
886 mutex_exit(&zp->z_lock);
887 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
892 * Not found create new znode/vnode
893 * but only if file exists.
895 * There is a small window where zfs_vget() could
896 * find this object while a file create is still in
897 * progress. This is checked for in zfs_znode_alloc()
899 * if zfs_znode_alloc() fails it will drop the hold on the
902 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
903 doi.doi_bonus_type, NULL);
909 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
914 zfs_rezget(znode_t *zp)
916 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
917 dmu_object_info_t doi;
919 uint64_t obj_num = zp->z_id;
921 sa_bulk_attr_t bulk[8];
926 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
928 mutex_enter(&zp->z_acl_lock);
929 if (zp->z_acl_cached) {
930 zfs_acl_free(zp->z_acl_cached);
931 zp->z_acl_cached = NULL;
934 mutex_exit(&zp->z_acl_lock);
935 ASSERT(zp->z_sa_hdl == NULL);
936 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
938 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
942 dmu_object_info_from_db(db, &doi);
943 if (doi.doi_bonus_type != DMU_OT_SA &&
944 (doi.doi_bonus_type != DMU_OT_ZNODE ||
945 (doi.doi_bonus_type == DMU_OT_ZNODE &&
946 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
947 sa_buf_rele(db, NULL);
948 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
952 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
954 /* reload cached values */
955 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
957 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
958 &zp->z_size, sizeof (zp->z_size));
959 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
960 &zp->z_links, sizeof (zp->z_links));
961 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
962 &zp->z_pflags, sizeof (zp->z_pflags));
963 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
964 &zp->z_atime, sizeof (zp->z_atime));
965 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
966 &zp->z_uid, sizeof (zp->z_uid));
967 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
968 &zp->z_gid, sizeof (zp->z_gid));
969 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
970 &mode, sizeof (mode));
972 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
973 zfs_znode_dmu_fini(zp);
974 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
980 if (gen != zp->z_gen) {
981 zfs_znode_dmu_fini(zp);
982 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
986 zp->z_unlinked = (zp->z_links == 0);
987 zp->z_blksz = doi.doi_data_block_size;
989 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
995 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
997 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
998 objset_t *os = zfsvfs->z_os;
999 uint64_t obj = zp->z_id;
1000 uint64_t acl_obj = zfs_external_acl(zp);
1002 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1004 VERIFY(!zp->z_is_sa);
1005 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1007 VERIFY(0 == dmu_object_free(os, obj, tx));
1008 zfs_znode_dmu_fini(zp);
1009 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1014 zfs_zinactive(znode_t *zp)
1016 vnode_t *vp = ZTOV(zp);
1017 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1018 uint64_t z_id = zp->z_id;
1020 ASSERT(zp->z_sa_hdl);
1023 * Don't allow a zfs_zget() while were trying to release this znode
1025 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1027 mutex_enter(&zp->z_lock);
1028 mutex_enter(&vp->v_lock);
1030 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1032 * If the hold count is greater than zero, somebody has
1033 * obtained a new reference on this znode while we were
1034 * processing it here, so we are done. If we still have
1035 * mapped pages then we are also done, since we don't
1036 * want to inactivate the znode until the pages get pushed.
1038 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1039 * this seems like it would leave the znode hanging with
1040 * no chance to go inactive...
1042 mutex_exit(&vp->v_lock);
1043 mutex_exit(&zp->z_lock);
1044 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1047 mutex_exit(&vp->v_lock);
1050 * If this was the last reference to a file with no links,
1051 * remove the file from the file system.
1053 if (zp->z_unlinked) {
1054 mutex_exit(&zp->z_lock);
1055 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1060 mutex_exit(&zp->z_lock);
1061 zfs_znode_dmu_fini(zp);
1062 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1067 zfs_znode_free(znode_t *zp)
1069 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1071 vn_invalid(ZTOV(zp));
1073 ASSERT(ZTOV(zp)->v_count == 0);
1075 mutex_enter(&zfsvfs->z_znodes_lock);
1076 POINTER_INVALIDATE(&zp->z_zfsvfs);
1077 list_remove(&zfsvfs->z_all_znodes, zp);
1078 mutex_exit(&zfsvfs->z_znodes_lock);
1080 if (zp->z_acl_cached) {
1081 zfs_acl_free(zp->z_acl_cached);
1082 zp->z_acl_cached = NULL;
1085 kmem_cache_free(znode_cache, zp);
1087 VFS_RELE(zfsvfs->z_vfs);
1091 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1092 uint64_t ctime[2], boolean_t have_tx)
1098 if (have_tx) { /* will sa_bulk_update happen really soon? */
1099 zp->z_atime_dirty = 0;
1102 zp->z_atime_dirty = 1;
1105 if (flag & AT_ATIME) {
1106 ZFS_TIME_ENCODE(&now, zp->z_atime);
1109 if (flag & AT_MTIME) {
1110 ZFS_TIME_ENCODE(&now, mtime);
1111 if (zp->z_zfsvfs->z_use_fuids) {
1112 zp->z_pflags |= (ZFS_ARCHIVE |
1117 if (flag & AT_CTIME) {
1118 ZFS_TIME_ENCODE(&now, ctime);
1119 if (zp->z_zfsvfs->z_use_fuids)
1120 zp->z_pflags |= ZFS_ARCHIVE;
1125 * Grow the block size for a file.
1127 * IN: zp - znode of file to free data in.
1128 * size - requested block size
1129 * tx - open transaction.
1131 * NOTE: this function assumes that the znode is write locked.
1134 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1139 if (size <= zp->z_blksz)
1142 * If the file size is already greater than the current blocksize,
1143 * we will not grow. If there is more than one block in a file,
1144 * the blocksize cannot change.
1146 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1149 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1152 if (error == ENOTSUP)
1154 ASSERT3U(error, ==, 0);
1156 /* What blocksize did we actually get? */
1157 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1162 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1163 * be calling back into the fs for a putpage(). E.g.: when truncating
1164 * a file, the pages being "thrown away* don't need to be written out.
1168 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1169 int flags, cred_t *cr)
1174 #endif /* HAVE_ZPL */
1177 * Increase the file length
1179 * IN: zp - znode of file to free data in.
1180 * end - new end-of-file
1182 * RETURN: 0 if success
1183 * error code if failure
1186 zfs_extend(znode_t *zp, uint64_t end)
1188 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1195 * We will change zp_size, lock the whole file.
1197 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1200 * Nothing to do if file already at desired length.
1202 if (end <= zp->z_size) {
1203 zfs_range_unlock(rl);
1207 tx = dmu_tx_create(zfsvfs->z_os);
1208 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1209 zfs_sa_upgrade_txholds(tx, zp);
1210 if (end > zp->z_blksz &&
1211 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1213 * We are growing the file past the current block size.
1215 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1216 ASSERT(!ISP2(zp->z_blksz));
1217 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1219 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1221 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1226 error = dmu_tx_assign(tx, TXG_NOWAIT);
1228 if (error == ERESTART) {
1234 zfs_range_unlock(rl);
1239 zfs_grow_blocksize(zp, newblksz, tx);
1243 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1244 &zp->z_size, sizeof (zp->z_size), tx));
1246 zfs_range_unlock(rl);
1254 * Free space in a file.
1256 * IN: zp - znode of file to free data in.
1257 * off - start of section to free.
1258 * len - length of section to free.
1260 * RETURN: 0 if success
1261 * error code if failure
1264 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1266 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1271 * Lock the range being freed.
1273 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1276 * Nothing to do if file already at desired length.
1278 if (off >= zp->z_size) {
1279 zfs_range_unlock(rl);
1283 if (off + len > zp->z_size)
1284 len = zp->z_size - off;
1286 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1288 zfs_range_unlock(rl);
1296 * IN: zp - znode of file to free data in.
1297 * end - new end-of-file.
1299 * RETURN: 0 if success
1300 * error code if failure
1303 zfs_trunc(znode_t *zp, uint64_t end)
1305 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1307 vnode_t *vp = ZTOV(zp);
1308 #endif /* HAVE_ZPL */
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(zfsvfs->z_os, zp->z_id, end, -1);
1330 zfs_range_unlock(rl);
1334 tx = dmu_tx_create(zfsvfs->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(zfsvfs),
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(zfsvfs),
1356 NULL, &zp->z_pflags, 8);
1358 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1364 * Clear any mapped pages in the truncated region. This has to
1365 * happen outside of the transaction to avoid the possibility of
1366 * a deadlock with someone trying to push a page that we are
1367 * about to invalidate.
1369 if (vn_has_cached_data(vp)) {
1371 uint64_t start = end & PAGEMASK;
1372 int poff = end & PAGEOFFSET;
1374 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1376 * We need to zero a partial page.
1378 pagezero(pp, poff, PAGESIZE - poff);
1382 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1383 B_INVAL | B_TRUNC, NULL);
1386 #endif /* HAVE_ZPL */
1388 zfs_range_unlock(rl);
1394 * Free space in a file
1396 * IN: zp - znode of file to free data in.
1397 * off - start of range
1398 * len - end of range (0 => EOF)
1399 * flag - current file open mode flags.
1400 * log - TRUE if this action should be logged
1402 * RETURN: 0 if success
1403 * error code if failure
1406 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1409 vnode_t *vp = ZTOV(zp);
1410 #endif /* HAVE_ZPL */
1412 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1413 zilog_t *zilog = zfsvfs->z_log;
1415 uint64_t mtime[2], ctime[2];
1416 sa_bulk_attr_t bulk[3];
1420 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1421 sizeof (mode))) != 0)
1424 if (off > zp->z_size) {
1425 error = zfs_extend(zp, off+len);
1426 if (error == 0 && log)
1434 * Check for any locks in the region to be freed.
1437 if (MANDLOCK(vp, (mode_t)mode)) {
1438 uint64_t length = (len ? len : zp->z_size - off);
1439 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1442 #endif /* HAVE_ZPL */
1445 error = zfs_trunc(zp, off);
1447 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1448 off + len > zp->z_size)
1449 error = zfs_extend(zp, off+len);
1454 tx = dmu_tx_create(zfsvfs->z_os);
1455 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1456 zfs_sa_upgrade_txholds(tx, zp);
1457 error = dmu_tx_assign(tx, TXG_NOWAIT);
1459 if (error == ERESTART) {
1468 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1469 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1470 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1471 NULL, &zp->z_pflags, 8);
1472 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1473 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1476 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1479 zfs_inode_update(zp);
1484 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1486 uint64_t moid, obj, sa_obj, version;
1487 uint64_t sense = ZFS_CASE_SENSITIVE;
1494 znode_t *rootzp = NULL;
1498 zfs_acl_ids_t acl_ids;
1503 #endif /* HAVE_ZPL */
1506 * First attempt to create master node.
1509 * In an empty objset, there are no blocks to read and thus
1510 * there can be no i/o errors (which we assert below).
1512 moid = MASTER_NODE_OBJ;
1513 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1514 DMU_OT_NONE, 0, tx);
1518 * Set starting attributes.
1520 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1522 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1523 /* For the moment we expect all zpl props to be uint64_ts */
1527 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1528 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1529 name = nvpair_name(elem);
1530 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1534 error = zap_update(os, moid, name, 8, 1, &val, tx);
1537 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1539 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1542 ASSERT(version != 0);
1543 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1546 * Create zap object used for SA attribute registration
1549 if (version >= ZPL_VERSION_SA) {
1550 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1551 DMU_OT_NONE, 0, tx);
1552 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1558 * Create a delete queue.
1560 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1562 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1567 * Create root znode. Create minimal znode/vnode/zfsvfs
1568 * to allow zfs_mknode to work.
1570 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1571 vattr.va_type = VDIR;
1572 vattr.va_mode = S_IFDIR|0755;
1573 vattr.va_uid = crgetuid(cr);
1574 vattr.va_gid = crgetgid(cr);
1576 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1577 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1578 rootzp->z_moved = 0;
1579 rootzp->z_unlinked = 0;
1580 rootzp->z_atime_dirty = 0;
1581 rootzp->z_is_sa = USE_SA(version, os);
1587 bzero(&zfsvfs, sizeof (zfsvfs_t));
1590 zfsvfs.z_parent = &zfsvfs;
1591 zfsvfs.z_version = version;
1592 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1593 zfsvfs.z_use_sa = USE_SA(version, os);
1594 zfsvfs.z_norm = norm;
1596 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1597 &zfsvfs.z_attr_table);
1602 * Fold case on file systems that are always or sometimes case
1605 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1606 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1608 /* XXX - This must be destroyed but I'm not quite sure yet so
1609 * I'm just annotating that fact when it's an issue. -Brian */
1610 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1611 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1612 offsetof(znode_t, z_link_node));
1614 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1615 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1617 rootzp->z_zfsvfs = &zfsvfs;
1618 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1619 cr, NULL, &acl_ids));
1620 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1621 ASSERT3P(zp, ==, rootzp);
1622 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1623 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1625 zfs_acl_ids_free(&acl_ids);
1626 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1628 ZTOV(rootzp)->v_count = 0;
1629 sa_handle_destroy(rootzp->z_sa_hdl);
1630 kmem_cache_free(znode_cache, rootzp);
1631 error = zfs_create_share_dir(&zfsvfs, tx);
1633 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1634 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1637 * Create root znode with code free of VFS dependencies
1639 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1640 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1642 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1643 dmu_buf_will_dirty(db, tx);
1646 * Initialize the znode physical data to zero.
1648 ASSERT(db->db_size >= sizeof (znode_phys_t));
1649 bzero(db->db_data, db->db_size);
1652 if (USE_FUIDS(version, os))
1653 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1655 pzp->zp_size = 2; /* "." and ".." */
1657 pzp->zp_parent = obj;
1658 pzp->zp_gen = dmu_tx_get_txg(tx);
1659 pzp->zp_mode = S_IFDIR | 0755;
1660 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1664 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1665 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1666 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1667 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1669 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1672 dmu_buf_rele(db, FTAG);
1673 #endif /* HAVE_ZPL */
1676 #endif /* _KERNEL */
1679 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1681 uint64_t sa_obj = 0;
1684 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1685 if (error != 0 && error != ENOENT)
1688 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1693 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1696 dmu_object_info_t doi;
1699 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1702 dmu_object_info_from_db(*db, &doi);
1703 if ((doi.doi_bonus_type != DMU_OT_SA &&
1704 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1705 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1706 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1707 sa_buf_rele(*db, FTAG);
1711 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1713 sa_buf_rele(*db, FTAG);
1721 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1723 sa_handle_destroy(hdl);
1724 sa_buf_rele(db, FTAG);
1728 * Given an object number, return its parent object number and whether
1729 * or not the object is an extended attribute directory.
1732 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1738 sa_bulk_attr_t bulk[3];
1742 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1743 &parent, sizeof (parent));
1744 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1745 &pflags, sizeof (pflags));
1746 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1747 &mode, sizeof (mode));
1749 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1753 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1759 * Given an object number, return some zpl level statistics
1762 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1765 sa_bulk_attr_t bulk[4];
1768 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1769 &sb->zs_mode, sizeof (sb->zs_mode));
1770 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1771 &sb->zs_gen, sizeof (sb->zs_gen));
1772 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1773 &sb->zs_links, sizeof (sb->zs_links));
1774 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1775 &sb->zs_ctime, sizeof (sb->zs_ctime));
1777 return (sa_bulk_lookup(hdl, bulk, count));
1781 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1782 sa_attr_type_t *sa_table, char *buf, int len)
1784 sa_handle_t *sa_hdl;
1785 sa_handle_t *prevhdl = NULL;
1786 dmu_buf_t *prevdb = NULL;
1787 dmu_buf_t *sa_db = NULL;
1788 char *path = buf + len - 1;
1796 char component[MAXNAMELEN + 2];
1801 zfs_release_sa_handle(prevhdl, prevdb);
1803 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1804 &is_xattrdir)) != 0)
1815 (void) sprintf(component + 1, "<xattrdir>");
1817 error = zap_value_search(osp, pobj, obj,
1818 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1823 complen = strlen(component);
1825 ASSERT(path >= buf);
1826 bcopy(component, path, complen);
1829 if (sa_hdl != hdl) {
1833 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1841 if (sa_hdl != NULL && sa_hdl != hdl) {
1842 ASSERT(sa_db != NULL);
1843 zfs_release_sa_handle(sa_hdl, sa_db);
1847 (void) memmove(buf, path, buf + len - path);
1853 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1855 sa_attr_type_t *sa_table;
1860 error = zfs_sa_setup(osp, &sa_table);
1864 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1868 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1870 zfs_release_sa_handle(hdl, db);
1875 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1878 char *path = buf + len - 1;
1879 sa_attr_type_t *sa_table;
1886 error = zfs_sa_setup(osp, &sa_table);
1890 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1894 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1896 zfs_release_sa_handle(hdl, db);
1900 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1902 zfs_release_sa_handle(hdl, db);
1906 #if defined(_KERNEL) && defined(HAVE_SPL)
1907 EXPORT_SYMBOL(zfs_create_fs);
1908 EXPORT_SYMBOL(zfs_obj_to_path);