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.
93 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
94 * be freed before it can be safely accessed.
96 krwlock_t zfsvfs_lock;
98 static kmem_cache_t *znode_cache = NULL;
102 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
105 * We should never drop all dbuf refs without first clearing
106 * the eviction callback.
108 panic("evicting znode %p\n", user_ptr);
113 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
117 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
119 zp->z_vnode = vn_alloc(kmflags);
120 if (zp->z_vnode == NULL) {
123 ZTOV(zp)->v_data = zp;
125 list_link_init(&zp->z_link_node);
127 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
128 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
129 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
130 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
132 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
133 avl_create(&zp->z_range_avl, zfs_range_compare,
134 sizeof (rl_t), offsetof(rl_t, r_node));
136 zp->z_dirlocks = NULL;
137 zp->z_acl_cached = NULL;
144 zfs_znode_cache_destructor(void *buf, void *arg)
148 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
149 ASSERT(ZTOV(zp)->v_data == zp);
151 ASSERT(!list_link_active(&zp->z_link_node));
152 mutex_destroy(&zp->z_lock);
153 rw_destroy(&zp->z_parent_lock);
154 rw_destroy(&zp->z_name_lock);
155 mutex_destroy(&zp->z_acl_lock);
156 avl_destroy(&zp->z_range_avl);
157 mutex_destroy(&zp->z_range_lock);
159 ASSERT(zp->z_dirlocks == NULL);
160 ASSERT(zp->z_acl_cached == NULL);
169 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
170 ASSERT(znode_cache == NULL);
171 znode_cache = kmem_cache_create("zfs_znode_cache",
172 sizeof (znode_t), 0, zfs_znode_cache_constructor,
173 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
180 * Cleanup vfs & vnode ops
182 zfs_remove_op_tables();
188 kmem_cache_destroy(znode_cache);
190 rw_destroy(&zfsvfs_lock);
193 struct vnodeops *zfs_dvnodeops;
194 struct vnodeops *zfs_fvnodeops;
195 struct vnodeops *zfs_symvnodeops;
196 struct vnodeops *zfs_xdvnodeops;
197 struct vnodeops *zfs_evnodeops;
198 struct vnodeops *zfs_sharevnodeops;
201 zfs_remove_op_tables()
207 (void) vfs_freevfsops_by_type(zfsfstype);
214 vn_freevnodeops(zfs_dvnodeops);
216 vn_freevnodeops(zfs_fvnodeops);
218 vn_freevnodeops(zfs_symvnodeops);
220 vn_freevnodeops(zfs_xdvnodeops);
222 vn_freevnodeops(zfs_evnodeops);
223 if (zfs_sharevnodeops)
224 vn_freevnodeops(zfs_sharevnodeops);
226 zfs_dvnodeops = NULL;
227 zfs_fvnodeops = NULL;
228 zfs_symvnodeops = NULL;
229 zfs_xdvnodeops = NULL;
230 zfs_evnodeops = NULL;
231 zfs_sharevnodeops = NULL;
234 extern const fs_operation_def_t zfs_dvnodeops_template[];
235 extern const fs_operation_def_t zfs_fvnodeops_template[];
236 extern const fs_operation_def_t zfs_xdvnodeops_template[];
237 extern const fs_operation_def_t zfs_symvnodeops_template[];
238 extern const fs_operation_def_t zfs_evnodeops_template[];
239 extern const fs_operation_def_t zfs_sharevnodeops_template[];
242 zfs_create_op_tables()
247 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
248 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
249 * In this case we just return as the ops vectors are already set up.
254 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
259 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
264 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
269 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
274 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
279 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
286 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
289 zfs_acl_ids_t acl_ids;
296 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
297 vattr.va_type = VDIR;
298 vattr.va_mode = S_IFDIR|0555;
299 vattr.va_uid = crgetuid(kcred);
300 vattr.va_gid = crgetgid(kcred);
302 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
303 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
304 sharezp->z_moved = 0;
305 sharezp->z_unlinked = 0;
306 sharezp->z_atime_dirty = 0;
307 sharezp->z_zfsvfs = zfsvfs;
308 sharezp->z_is_sa = zfsvfs->z_use_sa;
314 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
315 kcred, NULL, &acl_ids));
316 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
317 ASSERT3P(zp, ==, sharezp);
318 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
319 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
320 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
321 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
322 zfsvfs->z_shares_dir = sharezp->z_id;
324 zfs_acl_ids_free(&acl_ids);
325 ZTOV(sharezp)->v_count = 0;
326 sa_handle_destroy(sharezp->z_sa_hdl);
327 kmem_cache_free(znode_cache, sharezp);
332 #endif /* HAVE_SHARE */
336 * define a couple of values we need available
337 * for both 64 and 32 bit environments.
340 #define NBITSMINOR64 32
343 #define MAXMAJ64 0xffffffffUL
346 #define MAXMIN64 0xffffffffUL
350 * Create special expldev for ZFS private use.
351 * Can't use standard expldev since it doesn't do
352 * what we want. The standard expldev() takes a
353 * dev32_t in LP64 and expands it to a long dev_t.
354 * We need an interface that takes a dev32_t in ILP32
355 * and expands it to a long dev_t.
358 zfs_expldev(dev_t dev)
361 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
362 return (((uint64_t)major << NBITSMINOR64) |
363 ((minor_t)dev & MAXMIN32));
370 * Special cmpldev for ZFS private use.
371 * Can't use standard cmpldev since it takes
372 * a long dev_t and compresses it to dev32_t in
373 * LP64. We need to do a compaction of a long dev_t
374 * to a dev32_t in ILP32.
377 zfs_cmpldev(uint64_t dev)
380 minor_t minor = (minor_t)dev & MAXMIN64;
381 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
383 if (major > MAXMAJ32 || minor > MAXMIN32)
386 return (((dev32_t)major << NBITSMINOR32) | minor);
393 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
394 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
396 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
397 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
399 mutex_enter(&zp->z_lock);
401 ASSERT(zp->z_sa_hdl == NULL);
402 ASSERT(zp->z_acl_cached == NULL);
403 if (sa_hdl == NULL) {
404 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
405 SA_HDL_SHARED, &zp->z_sa_hdl));
407 zp->z_sa_hdl = sa_hdl;
408 sa_set_userp(sa_hdl, zp);
411 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
414 * Slap on VROOT if we are the root znode
416 if (zp->z_id == zfsvfs->z_root)
417 ZTOV(zp)->v_flag |= VROOT;
419 mutex_exit(&zp->z_lock);
424 zfs_znode_dmu_fini(znode_t *zp)
426 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
428 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
430 sa_handle_destroy(zp->z_sa_hdl);
435 * Construct a new znode/vnode and intialize.
437 * This does not do a call to dmu_set_user() that is
438 * up to the caller to do, in case you don't want to
442 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
443 dmu_object_type_t obj_type, sa_handle_t *hdl)
449 sa_bulk_attr_t bulk[9];
452 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
454 ASSERT(zp->z_dirlocks == NULL);
455 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
459 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
460 * the zfs_znode_move() callback.
464 zp->z_atime_dirty = 0;
466 zp->z_id = db->db_object;
468 zp->z_seq = 0x7A4653;
474 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
476 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
477 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
478 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
480 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
482 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
484 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
485 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
487 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
489 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
492 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
494 sa_handle_destroy(zp->z_sa_hdl);
495 kmem_cache_free(znode_cache, zp);
500 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
502 vp->v_type = IFTOVT((mode_t)mode);
504 switch (vp->v_type) {
506 if (zp->z_pflags & ZFS_XATTR) {
507 vn_setops(vp, zfs_xdvnodeops);
508 vp->v_flag |= V_XATTRDIR;
510 vn_setops(vp, zfs_dvnodeops);
512 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
518 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
519 &rdev, sizeof (rdev)) == 0);
521 vp->v_rdev = zfs_cmpldev(rdev);
527 vn_setops(vp, zfs_fvnodeops);
530 vp->v_flag |= VMODSORT;
531 if (parent == zfsvfs->z_shares_dir) {
532 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
533 vn_setops(vp, zfs_sharevnodeops);
535 vn_setops(vp, zfs_fvnodeops);
539 vn_setops(vp, zfs_symvnodeops);
542 vn_setops(vp, zfs_evnodeops);
546 mutex_enter(&zfsvfs->z_znodes_lock);
547 list_insert_tail(&zfsvfs->z_all_znodes, zp);
550 * Everything else must be valid before assigning z_zfsvfs makes the
551 * znode eligible for zfs_znode_move().
553 zp->z_zfsvfs = zfsvfs;
554 mutex_exit(&zfsvfs->z_znodes_lock);
556 VFS_HOLD(zfsvfs->z_vfs);
560 static uint64_t empty_xattr;
561 static uint64_t pad[4];
562 static zfs_acl_phys_t acl_phys;
564 * Create a new DMU object to hold a zfs znode.
566 * IN: dzp - parent directory for new znode
567 * vap - file attributes for new znode
568 * tx - dmu transaction id for zap operations
569 * cr - credentials of caller
571 * IS_ROOT_NODE - new object will be root
572 * IS_XATTR - new object is an attribute
573 * bonuslen - length of bonus buffer
574 * setaclp - File/Dir initial ACL
575 * fuidp - Tracks fuid allocation.
577 * OUT: zpp - allocated znode
581 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
582 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
584 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
585 uint64_t mode, size, links, parent, pflags;
586 uint64_t dzp_pflags = 0;
588 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
595 dmu_object_type_t obj_type;
596 sa_bulk_attr_t *sa_attrs;
598 zfs_acl_locator_cb_t locate = { 0 };
600 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
602 if (zfsvfs->z_replay) {
603 obj = vap->va_nodeid;
604 now = vap->va_ctime; /* see zfs_replay_create() */
605 gen = vap->va_nblocks; /* ditto */
609 gen = dmu_tx_get_txg(tx);
612 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
613 bonuslen = (obj_type == DMU_OT_SA) ?
614 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
617 * Create a new DMU object.
620 * There's currently no mechanism for pre-reading the blocks that will
621 * be needed to allocate a new object, so we accept the small chance
622 * that there will be an i/o error and we will fail one of the
625 if (vap->va_type == VDIR) {
626 if (zfsvfs->z_replay) {
627 err = zap_create_claim_norm(zfsvfs->z_os, obj,
628 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
629 obj_type, bonuslen, tx);
630 ASSERT3U(err, ==, 0);
632 obj = zap_create_norm(zfsvfs->z_os,
633 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
634 obj_type, bonuslen, tx);
637 if (zfsvfs->z_replay) {
638 err = dmu_object_claim(zfsvfs->z_os, obj,
639 DMU_OT_PLAIN_FILE_CONTENTS, 0,
640 obj_type, bonuslen, tx);
641 ASSERT3U(err, ==, 0);
643 obj = dmu_object_alloc(zfsvfs->z_os,
644 DMU_OT_PLAIN_FILE_CONTENTS, 0,
645 obj_type, bonuslen, tx);
649 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
650 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
653 * If this is the root, fix up the half-initialized parent pointer
654 * to reference the just-allocated physical data area.
656 if (flag & IS_ROOT_NODE) {
659 dzp_pflags = dzp->z_pflags;
663 * If parent is an xattr, so am I.
665 if (dzp_pflags & ZFS_XATTR) {
669 if (zfsvfs->z_use_fuids)
670 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
674 if (vap->va_type == VDIR) {
675 size = 2; /* contents ("." and "..") */
676 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
681 if (vap->va_type == VBLK || vap->va_type == VCHR) {
682 rdev = zfs_expldev(vap->va_rdev);
686 mode = acl_ids->z_mode;
691 * No execs denied will be deterimed when zfs_mode_compute() is called.
693 pflags |= acl_ids->z_aclp->z_hints &
694 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
695 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
697 ZFS_TIME_ENCODE(&now, crtime);
698 ZFS_TIME_ENCODE(&now, ctime);
700 if (vap->va_mask & AT_ATIME) {
701 ZFS_TIME_ENCODE(&vap->va_atime, atime);
703 ZFS_TIME_ENCODE(&now, atime);
706 if (vap->va_mask & AT_MTIME) {
707 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
709 ZFS_TIME_ENCODE(&now, mtime);
712 /* Now add in all of the "SA" attributes */
713 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
717 * Setup the array of attributes to be replaced/set on the new file
719 * order for DMU_OT_ZNODE is critical since it needs to be constructed
720 * in the old znode_phys_t format. Don't change this ordering
722 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
724 if (obj_type == DMU_OT_ZNODE) {
725 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
727 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
729 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
731 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
733 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
735 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
737 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
739 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
742 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
744 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
746 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
748 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
749 &acl_ids->z_fuid, 8);
750 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
751 &acl_ids->z_fgid, 8);
752 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
754 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
756 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
758 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
760 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
762 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
766 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
768 if (obj_type == DMU_OT_ZNODE) {
769 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
772 if (obj_type == DMU_OT_ZNODE ||
773 (vap->va_type == VBLK || vap->va_type == VCHR)) {
774 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
778 if (obj_type == DMU_OT_ZNODE) {
779 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
781 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
782 &acl_ids->z_fuid, 8);
783 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
784 &acl_ids->z_fgid, 8);
785 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
786 sizeof (uint64_t) * 4);
787 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
788 &acl_phys, sizeof (zfs_acl_phys_t));
789 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
790 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
791 &acl_ids->z_aclp->z_acl_count, 8);
792 locate.cb_aclp = acl_ids->z_aclp;
793 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
794 zfs_acl_data_locator, &locate,
795 acl_ids->z_aclp->z_acl_bytes);
796 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
797 acl_ids->z_fuid, acl_ids->z_fgid);
800 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
802 if (!(flag & IS_ROOT_NODE)) {
803 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
804 ASSERT(*zpp != NULL);
807 * If we are creating the root node, the "parent" we
808 * passed in is the znode for the root.
812 (*zpp)->z_sa_hdl = sa_hdl;
815 (*zpp)->z_pflags = pflags;
816 (*zpp)->z_mode = mode;
818 if (vap->va_mask & AT_XVATTR)
819 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
821 if (obj_type == DMU_OT_ZNODE ||
822 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
823 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
824 ASSERT3S(err, ==, 0);
826 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
827 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
831 * zfs_xvattr_set only updates the in-core attributes
832 * it is assumed the caller will be doing an sa_bulk_update
833 * to push the changes out
836 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
840 xoap = xva_getxoptattr(xvap);
843 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
845 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
846 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
847 ×, sizeof (times), tx);
848 XVA_SET_RTN(xvap, XAT_CREATETIME);
850 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
851 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
853 XVA_SET_RTN(xvap, XAT_READONLY);
855 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
856 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
858 XVA_SET_RTN(xvap, XAT_HIDDEN);
860 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
861 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
863 XVA_SET_RTN(xvap, XAT_SYSTEM);
865 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
866 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
868 XVA_SET_RTN(xvap, XAT_ARCHIVE);
870 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
871 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
873 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
875 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
876 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
878 XVA_SET_RTN(xvap, XAT_NOUNLINK);
880 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
881 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
883 XVA_SET_RTN(xvap, XAT_APPENDONLY);
885 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
886 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
888 XVA_SET_RTN(xvap, XAT_NODUMP);
890 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
891 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
893 XVA_SET_RTN(xvap, XAT_OPAQUE);
895 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
896 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
897 xoap->xoa_av_quarantined, zp->z_pflags, tx);
898 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
900 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
901 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
903 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
905 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
906 zfs_sa_set_scanstamp(zp, xvap, tx);
907 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
909 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
910 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
912 XVA_SET_RTN(xvap, XAT_REPARSE);
914 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
915 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
917 XVA_SET_RTN(xvap, XAT_OFFLINE);
919 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
920 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
922 XVA_SET_RTN(xvap, XAT_SPARSE);
927 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
929 dmu_object_info_t doi;
937 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
939 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
941 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
945 dmu_object_info_from_db(db, &doi);
946 if (doi.doi_bonus_type != DMU_OT_SA &&
947 (doi.doi_bonus_type != DMU_OT_ZNODE ||
948 (doi.doi_bonus_type == DMU_OT_ZNODE &&
949 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
950 sa_buf_rele(db, NULL);
951 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
955 hdl = dmu_buf_get_user(db);
957 zp = sa_get_userdata(hdl);
961 * Since "SA" does immediate eviction we
962 * should never find a sa handle that doesn't
963 * know about the znode.
966 ASSERT3P(zp, !=, NULL);
968 mutex_enter(&zp->z_lock);
969 ASSERT3U(zp->z_id, ==, obj_num);
970 if (zp->z_unlinked) {
977 sa_buf_rele(db, NULL);
978 mutex_exit(&zp->z_lock);
979 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
984 * Not found create new znode/vnode
985 * but only if file exists.
987 * There is a small window where zfs_vget() could
988 * find this object while a file create is still in
989 * progress. This is checked for in zfs_znode_alloc()
991 * if zfs_znode_alloc() fails it will drop the hold on the
994 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
995 doi.doi_bonus_type, NULL);
1001 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1006 zfs_rezget(znode_t *zp)
1008 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1009 dmu_object_info_t doi;
1011 uint64_t obj_num = zp->z_id;
1013 sa_bulk_attr_t bulk[8];
1018 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1020 mutex_enter(&zp->z_acl_lock);
1021 if (zp->z_acl_cached) {
1022 zfs_acl_free(zp->z_acl_cached);
1023 zp->z_acl_cached = NULL;
1026 mutex_exit(&zp->z_acl_lock);
1027 ASSERT(zp->z_sa_hdl == NULL);
1028 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1030 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1034 dmu_object_info_from_db(db, &doi);
1035 if (doi.doi_bonus_type != DMU_OT_SA &&
1036 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1037 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1038 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1039 sa_buf_rele(db, NULL);
1040 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1044 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1046 /* reload cached values */
1047 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1048 &gen, sizeof (gen));
1049 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1050 &zp->z_size, sizeof (zp->z_size));
1051 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1052 &zp->z_links, sizeof (zp->z_links));
1053 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1054 &zp->z_pflags, sizeof (zp->z_pflags));
1055 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1056 &zp->z_atime, sizeof (zp->z_atime));
1057 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1058 &zp->z_uid, sizeof (zp->z_uid));
1059 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1060 &zp->z_gid, sizeof (zp->z_gid));
1061 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1062 &mode, sizeof (mode));
1064 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1065 zfs_znode_dmu_fini(zp);
1066 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1072 if (gen != zp->z_gen) {
1073 zfs_znode_dmu_fini(zp);
1074 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1078 zp->z_unlinked = (zp->z_links == 0);
1079 zp->z_blksz = doi.doi_data_block_size;
1081 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1087 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1089 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1090 objset_t *os = zfsvfs->z_os;
1091 uint64_t obj = zp->z_id;
1092 uint64_t acl_obj = zfs_external_acl(zp);
1094 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1096 VERIFY(!zp->z_is_sa);
1097 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1099 VERIFY(0 == dmu_object_free(os, obj, tx));
1100 zfs_znode_dmu_fini(zp);
1101 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1106 zfs_zinactive(znode_t *zp)
1108 vnode_t *vp = ZTOV(zp);
1109 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1110 uint64_t z_id = zp->z_id;
1112 ASSERT(zp->z_sa_hdl);
1115 * Don't allow a zfs_zget() while were trying to release this znode
1117 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1119 mutex_enter(&zp->z_lock);
1120 mutex_enter(&vp->v_lock);
1122 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1124 * If the hold count is greater than zero, somebody has
1125 * obtained a new reference on this znode while we were
1126 * processing it here, so we are done. If we still have
1127 * mapped pages then we are also done, since we don't
1128 * want to inactivate the znode until the pages get pushed.
1130 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1131 * this seems like it would leave the znode hanging with
1132 * no chance to go inactive...
1134 mutex_exit(&vp->v_lock);
1135 mutex_exit(&zp->z_lock);
1136 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1139 mutex_exit(&vp->v_lock);
1142 * If this was the last reference to a file with no links,
1143 * remove the file from the file system.
1145 if (zp->z_unlinked) {
1146 mutex_exit(&zp->z_lock);
1147 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1152 mutex_exit(&zp->z_lock);
1153 zfs_znode_dmu_fini(zp);
1154 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1159 zfs_znode_free(znode_t *zp)
1161 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1163 vn_invalid(ZTOV(zp));
1165 ASSERT(ZTOV(zp)->v_count == 0);
1167 mutex_enter(&zfsvfs->z_znodes_lock);
1168 POINTER_INVALIDATE(&zp->z_zfsvfs);
1169 list_remove(&zfsvfs->z_all_znodes, zp);
1170 mutex_exit(&zfsvfs->z_znodes_lock);
1172 if (zp->z_acl_cached) {
1173 zfs_acl_free(zp->z_acl_cached);
1174 zp->z_acl_cached = NULL;
1177 kmem_cache_free(znode_cache, zp);
1179 VFS_RELE(zfsvfs->z_vfs);
1183 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1184 uint64_t ctime[2], boolean_t have_tx)
1190 if (have_tx) { /* will sa_bulk_update happen really soon? */
1191 zp->z_atime_dirty = 0;
1194 zp->z_atime_dirty = 1;
1197 if (flag & AT_ATIME) {
1198 ZFS_TIME_ENCODE(&now, zp->z_atime);
1201 if (flag & AT_MTIME) {
1202 ZFS_TIME_ENCODE(&now, mtime);
1203 if (zp->z_zfsvfs->z_use_fuids) {
1204 zp->z_pflags |= (ZFS_ARCHIVE |
1209 if (flag & AT_CTIME) {
1210 ZFS_TIME_ENCODE(&now, ctime);
1211 if (zp->z_zfsvfs->z_use_fuids)
1212 zp->z_pflags |= ZFS_ARCHIVE;
1217 * Grow the block size for a file.
1219 * IN: zp - znode of file to free data in.
1220 * size - requested block size
1221 * tx - open transaction.
1223 * NOTE: this function assumes that the znode is write locked.
1226 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1231 if (size <= zp->z_blksz)
1234 * If the file size is already greater than the current blocksize,
1235 * we will not grow. If there is more than one block in a file,
1236 * the blocksize cannot change.
1238 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1241 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1244 if (error == ENOTSUP)
1246 ASSERT3U(error, ==, 0);
1248 /* What blocksize did we actually get? */
1249 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1253 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1254 * be calling back into the fs for a putpage(). E.g.: when truncating
1255 * a file, the pages being "thrown away* don't need to be written out.
1259 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1260 int flags, cred_t *cr)
1267 * Increase the file length
1269 * IN: zp - znode of file to free data in.
1270 * end - new end-of-file
1272 * RETURN: 0 if success
1273 * error code if failure
1276 zfs_extend(znode_t *zp, uint64_t end)
1278 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1285 * We will change zp_size, lock the whole file.
1287 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1290 * Nothing to do if file already at desired length.
1292 if (end <= zp->z_size) {
1293 zfs_range_unlock(rl);
1297 tx = dmu_tx_create(zfsvfs->z_os);
1298 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1299 zfs_sa_upgrade_txholds(tx, zp);
1300 if (end > zp->z_blksz &&
1301 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1303 * We are growing the file past the current block size.
1305 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1306 ASSERT(!ISP2(zp->z_blksz));
1307 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1309 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1311 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1316 error = dmu_tx_assign(tx, TXG_NOWAIT);
1318 if (error == ERESTART) {
1324 zfs_range_unlock(rl);
1329 zfs_grow_blocksize(zp, newblksz, tx);
1333 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1334 &zp->z_size, sizeof (zp->z_size), tx));
1336 zfs_range_unlock(rl);
1344 * Free space in a file.
1346 * IN: zp - znode of file to free data in.
1347 * off - start of section to free.
1348 * len - length of section to free.
1350 * RETURN: 0 if success
1351 * error code if failure
1354 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1356 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1361 * Lock the range being freed.
1363 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1366 * Nothing to do if file already at desired length.
1368 if (off >= zp->z_size) {
1369 zfs_range_unlock(rl);
1373 if (off + len > zp->z_size)
1374 len = zp->z_size - off;
1376 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1378 zfs_range_unlock(rl);
1386 * IN: zp - znode of file to free data in.
1387 * end - new end-of-file.
1389 * RETURN: 0 if success
1390 * error code if failure
1393 zfs_trunc(znode_t *zp, uint64_t end)
1395 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1396 vnode_t *vp = ZTOV(zp);
1400 sa_bulk_attr_t bulk[2];
1404 * We will change zp_size, lock the whole file.
1406 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1409 * Nothing to do if file already at desired length.
1411 if (end >= zp->z_size) {
1412 zfs_range_unlock(rl);
1416 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1418 zfs_range_unlock(rl);
1422 tx = dmu_tx_create(zfsvfs->z_os);
1423 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1424 zfs_sa_upgrade_txholds(tx, zp);
1425 error = dmu_tx_assign(tx, TXG_NOWAIT);
1427 if (error == ERESTART) {
1433 zfs_range_unlock(rl);
1438 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1439 NULL, &zp->z_size, sizeof (zp->z_size));
1442 zp->z_pflags &= ~ZFS_SPARSE;
1443 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1444 NULL, &zp->z_pflags, 8);
1446 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1451 * Clear any mapped pages in the truncated region. This has to
1452 * happen outside of the transaction to avoid the possibility of
1453 * a deadlock with someone trying to push a page that we are
1454 * about to invalidate.
1456 if (vn_has_cached_data(vp)) {
1458 uint64_t start = end & PAGEMASK;
1459 int poff = end & PAGEOFFSET;
1461 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1463 * We need to zero a partial page.
1465 pagezero(pp, poff, PAGESIZE - poff);
1469 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1470 B_INVAL | B_TRUNC, NULL);
1474 zfs_range_unlock(rl);
1480 * Free space in a file
1482 * IN: zp - znode of file to free data in.
1483 * off - start of range
1484 * len - end of range (0 => EOF)
1485 * flag - current file open mode flags.
1486 * log - TRUE if this action should be logged
1488 * RETURN: 0 if success
1489 * error code if failure
1492 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1494 vnode_t *vp = ZTOV(zp);
1496 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1497 zilog_t *zilog = zfsvfs->z_log;
1499 uint64_t mtime[2], ctime[2];
1500 sa_bulk_attr_t bulk[3];
1504 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1505 sizeof (mode))) != 0)
1508 if (off > zp->z_size) {
1509 error = zfs_extend(zp, off+len);
1510 if (error == 0 && log)
1517 * Check for any locks in the region to be freed.
1520 if (MANDLOCK(vp, (mode_t)mode)) {
1521 uint64_t length = (len ? len : zp->z_size - off);
1522 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1527 error = zfs_trunc(zp, off);
1529 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1530 off + len > zp->z_size)
1531 error = zfs_extend(zp, off+len);
1536 tx = dmu_tx_create(zfsvfs->z_os);
1537 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1538 zfs_sa_upgrade_txholds(tx, zp);
1539 error = dmu_tx_assign(tx, TXG_NOWAIT);
1541 if (error == ERESTART) {
1550 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1551 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1552 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1553 NULL, &zp->z_pflags, 8);
1554 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1555 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1558 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1563 #endif /* HAVE_ZPL */
1566 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1568 uint64_t moid, obj, sa_obj, version;
1569 uint64_t sense = ZFS_CASE_SENSITIVE;
1576 znode_t *rootzp = NULL;
1580 zfs_acl_ids_t acl_ids;
1585 #endif /* HAVE_ZPL */
1588 * First attempt to create master node.
1591 * In an empty objset, there are no blocks to read and thus
1592 * there can be no i/o errors (which we assert below).
1594 moid = MASTER_NODE_OBJ;
1595 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1596 DMU_OT_NONE, 0, tx);
1600 * Set starting attributes.
1602 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1604 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1605 /* For the moment we expect all zpl props to be uint64_ts */
1609 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1610 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1611 name = nvpair_name(elem);
1612 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1616 error = zap_update(os, moid, name, 8, 1, &val, tx);
1619 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1621 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1624 ASSERT(version != 0);
1625 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1628 * Create zap object used for SA attribute registration
1631 if (version >= ZPL_VERSION_SA) {
1632 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1633 DMU_OT_NONE, 0, tx);
1634 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1640 * Create a delete queue.
1642 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1644 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1649 * Create root znode. Create minimal znode/vnode/zfsvfs
1650 * to allow zfs_mknode to work.
1652 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1653 vattr.va_type = VDIR;
1654 vattr.va_mode = S_IFDIR|0755;
1655 vattr.va_uid = crgetuid(cr);
1656 vattr.va_gid = crgetgid(cr);
1658 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1659 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1660 rootzp->z_moved = 0;
1661 rootzp->z_unlinked = 0;
1662 rootzp->z_atime_dirty = 0;
1663 rootzp->z_is_sa = USE_SA(version, os);
1669 bzero(&zfsvfs, sizeof (zfsvfs_t));
1672 zfsvfs.z_parent = &zfsvfs;
1673 zfsvfs.z_version = version;
1674 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1675 zfsvfs.z_use_sa = USE_SA(version, os);
1676 zfsvfs.z_norm = norm;
1678 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1679 &zfsvfs.z_attr_table);
1684 * Fold case on file systems that are always or sometimes case
1687 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1688 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1690 /* XXX - This must be destroyed but I'm not quite sure yet so
1691 * I'm just annotating that fact when it's an issue. -Brian */
1692 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1693 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1694 offsetof(znode_t, z_link_node));
1696 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1697 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1699 rootzp->z_zfsvfs = &zfsvfs;
1700 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1701 cr, NULL, &acl_ids));
1702 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1703 ASSERT3P(zp, ==, rootzp);
1704 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1705 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1707 zfs_acl_ids_free(&acl_ids);
1708 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1710 ZTOV(rootzp)->v_count = 0;
1711 sa_handle_destroy(rootzp->z_sa_hdl);
1712 kmem_cache_free(znode_cache, rootzp);
1713 error = zfs_create_share_dir(&zfsvfs, tx);
1715 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1716 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1719 * Create root znode with code free of VFS dependencies
1721 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1722 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1724 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1725 dmu_buf_will_dirty(db, tx);
1728 * Initialize the znode physical data to zero.
1730 ASSERT(db->db_size >= sizeof (znode_phys_t));
1731 bzero(db->db_data, db->db_size);
1734 if (USE_FUIDS(version, os))
1735 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1737 pzp->zp_size = 2; /* "." and ".." */
1739 pzp->zp_parent = obj;
1740 pzp->zp_gen = dmu_tx_get_txg(tx);
1741 pzp->zp_mode = S_IFDIR | 0755;
1742 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1746 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1747 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1748 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1749 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1751 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1754 dmu_buf_rele(db, FTAG);
1755 #endif /* HAVE_ZPL */
1758 #endif /* _KERNEL */
1761 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1763 uint64_t sa_obj = 0;
1766 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1767 if (error != 0 && error != ENOENT)
1770 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1775 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1778 dmu_object_info_t doi;
1781 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1784 dmu_object_info_from_db(*db, &doi);
1785 if ((doi.doi_bonus_type != DMU_OT_SA &&
1786 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1787 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1788 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1789 sa_buf_rele(*db, FTAG);
1793 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1795 sa_buf_rele(*db, FTAG);
1803 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1805 sa_handle_destroy(hdl);
1806 sa_buf_rele(db, FTAG);
1810 * Given an object number, return its parent object number and whether
1811 * or not the object is an extended attribute directory.
1814 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1820 sa_bulk_attr_t bulk[3];
1824 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1825 &parent, sizeof (parent));
1826 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1827 &pflags, sizeof (pflags));
1828 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1829 &mode, sizeof (mode));
1831 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1835 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1841 * Given an object number, return some zpl level statistics
1844 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1847 sa_bulk_attr_t bulk[4];
1850 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1851 &sb->zs_mode, sizeof (sb->zs_mode));
1852 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1853 &sb->zs_gen, sizeof (sb->zs_gen));
1854 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1855 &sb->zs_links, sizeof (sb->zs_links));
1856 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1857 &sb->zs_ctime, sizeof (sb->zs_ctime));
1859 return (sa_bulk_lookup(hdl, bulk, count));
1863 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1864 sa_attr_type_t *sa_table, char *buf, int len)
1866 sa_handle_t *sa_hdl;
1867 sa_handle_t *prevhdl = NULL;
1868 dmu_buf_t *prevdb = NULL;
1869 dmu_buf_t *sa_db = NULL;
1870 char *path = buf + len - 1;
1878 char component[MAXNAMELEN + 2];
1883 zfs_release_sa_handle(prevhdl, prevdb);
1885 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1886 &is_xattrdir)) != 0)
1897 (void) sprintf(component + 1, "<xattrdir>");
1899 error = zap_value_search(osp, pobj, obj,
1900 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1905 complen = strlen(component);
1907 ASSERT(path >= buf);
1908 bcopy(component, path, complen);
1911 if (sa_hdl != hdl) {
1915 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1923 if (sa_hdl != NULL && sa_hdl != hdl) {
1924 ASSERT(sa_db != NULL);
1925 zfs_release_sa_handle(sa_hdl, sa_db);
1929 (void) memmove(buf, path, buf + len - path);
1935 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1937 sa_attr_type_t *sa_table;
1942 error = zfs_sa_setup(osp, &sa_table);
1946 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1950 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1952 zfs_release_sa_handle(hdl, db);
1957 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1960 char *path = buf + len - 1;
1961 sa_attr_type_t *sa_table;
1968 error = zfs_sa_setup(osp, &sa_table);
1972 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1976 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1978 zfs_release_sa_handle(hdl, db);
1982 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1984 zfs_release_sa_handle(hdl, db);
1988 #if defined(_KERNEL) && defined(HAVE_SPL)
1989 EXPORT_SYMBOL(zfs_create_fs);
1990 EXPORT_SYMBOL(zfs_obj_to_path);