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);
168 * Cleanup vfs & vnode ops
171 zfs_remove_op_tables();
172 #endif /* HAVE_ZPL */
178 kmem_cache_destroy(znode_cache);
180 rw_destroy(&zfsvfs_lock);
184 struct vnodeops *zfs_dvnodeops;
185 struct vnodeops *zfs_fvnodeops;
186 struct vnodeops *zfs_symvnodeops;
187 struct vnodeops *zfs_xdvnodeops;
188 struct vnodeops *zfs_evnodeops;
189 struct vnodeops *zfs_sharevnodeops;
192 zfs_remove_op_tables()
198 (void) vfs_freevfsops_by_type(zfsfstype);
205 vn_freevnodeops(zfs_dvnodeops);
207 vn_freevnodeops(zfs_fvnodeops);
209 vn_freevnodeops(zfs_symvnodeops);
211 vn_freevnodeops(zfs_xdvnodeops);
213 vn_freevnodeops(zfs_evnodeops);
214 if (zfs_sharevnodeops)
215 vn_freevnodeops(zfs_sharevnodeops);
217 zfs_dvnodeops = NULL;
218 zfs_fvnodeops = NULL;
219 zfs_symvnodeops = NULL;
220 zfs_xdvnodeops = NULL;
221 zfs_evnodeops = NULL;
222 zfs_sharevnodeops = NULL;
225 extern const fs_operation_def_t zfs_dvnodeops_template[];
226 extern const fs_operation_def_t zfs_fvnodeops_template[];
227 extern const fs_operation_def_t zfs_xdvnodeops_template[];
228 extern const fs_operation_def_t zfs_symvnodeops_template[];
229 extern const fs_operation_def_t zfs_evnodeops_template[];
230 extern const fs_operation_def_t zfs_sharevnodeops_template[];
233 zfs_create_op_tables()
238 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
239 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
240 * In this case we just return as the ops vectors are already set up.
245 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
250 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
255 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
260 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
265 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
270 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
277 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
280 zfs_acl_ids_t acl_ids;
287 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
288 vattr.va_type = VDIR;
289 vattr.va_mode = S_IFDIR|0555;
290 vattr.va_uid = crgetuid(kcred);
291 vattr.va_gid = crgetgid(kcred);
293 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
294 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
295 sharezp->z_moved = 0;
296 sharezp->z_unlinked = 0;
297 sharezp->z_atime_dirty = 0;
298 sharezp->z_zfsvfs = zfsvfs;
299 sharezp->z_is_sa = zfsvfs->z_use_sa;
305 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
306 kcred, NULL, &acl_ids));
307 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
308 ASSERT3P(zp, ==, sharezp);
309 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
310 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
311 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
312 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
313 zfsvfs->z_shares_dir = sharezp->z_id;
315 zfs_acl_ids_free(&acl_ids);
316 ZTOV(sharezp)->v_count = 0;
317 sa_handle_destroy(sharezp->z_sa_hdl);
318 kmem_cache_free(znode_cache, sharezp);
323 #endif /* HAVE_SHARE */
327 * define a couple of values we need available
328 * for both 64 and 32 bit environments.
331 #define NBITSMINOR64 32
334 #define MAXMAJ64 0xffffffffUL
337 #define MAXMIN64 0xffffffffUL
340 #endif /* HAVE_ZPL */
343 * Create special expldev for ZFS private use.
344 * Can't use standard expldev since it doesn't do
345 * what we want. The standard expldev() takes a
346 * dev32_t in LP64 and expands it to a long dev_t.
347 * We need an interface that takes a dev32_t in ILP32
348 * and expands it to a long dev_t.
351 zfs_expldev(dev_t dev)
354 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
355 return (((uint64_t)major << NBITSMINOR64) |
356 ((minor_t)dev & MAXMIN32));
363 * Special cmpldev for ZFS private use.
364 * Can't use standard cmpldev since it takes
365 * a long dev_t and compresses it to dev32_t in
366 * LP64. We need to do a compaction of a long dev_t
367 * to a dev32_t in ILP32.
370 zfs_cmpldev(uint64_t dev)
373 minor_t minor = (minor_t)dev & MAXMIN64;
374 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
376 if (major > MAXMAJ32 || minor > MAXMIN32)
379 return (((dev32_t)major << NBITSMINOR32) | minor);
386 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
387 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
389 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
390 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
392 mutex_enter(&zp->z_lock);
394 ASSERT(zp->z_sa_hdl == NULL);
395 ASSERT(zp->z_acl_cached == NULL);
396 if (sa_hdl == NULL) {
397 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
398 SA_HDL_SHARED, &zp->z_sa_hdl));
400 zp->z_sa_hdl = sa_hdl;
401 sa_set_userp(sa_hdl, zp);
404 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
407 * Slap on VROOT if we are the root znode
409 if (zp->z_id == zfsvfs->z_root)
410 ZTOV(zp)->v_flag |= VROOT;
412 mutex_exit(&zp->z_lock);
417 zfs_znode_dmu_fini(znode_t *zp)
419 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
421 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
423 sa_handle_destroy(zp->z_sa_hdl);
428 * Construct a new znode+inode and initialize.
430 * This does not do a call to dmu_set_user() that is
431 * up to the caller to do, in case you don't want to
435 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
436 dmu_object_type_t obj_type, sa_handle_t *hdl)
441 sa_bulk_attr_t bulk[9];
444 ASSERT(zfsvfs != NULL);
445 ASSERT(zfsvfs->z_vfs != NULL);
446 ASSERT(zfsvfs->z_vfs->mnt_sb != NULL);
448 inode = iget_locked(zfsvfs->z_vfs->mnt_sb, db->db_object);
451 ASSERT(inode->i_state & I_NEW);
452 ASSERT(zp->z_dirlocks == NULL);
453 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
457 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
458 * the zfs_znode_move() callback.
462 zp->z_atime_dirty = 0;
464 zp->z_id = db->db_object;
466 zp->z_seq = 0x7A4653;
469 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
471 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
473 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
475 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
477 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
479 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
481 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
483 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
485 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
487 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
490 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
492 sa_handle_destroy(zp->z_sa_hdl);
497 inode->i_mode = (umode_t)zp->z_mode;
498 if ((S_ISCHR(inode->i_mode)) || (S_ISBLK(inode->i_mode))) {
500 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
501 &rdev, sizeof (rdev)) == 0);
502 inode->i_rdev = zfs_cmpldev(rdev);
505 /* zp->z_set_ops_inode() must be set in sb->alloc_inode() */
506 ASSERT(zp->z_set_ops_inode != NULL);
507 zp->z_set_ops_inode(inode);
508 unlock_new_inode(inode);
510 mutex_enter(&zfsvfs->z_znodes_lock);
511 list_insert_tail(&zfsvfs->z_all_znodes, zp);
514 * Everything else must be valid before assigning z_zfsvfs makes the
515 * znode eligible for zfs_znode_move().
517 zp->z_zfsvfs = zfsvfs;
518 mutex_exit(&zfsvfs->z_znodes_lock);
520 VFS_HOLD(zfsvfs->z_vfs);
524 static uint64_t empty_xattr;
525 static uint64_t pad[4];
526 static zfs_acl_phys_t acl_phys;
528 * Create a new DMU object to hold a zfs znode.
530 * IN: dzp - parent directory for new znode
531 * vap - file attributes for new znode
532 * tx - dmu transaction id for zap operations
533 * cr - credentials of caller
535 * IS_ROOT_NODE - new object will be root
536 * IS_XATTR - new object is an attribute
537 * bonuslen - length of bonus buffer
538 * setaclp - File/Dir initial ACL
539 * fuidp - Tracks fuid allocation.
541 * OUT: zpp - allocated znode
545 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
546 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
548 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
549 uint64_t mode, size, links, parent, pflags;
550 uint64_t dzp_pflags = 0;
552 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
559 dmu_object_type_t obj_type;
560 sa_bulk_attr_t *sa_attrs;
562 zfs_acl_locator_cb_t locate = { 0 };
564 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
566 if (zfsvfs->z_replay) {
567 obj = vap->va_nodeid;
568 now = vap->va_ctime; /* see zfs_replay_create() */
569 gen = vap->va_nblocks; /* ditto */
573 gen = dmu_tx_get_txg(tx);
576 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
577 bonuslen = (obj_type == DMU_OT_SA) ?
578 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
581 * Create a new DMU object.
584 * There's currently no mechanism for pre-reading the blocks that will
585 * be needed to allocate a new object, so we accept the small chance
586 * that there will be an i/o error and we will fail one of the
589 if (vap->va_type == VDIR) {
590 if (zfsvfs->z_replay) {
591 err = zap_create_claim_norm(zfsvfs->z_os, obj,
592 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
593 obj_type, bonuslen, tx);
594 ASSERT3U(err, ==, 0);
596 obj = zap_create_norm(zfsvfs->z_os,
597 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
598 obj_type, bonuslen, tx);
601 if (zfsvfs->z_replay) {
602 err = dmu_object_claim(zfsvfs->z_os, obj,
603 DMU_OT_PLAIN_FILE_CONTENTS, 0,
604 obj_type, bonuslen, tx);
605 ASSERT3U(err, ==, 0);
607 obj = dmu_object_alloc(zfsvfs->z_os,
608 DMU_OT_PLAIN_FILE_CONTENTS, 0,
609 obj_type, bonuslen, tx);
613 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
614 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
617 * If this is the root, fix up the half-initialized parent pointer
618 * to reference the just-allocated physical data area.
620 if (flag & IS_ROOT_NODE) {
623 dzp_pflags = dzp->z_pflags;
627 * If parent is an xattr, so am I.
629 if (dzp_pflags & ZFS_XATTR) {
633 if (zfsvfs->z_use_fuids)
634 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
638 if (vap->va_type == VDIR) {
639 size = 2; /* contents ("." and "..") */
640 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
645 if (vap->va_type == VBLK || vap->va_type == VCHR) {
646 rdev = zfs_expldev(vap->va_rdev);
650 mode = acl_ids->z_mode;
655 * No execs denied will be deterimed when zfs_mode_compute() is called.
657 pflags |= acl_ids->z_aclp->z_hints &
658 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
659 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
661 ZFS_TIME_ENCODE(&now, crtime);
662 ZFS_TIME_ENCODE(&now, ctime);
664 if (vap->va_mask & AT_ATIME) {
665 ZFS_TIME_ENCODE(&vap->va_atime, atime);
667 ZFS_TIME_ENCODE(&now, atime);
670 if (vap->va_mask & AT_MTIME) {
671 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
673 ZFS_TIME_ENCODE(&now, mtime);
676 /* Now add in all of the "SA" attributes */
677 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
681 * Setup the array of attributes to be replaced/set on the new file
683 * order for DMU_OT_ZNODE is critical since it needs to be constructed
684 * in the old znode_phys_t format. Don't change this ordering
686 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
688 if (obj_type == DMU_OT_ZNODE) {
689 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
691 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
693 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
695 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
697 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
699 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
701 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
703 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
706 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
708 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
710 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
712 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
713 &acl_ids->z_fuid, 8);
714 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
715 &acl_ids->z_fgid, 8);
716 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
718 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
720 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
722 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
724 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
726 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
730 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
732 if (obj_type == DMU_OT_ZNODE) {
733 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
736 if (obj_type == DMU_OT_ZNODE ||
737 (vap->va_type == VBLK || vap->va_type == VCHR)) {
738 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
742 if (obj_type == DMU_OT_ZNODE) {
743 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
745 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
746 &acl_ids->z_fuid, 8);
747 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
748 &acl_ids->z_fgid, 8);
749 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
750 sizeof (uint64_t) * 4);
751 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
752 &acl_phys, sizeof (zfs_acl_phys_t));
753 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
754 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
755 &acl_ids->z_aclp->z_acl_count, 8);
756 locate.cb_aclp = acl_ids->z_aclp;
757 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
758 zfs_acl_data_locator, &locate,
759 acl_ids->z_aclp->z_acl_bytes);
760 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
761 acl_ids->z_fuid, acl_ids->z_fgid);
764 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
766 if (!(flag & IS_ROOT_NODE)) {
767 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
768 ASSERT(*zpp != NULL);
771 * If we are creating the root node, the "parent" we
772 * passed in is the znode for the root.
776 (*zpp)->z_sa_hdl = sa_hdl;
779 (*zpp)->z_pflags = pflags;
780 (*zpp)->z_mode = mode;
782 if (vap->va_mask & AT_XVATTR)
783 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
785 if (obj_type == DMU_OT_ZNODE ||
786 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
787 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
788 ASSERT3S(err, ==, 0);
790 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
791 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
795 * zfs_xvattr_set only updates the in-core attributes
796 * it is assumed the caller will be doing an sa_bulk_update
797 * to push the changes out
800 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
805 xoap = xva_getxoptattr(xvap);
808 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
810 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
811 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
812 ×, sizeof (times), tx);
813 XVA_SET_RTN(xvap, XAT_CREATETIME);
815 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
816 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
818 XVA_SET_RTN(xvap, XAT_READONLY);
820 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
821 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
823 XVA_SET_RTN(xvap, XAT_HIDDEN);
825 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
826 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
828 XVA_SET_RTN(xvap, XAT_SYSTEM);
830 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
831 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
833 XVA_SET_RTN(xvap, XAT_ARCHIVE);
835 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
836 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
838 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
840 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
841 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
843 XVA_SET_RTN(xvap, XAT_NOUNLINK);
845 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
846 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
848 XVA_SET_RTN(xvap, XAT_APPENDONLY);
850 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
851 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
853 XVA_SET_RTN(xvap, XAT_NODUMP);
855 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
856 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
858 XVA_SET_RTN(xvap, XAT_OPAQUE);
860 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
861 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
862 xoap->xoa_av_quarantined, zp->z_pflags, tx);
863 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
865 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
866 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
868 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
870 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
871 zfs_sa_set_scanstamp(zp, xvap, tx);
872 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
874 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
875 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
877 XVA_SET_RTN(xvap, XAT_REPARSE);
879 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
880 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
882 XVA_SET_RTN(xvap, XAT_OFFLINE);
884 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
885 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
887 XVA_SET_RTN(xvap, XAT_SPARSE);
889 #endif /* HAVE_XVATTR */
893 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
895 dmu_object_info_t doi;
903 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
905 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
907 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
911 dmu_object_info_from_db(db, &doi);
912 if (doi.doi_bonus_type != DMU_OT_SA &&
913 (doi.doi_bonus_type != DMU_OT_ZNODE ||
914 (doi.doi_bonus_type == DMU_OT_ZNODE &&
915 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
916 sa_buf_rele(db, NULL);
917 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
921 hdl = dmu_buf_get_user(db);
923 zp = sa_get_userdata(hdl);
927 * Since "SA" does immediate eviction we
928 * should never find a sa handle that doesn't
929 * know about the znode.
932 ASSERT3P(zp, !=, NULL);
934 mutex_enter(&zp->z_lock);
935 ASSERT3U(zp->z_id, ==, obj_num);
936 if (zp->z_unlinked) {
943 sa_buf_rele(db, NULL);
944 mutex_exit(&zp->z_lock);
945 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
950 * Not found create new znode/vnode
951 * but only if file exists.
953 * There is a small window where zfs_vget() could
954 * find this object while a file create is still in
955 * progress. This is checked for in zfs_znode_alloc()
957 * if zfs_znode_alloc() fails it will drop the hold on the
960 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
961 doi.doi_bonus_type, NULL);
967 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
972 zfs_rezget(znode_t *zp)
974 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
975 dmu_object_info_t doi;
977 uint64_t obj_num = zp->z_id;
979 sa_bulk_attr_t bulk[8];
984 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
986 mutex_enter(&zp->z_acl_lock);
987 if (zp->z_acl_cached) {
988 zfs_acl_free(zp->z_acl_cached);
989 zp->z_acl_cached = NULL;
992 mutex_exit(&zp->z_acl_lock);
993 ASSERT(zp->z_sa_hdl == NULL);
994 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
996 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1000 dmu_object_info_from_db(db, &doi);
1001 if (doi.doi_bonus_type != DMU_OT_SA &&
1002 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1003 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1004 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1005 sa_buf_rele(db, NULL);
1006 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1010 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1012 /* reload cached values */
1013 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1014 &gen, sizeof (gen));
1015 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1016 &zp->z_size, sizeof (zp->z_size));
1017 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1018 &zp->z_links, sizeof (zp->z_links));
1019 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1020 &zp->z_pflags, sizeof (zp->z_pflags));
1021 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1022 &zp->z_atime, sizeof (zp->z_atime));
1023 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1024 &zp->z_uid, sizeof (zp->z_uid));
1025 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1026 &zp->z_gid, sizeof (zp->z_gid));
1027 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1028 &mode, sizeof (mode));
1030 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1031 zfs_znode_dmu_fini(zp);
1032 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1038 if (gen != zp->z_gen) {
1039 zfs_znode_dmu_fini(zp);
1040 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1044 zp->z_unlinked = (zp->z_links == 0);
1045 zp->z_blksz = doi.doi_data_block_size;
1047 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1053 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1055 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1056 objset_t *os = zfsvfs->z_os;
1057 uint64_t obj = zp->z_id;
1058 uint64_t acl_obj = zfs_external_acl(zp);
1060 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1062 VERIFY(!zp->z_is_sa);
1063 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1065 VERIFY(0 == dmu_object_free(os, obj, tx));
1066 zfs_znode_dmu_fini(zp);
1067 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1072 zfs_zinactive(znode_t *zp)
1074 vnode_t *vp = ZTOV(zp);
1075 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1076 uint64_t z_id = zp->z_id;
1078 ASSERT(zp->z_sa_hdl);
1081 * Don't allow a zfs_zget() while were trying to release this znode
1083 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1085 mutex_enter(&zp->z_lock);
1086 mutex_enter(&vp->v_lock);
1088 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1090 * If the hold count is greater than zero, somebody has
1091 * obtained a new reference on this znode while we were
1092 * processing it here, so we are done. If we still have
1093 * mapped pages then we are also done, since we don't
1094 * want to inactivate the znode until the pages get pushed.
1096 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1097 * this seems like it would leave the znode hanging with
1098 * no chance to go inactive...
1100 mutex_exit(&vp->v_lock);
1101 mutex_exit(&zp->z_lock);
1102 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1105 mutex_exit(&vp->v_lock);
1108 * If this was the last reference to a file with no links,
1109 * remove the file from the file system.
1111 if (zp->z_unlinked) {
1112 mutex_exit(&zp->z_lock);
1113 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1118 mutex_exit(&zp->z_lock);
1119 zfs_znode_dmu_fini(zp);
1120 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1125 zfs_znode_free(znode_t *zp)
1127 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1129 vn_invalid(ZTOV(zp));
1131 ASSERT(ZTOV(zp)->v_count == 0);
1133 mutex_enter(&zfsvfs->z_znodes_lock);
1134 POINTER_INVALIDATE(&zp->z_zfsvfs);
1135 list_remove(&zfsvfs->z_all_znodes, zp);
1136 mutex_exit(&zfsvfs->z_znodes_lock);
1138 if (zp->z_acl_cached) {
1139 zfs_acl_free(zp->z_acl_cached);
1140 zp->z_acl_cached = NULL;
1143 kmem_cache_free(znode_cache, zp);
1145 VFS_RELE(zfsvfs->z_vfs);
1149 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1150 uint64_t ctime[2], boolean_t have_tx)
1156 if (have_tx) { /* will sa_bulk_update happen really soon? */
1157 zp->z_atime_dirty = 0;
1160 zp->z_atime_dirty = 1;
1163 if (flag & AT_ATIME) {
1164 ZFS_TIME_ENCODE(&now, zp->z_atime);
1167 if (flag & AT_MTIME) {
1168 ZFS_TIME_ENCODE(&now, mtime);
1169 if (zp->z_zfsvfs->z_use_fuids) {
1170 zp->z_pflags |= (ZFS_ARCHIVE |
1175 if (flag & AT_CTIME) {
1176 ZFS_TIME_ENCODE(&now, ctime);
1177 if (zp->z_zfsvfs->z_use_fuids)
1178 zp->z_pflags |= ZFS_ARCHIVE;
1183 * Grow the block size for a file.
1185 * IN: zp - znode of file to free data in.
1186 * size - requested block size
1187 * tx - open transaction.
1189 * NOTE: this function assumes that the znode is write locked.
1192 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1197 if (size <= zp->z_blksz)
1200 * If the file size is already greater than the current blocksize,
1201 * we will not grow. If there is more than one block in a file,
1202 * the blocksize cannot change.
1204 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1207 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1210 if (error == ENOTSUP)
1212 ASSERT3U(error, ==, 0);
1214 /* What blocksize did we actually get? */
1215 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1220 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1221 * be calling back into the fs for a putpage(). E.g.: when truncating
1222 * a file, the pages being "thrown away* don't need to be written out.
1226 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1227 int flags, cred_t *cr)
1232 #endif /* HAVE_ZPL */
1235 * Increase the file length
1237 * IN: zp - znode of file to free data in.
1238 * end - new end-of-file
1240 * RETURN: 0 if success
1241 * error code if failure
1244 zfs_extend(znode_t *zp, uint64_t end)
1246 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1253 * We will change zp_size, lock the whole file.
1255 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1258 * Nothing to do if file already at desired length.
1260 if (end <= zp->z_size) {
1261 zfs_range_unlock(rl);
1265 tx = dmu_tx_create(zfsvfs->z_os);
1266 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1267 zfs_sa_upgrade_txholds(tx, zp);
1268 if (end > zp->z_blksz &&
1269 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1271 * We are growing the file past the current block size.
1273 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1274 ASSERT(!ISP2(zp->z_blksz));
1275 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1277 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1279 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1284 error = dmu_tx_assign(tx, TXG_NOWAIT);
1286 if (error == ERESTART) {
1292 zfs_range_unlock(rl);
1297 zfs_grow_blocksize(zp, newblksz, tx);
1301 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1302 &zp->z_size, sizeof (zp->z_size), tx));
1304 zfs_range_unlock(rl);
1312 * Free space in a file.
1314 * IN: zp - znode of file to free data in.
1315 * off - start of section to free.
1316 * len - length of section to free.
1318 * RETURN: 0 if success
1319 * error code if failure
1322 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1324 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1329 * Lock the range being freed.
1331 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1334 * Nothing to do if file already at desired length.
1336 if (off >= zp->z_size) {
1337 zfs_range_unlock(rl);
1341 if (off + len > zp->z_size)
1342 len = zp->z_size - off;
1344 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1346 zfs_range_unlock(rl);
1354 * IN: zp - znode of file to free data in.
1355 * end - new end-of-file.
1357 * RETURN: 0 if success
1358 * error code if failure
1361 zfs_trunc(znode_t *zp, uint64_t end)
1363 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1365 vnode_t *vp = ZTOV(zp);
1366 #endif /* HAVE_ZPL */
1370 sa_bulk_attr_t bulk[2];
1374 * We will change zp_size, lock the whole file.
1376 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1379 * Nothing to do if file already at desired length.
1381 if (end >= zp->z_size) {
1382 zfs_range_unlock(rl);
1386 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1388 zfs_range_unlock(rl);
1392 tx = dmu_tx_create(zfsvfs->z_os);
1393 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1394 zfs_sa_upgrade_txholds(tx, zp);
1395 error = dmu_tx_assign(tx, TXG_NOWAIT);
1397 if (error == ERESTART) {
1403 zfs_range_unlock(rl);
1408 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1409 NULL, &zp->z_size, sizeof (zp->z_size));
1412 zp->z_pflags &= ~ZFS_SPARSE;
1413 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1414 NULL, &zp->z_pflags, 8);
1416 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1422 * Clear any mapped pages in the truncated region. This has to
1423 * happen outside of the transaction to avoid the possibility of
1424 * a deadlock with someone trying to push a page that we are
1425 * about to invalidate.
1427 if (vn_has_cached_data(vp)) {
1429 uint64_t start = end & PAGEMASK;
1430 int poff = end & PAGEOFFSET;
1432 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1434 * We need to zero a partial page.
1436 pagezero(pp, poff, PAGESIZE - poff);
1440 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1441 B_INVAL | B_TRUNC, NULL);
1444 #endif /* HAVE_ZPL */
1446 zfs_range_unlock(rl);
1452 * Free space in a file
1454 * IN: zp - znode of file to free data in.
1455 * off - start of range
1456 * len - end of range (0 => EOF)
1457 * flag - current file open mode flags.
1458 * log - TRUE if this action should be logged
1460 * RETURN: 0 if success
1461 * error code if failure
1464 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1467 vnode_t *vp = ZTOV(zp);
1468 #endif /* HAVE_ZPL */
1470 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1471 zilog_t *zilog = zfsvfs->z_log;
1473 uint64_t mtime[2], ctime[2];
1474 sa_bulk_attr_t bulk[3];
1478 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1479 sizeof (mode))) != 0)
1482 if (off > zp->z_size) {
1483 error = zfs_extend(zp, off+len);
1484 if (error == 0 && log)
1492 * Check for any locks in the region to be freed.
1495 if (MANDLOCK(vp, (mode_t)mode)) {
1496 uint64_t length = (len ? len : zp->z_size - off);
1497 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1500 #endif /* HAVE_ZPL */
1503 error = zfs_trunc(zp, off);
1505 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1506 off + len > zp->z_size)
1507 error = zfs_extend(zp, off+len);
1512 tx = dmu_tx_create(zfsvfs->z_os);
1513 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1514 zfs_sa_upgrade_txholds(tx, zp);
1515 error = dmu_tx_assign(tx, TXG_NOWAIT);
1517 if (error == ERESTART) {
1526 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1527 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1528 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1529 NULL, &zp->z_pflags, 8);
1530 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1531 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1534 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1541 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1543 uint64_t moid, obj, sa_obj, version;
1544 uint64_t sense = ZFS_CASE_SENSITIVE;
1551 znode_t *rootzp = NULL;
1555 zfs_acl_ids_t acl_ids;
1560 #endif /* HAVE_ZPL */
1563 * First attempt to create master node.
1566 * In an empty objset, there are no blocks to read and thus
1567 * there can be no i/o errors (which we assert below).
1569 moid = MASTER_NODE_OBJ;
1570 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1571 DMU_OT_NONE, 0, tx);
1575 * Set starting attributes.
1577 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1579 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1580 /* For the moment we expect all zpl props to be uint64_ts */
1584 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1585 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1586 name = nvpair_name(elem);
1587 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1591 error = zap_update(os, moid, name, 8, 1, &val, tx);
1594 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1596 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1599 ASSERT(version != 0);
1600 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1603 * Create zap object used for SA attribute registration
1606 if (version >= ZPL_VERSION_SA) {
1607 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1608 DMU_OT_NONE, 0, tx);
1609 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1615 * Create a delete queue.
1617 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1619 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1624 * Create root znode. Create minimal znode/vnode/zfsvfs
1625 * to allow zfs_mknode to work.
1627 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1628 vattr.va_type = VDIR;
1629 vattr.va_mode = S_IFDIR|0755;
1630 vattr.va_uid = crgetuid(cr);
1631 vattr.va_gid = crgetgid(cr);
1633 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1634 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1635 rootzp->z_moved = 0;
1636 rootzp->z_unlinked = 0;
1637 rootzp->z_atime_dirty = 0;
1638 rootzp->z_is_sa = USE_SA(version, os);
1644 bzero(&zfsvfs, sizeof (zfsvfs_t));
1647 zfsvfs.z_parent = &zfsvfs;
1648 zfsvfs.z_version = version;
1649 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1650 zfsvfs.z_use_sa = USE_SA(version, os);
1651 zfsvfs.z_norm = norm;
1653 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1654 &zfsvfs.z_attr_table);
1659 * Fold case on file systems that are always or sometimes case
1662 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1663 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1665 /* XXX - This must be destroyed but I'm not quite sure yet so
1666 * I'm just annotating that fact when it's an issue. -Brian */
1667 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1668 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1669 offsetof(znode_t, z_link_node));
1671 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1672 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1674 rootzp->z_zfsvfs = &zfsvfs;
1675 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1676 cr, NULL, &acl_ids));
1677 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1678 ASSERT3P(zp, ==, rootzp);
1679 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1680 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1682 zfs_acl_ids_free(&acl_ids);
1683 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1685 ZTOV(rootzp)->v_count = 0;
1686 sa_handle_destroy(rootzp->z_sa_hdl);
1687 kmem_cache_free(znode_cache, rootzp);
1688 error = zfs_create_share_dir(&zfsvfs, tx);
1690 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1691 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1694 * Create root znode with code free of VFS dependencies
1696 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1697 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1699 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1700 dmu_buf_will_dirty(db, tx);
1703 * Initialize the znode physical data to zero.
1705 ASSERT(db->db_size >= sizeof (znode_phys_t));
1706 bzero(db->db_data, db->db_size);
1709 if (USE_FUIDS(version, os))
1710 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1712 pzp->zp_size = 2; /* "." and ".." */
1714 pzp->zp_parent = obj;
1715 pzp->zp_gen = dmu_tx_get_txg(tx);
1716 pzp->zp_mode = S_IFDIR | 0755;
1717 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1721 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1722 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1723 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1724 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1726 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1729 dmu_buf_rele(db, FTAG);
1730 #endif /* HAVE_ZPL */
1733 #endif /* _KERNEL */
1736 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1738 uint64_t sa_obj = 0;
1741 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1742 if (error != 0 && error != ENOENT)
1745 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1750 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1753 dmu_object_info_t doi;
1756 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1759 dmu_object_info_from_db(*db, &doi);
1760 if ((doi.doi_bonus_type != DMU_OT_SA &&
1761 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1762 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1763 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1764 sa_buf_rele(*db, FTAG);
1768 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1770 sa_buf_rele(*db, FTAG);
1778 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1780 sa_handle_destroy(hdl);
1781 sa_buf_rele(db, FTAG);
1785 * Given an object number, return its parent object number and whether
1786 * or not the object is an extended attribute directory.
1789 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1795 sa_bulk_attr_t bulk[3];
1799 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1800 &parent, sizeof (parent));
1801 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1802 &pflags, sizeof (pflags));
1803 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1804 &mode, sizeof (mode));
1806 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1810 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1816 * Given an object number, return some zpl level statistics
1819 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1822 sa_bulk_attr_t bulk[4];
1825 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1826 &sb->zs_mode, sizeof (sb->zs_mode));
1827 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1828 &sb->zs_gen, sizeof (sb->zs_gen));
1829 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1830 &sb->zs_links, sizeof (sb->zs_links));
1831 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1832 &sb->zs_ctime, sizeof (sb->zs_ctime));
1834 return (sa_bulk_lookup(hdl, bulk, count));
1838 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1839 sa_attr_type_t *sa_table, char *buf, int len)
1841 sa_handle_t *sa_hdl;
1842 sa_handle_t *prevhdl = NULL;
1843 dmu_buf_t *prevdb = NULL;
1844 dmu_buf_t *sa_db = NULL;
1845 char *path = buf + len - 1;
1853 char component[MAXNAMELEN + 2];
1858 zfs_release_sa_handle(prevhdl, prevdb);
1860 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1861 &is_xattrdir)) != 0)
1872 (void) sprintf(component + 1, "<xattrdir>");
1874 error = zap_value_search(osp, pobj, obj,
1875 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1880 complen = strlen(component);
1882 ASSERT(path >= buf);
1883 bcopy(component, path, complen);
1886 if (sa_hdl != hdl) {
1890 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1898 if (sa_hdl != NULL && sa_hdl != hdl) {
1899 ASSERT(sa_db != NULL);
1900 zfs_release_sa_handle(sa_hdl, sa_db);
1904 (void) memmove(buf, path, buf + len - path);
1910 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1912 sa_attr_type_t *sa_table;
1917 error = zfs_sa_setup(osp, &sa_table);
1921 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1925 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1927 zfs_release_sa_handle(hdl, db);
1932 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1935 char *path = buf + len - 1;
1936 sa_attr_type_t *sa_table;
1943 error = zfs_sa_setup(osp, &sa_table);
1947 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1951 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1953 zfs_release_sa_handle(hdl, db);
1957 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1959 zfs_release_sa_handle(hdl, db);
1963 #if defined(_KERNEL) && defined(HAVE_SPL)
1964 EXPORT_SYMBOL(zfs_create_fs);
1965 EXPORT_SYMBOL(zfs_obj_to_path);