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/vnode and intialize.
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)
443 sa_bulk_attr_t bulk[9];
445 #endif /* HAVE_ZPL */
447 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
449 ASSERT(zp->z_dirlocks == NULL);
450 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
454 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
455 * the zfs_znode_move() callback.
459 zp->z_atime_dirty = 0;
461 zp->z_id = db->db_object;
463 zp->z_seq = 0x7A4653;
470 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
472 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
473 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
474 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
476 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
478 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
480 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
481 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
483 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
485 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
488 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
490 sa_handle_destroy(zp->z_sa_hdl);
491 kmem_cache_free(znode_cache, zp);
496 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
498 vp->v_type = IFTOVT((mode_t)mode);
500 switch (vp->v_type) {
502 if (zp->z_pflags & ZFS_XATTR) {
503 vn_setops(vp, zfs_xdvnodeops);
504 vp->v_flag |= V_XATTRDIR;
506 vn_setops(vp, zfs_dvnodeops);
508 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
514 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
515 &rdev, sizeof (rdev)) == 0);
517 vp->v_rdev = zfs_cmpldev(rdev);
523 vn_setops(vp, zfs_fvnodeops);
526 vp->v_flag |= VMODSORT;
527 if (parent == zfsvfs->z_shares_dir) {
528 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
529 vn_setops(vp, zfs_sharevnodeops);
531 vn_setops(vp, zfs_fvnodeops);
535 vn_setops(vp, zfs_symvnodeops);
538 vn_setops(vp, zfs_evnodeops);
541 #endif /* HAVE_ZPL */
542 mutex_enter(&zfsvfs->z_znodes_lock);
543 list_insert_tail(&zfsvfs->z_all_znodes, zp);
546 * Everything else must be valid before assigning z_zfsvfs makes the
547 * znode eligible for zfs_znode_move().
549 zp->z_zfsvfs = zfsvfs;
550 mutex_exit(&zfsvfs->z_znodes_lock);
552 VFS_HOLD(zfsvfs->z_vfs);
556 static uint64_t empty_xattr;
557 static uint64_t pad[4];
558 static zfs_acl_phys_t acl_phys;
560 * Create a new DMU object to hold a zfs znode.
562 * IN: dzp - parent directory for new znode
563 * vap - file attributes for new znode
564 * tx - dmu transaction id for zap operations
565 * cr - credentials of caller
567 * IS_ROOT_NODE - new object will be root
568 * IS_XATTR - new object is an attribute
569 * bonuslen - length of bonus buffer
570 * setaclp - File/Dir initial ACL
571 * fuidp - Tracks fuid allocation.
573 * OUT: zpp - allocated znode
577 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
578 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
580 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
581 uint64_t mode, size, links, parent, pflags;
582 uint64_t dzp_pflags = 0;
584 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
591 dmu_object_type_t obj_type;
592 sa_bulk_attr_t *sa_attrs;
594 zfs_acl_locator_cb_t locate = { 0 };
596 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
598 if (zfsvfs->z_replay) {
599 obj = vap->va_nodeid;
600 now = vap->va_ctime; /* see zfs_replay_create() */
601 gen = vap->va_nblocks; /* ditto */
605 gen = dmu_tx_get_txg(tx);
608 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
609 bonuslen = (obj_type == DMU_OT_SA) ?
610 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
613 * Create a new DMU object.
616 * There's currently no mechanism for pre-reading the blocks that will
617 * be needed to allocate a new object, so we accept the small chance
618 * that there will be an i/o error and we will fail one of the
621 if (vap->va_type == VDIR) {
622 if (zfsvfs->z_replay) {
623 err = zap_create_claim_norm(zfsvfs->z_os, obj,
624 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
625 obj_type, bonuslen, tx);
626 ASSERT3U(err, ==, 0);
628 obj = zap_create_norm(zfsvfs->z_os,
629 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
630 obj_type, bonuslen, tx);
633 if (zfsvfs->z_replay) {
634 err = dmu_object_claim(zfsvfs->z_os, obj,
635 DMU_OT_PLAIN_FILE_CONTENTS, 0,
636 obj_type, bonuslen, tx);
637 ASSERT3U(err, ==, 0);
639 obj = dmu_object_alloc(zfsvfs->z_os,
640 DMU_OT_PLAIN_FILE_CONTENTS, 0,
641 obj_type, bonuslen, tx);
645 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
646 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
649 * If this is the root, fix up the half-initialized parent pointer
650 * to reference the just-allocated physical data area.
652 if (flag & IS_ROOT_NODE) {
655 dzp_pflags = dzp->z_pflags;
659 * If parent is an xattr, so am I.
661 if (dzp_pflags & ZFS_XATTR) {
665 if (zfsvfs->z_use_fuids)
666 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
670 if (vap->va_type == VDIR) {
671 size = 2; /* contents ("." and "..") */
672 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
677 if (vap->va_type == VBLK || vap->va_type == VCHR) {
678 rdev = zfs_expldev(vap->va_rdev);
682 mode = acl_ids->z_mode;
687 * No execs denied will be deterimed when zfs_mode_compute() is called.
689 pflags |= acl_ids->z_aclp->z_hints &
690 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
691 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
693 ZFS_TIME_ENCODE(&now, crtime);
694 ZFS_TIME_ENCODE(&now, ctime);
696 if (vap->va_mask & AT_ATIME) {
697 ZFS_TIME_ENCODE(&vap->va_atime, atime);
699 ZFS_TIME_ENCODE(&now, atime);
702 if (vap->va_mask & AT_MTIME) {
703 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
705 ZFS_TIME_ENCODE(&now, mtime);
708 /* Now add in all of the "SA" attributes */
709 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
713 * Setup the array of attributes to be replaced/set on the new file
715 * order for DMU_OT_ZNODE is critical since it needs to be constructed
716 * in the old znode_phys_t format. Don't change this ordering
718 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
720 if (obj_type == DMU_OT_ZNODE) {
721 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
723 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
725 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
727 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
729 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
731 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
733 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
735 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
738 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
740 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
742 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
744 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
745 &acl_ids->z_fuid, 8);
746 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
747 &acl_ids->z_fgid, 8);
748 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
750 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
752 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
754 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
756 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
758 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
762 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
764 if (obj_type == DMU_OT_ZNODE) {
765 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
768 if (obj_type == DMU_OT_ZNODE ||
769 (vap->va_type == VBLK || vap->va_type == VCHR)) {
770 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
774 if (obj_type == DMU_OT_ZNODE) {
775 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
777 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
778 &acl_ids->z_fuid, 8);
779 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
780 &acl_ids->z_fgid, 8);
781 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
782 sizeof (uint64_t) * 4);
783 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
784 &acl_phys, sizeof (zfs_acl_phys_t));
785 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
786 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
787 &acl_ids->z_aclp->z_acl_count, 8);
788 locate.cb_aclp = acl_ids->z_aclp;
789 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
790 zfs_acl_data_locator, &locate,
791 acl_ids->z_aclp->z_acl_bytes);
792 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
793 acl_ids->z_fuid, acl_ids->z_fgid);
796 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
798 if (!(flag & IS_ROOT_NODE)) {
799 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
800 ASSERT(*zpp != NULL);
803 * If we are creating the root node, the "parent" we
804 * passed in is the znode for the root.
808 (*zpp)->z_sa_hdl = sa_hdl;
811 (*zpp)->z_pflags = pflags;
812 (*zpp)->z_mode = mode;
814 if (vap->va_mask & AT_XVATTR)
815 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
817 if (obj_type == DMU_OT_ZNODE ||
818 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
819 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
820 ASSERT3S(err, ==, 0);
822 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
823 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
827 * zfs_xvattr_set only updates the in-core attributes
828 * it is assumed the caller will be doing an sa_bulk_update
829 * to push the changes out
832 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
837 xoap = xva_getxoptattr(xvap);
840 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
842 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
843 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
844 ×, sizeof (times), tx);
845 XVA_SET_RTN(xvap, XAT_CREATETIME);
847 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
848 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
850 XVA_SET_RTN(xvap, XAT_READONLY);
852 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
853 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
855 XVA_SET_RTN(xvap, XAT_HIDDEN);
857 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
858 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
860 XVA_SET_RTN(xvap, XAT_SYSTEM);
862 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
863 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
865 XVA_SET_RTN(xvap, XAT_ARCHIVE);
867 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
868 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
870 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
872 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
873 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
875 XVA_SET_RTN(xvap, XAT_NOUNLINK);
877 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
878 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
880 XVA_SET_RTN(xvap, XAT_APPENDONLY);
882 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
883 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
885 XVA_SET_RTN(xvap, XAT_NODUMP);
887 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
888 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
890 XVA_SET_RTN(xvap, XAT_OPAQUE);
892 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
893 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
894 xoap->xoa_av_quarantined, zp->z_pflags, tx);
895 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
897 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
898 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
900 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
902 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
903 zfs_sa_set_scanstamp(zp, xvap, tx);
904 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
906 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
907 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
909 XVA_SET_RTN(xvap, XAT_REPARSE);
911 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
912 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
914 XVA_SET_RTN(xvap, XAT_OFFLINE);
916 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
917 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
919 XVA_SET_RTN(xvap, XAT_SPARSE);
921 #endif /* HAVE_XVATTR */
925 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
927 dmu_object_info_t doi;
935 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
937 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
939 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
943 dmu_object_info_from_db(db, &doi);
944 if (doi.doi_bonus_type != DMU_OT_SA &&
945 (doi.doi_bonus_type != DMU_OT_ZNODE ||
946 (doi.doi_bonus_type == DMU_OT_ZNODE &&
947 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
948 sa_buf_rele(db, NULL);
949 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
953 hdl = dmu_buf_get_user(db);
955 zp = sa_get_userdata(hdl);
959 * Since "SA" does immediate eviction we
960 * should never find a sa handle that doesn't
961 * know about the znode.
964 ASSERT3P(zp, !=, NULL);
966 mutex_enter(&zp->z_lock);
967 ASSERT3U(zp->z_id, ==, obj_num);
968 if (zp->z_unlinked) {
975 sa_buf_rele(db, NULL);
976 mutex_exit(&zp->z_lock);
977 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
982 * Not found create new znode/vnode
983 * but only if file exists.
985 * There is a small window where zfs_vget() could
986 * find this object while a file create is still in
987 * progress. This is checked for in zfs_znode_alloc()
989 * if zfs_znode_alloc() fails it will drop the hold on the
992 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
993 doi.doi_bonus_type, NULL);
999 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1004 zfs_rezget(znode_t *zp)
1006 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1007 dmu_object_info_t doi;
1009 uint64_t obj_num = zp->z_id;
1011 sa_bulk_attr_t bulk[8];
1016 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1018 mutex_enter(&zp->z_acl_lock);
1019 if (zp->z_acl_cached) {
1020 zfs_acl_free(zp->z_acl_cached);
1021 zp->z_acl_cached = NULL;
1024 mutex_exit(&zp->z_acl_lock);
1025 ASSERT(zp->z_sa_hdl == NULL);
1026 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1028 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1032 dmu_object_info_from_db(db, &doi);
1033 if (doi.doi_bonus_type != DMU_OT_SA &&
1034 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1035 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1036 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1037 sa_buf_rele(db, NULL);
1038 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1042 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1044 /* reload cached values */
1045 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1046 &gen, sizeof (gen));
1047 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1048 &zp->z_size, sizeof (zp->z_size));
1049 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1050 &zp->z_links, sizeof (zp->z_links));
1051 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1052 &zp->z_pflags, sizeof (zp->z_pflags));
1053 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1054 &zp->z_atime, sizeof (zp->z_atime));
1055 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1056 &zp->z_uid, sizeof (zp->z_uid));
1057 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1058 &zp->z_gid, sizeof (zp->z_gid));
1059 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1060 &mode, sizeof (mode));
1062 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1063 zfs_znode_dmu_fini(zp);
1064 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1070 if (gen != zp->z_gen) {
1071 zfs_znode_dmu_fini(zp);
1072 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1076 zp->z_unlinked = (zp->z_links == 0);
1077 zp->z_blksz = doi.doi_data_block_size;
1079 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1085 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1087 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1088 objset_t *os = zfsvfs->z_os;
1089 uint64_t obj = zp->z_id;
1090 uint64_t acl_obj = zfs_external_acl(zp);
1092 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1094 VERIFY(!zp->z_is_sa);
1095 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1097 VERIFY(0 == dmu_object_free(os, obj, tx));
1098 zfs_znode_dmu_fini(zp);
1099 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1104 zfs_zinactive(znode_t *zp)
1106 vnode_t *vp = ZTOV(zp);
1107 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1108 uint64_t z_id = zp->z_id;
1110 ASSERT(zp->z_sa_hdl);
1113 * Don't allow a zfs_zget() while were trying to release this znode
1115 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1117 mutex_enter(&zp->z_lock);
1118 mutex_enter(&vp->v_lock);
1120 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1122 * If the hold count is greater than zero, somebody has
1123 * obtained a new reference on this znode while we were
1124 * processing it here, so we are done. If we still have
1125 * mapped pages then we are also done, since we don't
1126 * want to inactivate the znode until the pages get pushed.
1128 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1129 * this seems like it would leave the znode hanging with
1130 * no chance to go inactive...
1132 mutex_exit(&vp->v_lock);
1133 mutex_exit(&zp->z_lock);
1134 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1137 mutex_exit(&vp->v_lock);
1140 * If this was the last reference to a file with no links,
1141 * remove the file from the file system.
1143 if (zp->z_unlinked) {
1144 mutex_exit(&zp->z_lock);
1145 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1150 mutex_exit(&zp->z_lock);
1151 zfs_znode_dmu_fini(zp);
1152 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1157 zfs_znode_free(znode_t *zp)
1159 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1161 vn_invalid(ZTOV(zp));
1163 ASSERT(ZTOV(zp)->v_count == 0);
1165 mutex_enter(&zfsvfs->z_znodes_lock);
1166 POINTER_INVALIDATE(&zp->z_zfsvfs);
1167 list_remove(&zfsvfs->z_all_znodes, zp);
1168 mutex_exit(&zfsvfs->z_znodes_lock);
1170 if (zp->z_acl_cached) {
1171 zfs_acl_free(zp->z_acl_cached);
1172 zp->z_acl_cached = NULL;
1175 kmem_cache_free(znode_cache, zp);
1177 VFS_RELE(zfsvfs->z_vfs);
1181 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1182 uint64_t ctime[2], boolean_t have_tx)
1188 if (have_tx) { /* will sa_bulk_update happen really soon? */
1189 zp->z_atime_dirty = 0;
1192 zp->z_atime_dirty = 1;
1195 if (flag & AT_ATIME) {
1196 ZFS_TIME_ENCODE(&now, zp->z_atime);
1199 if (flag & AT_MTIME) {
1200 ZFS_TIME_ENCODE(&now, mtime);
1201 if (zp->z_zfsvfs->z_use_fuids) {
1202 zp->z_pflags |= (ZFS_ARCHIVE |
1207 if (flag & AT_CTIME) {
1208 ZFS_TIME_ENCODE(&now, ctime);
1209 if (zp->z_zfsvfs->z_use_fuids)
1210 zp->z_pflags |= ZFS_ARCHIVE;
1215 * Grow the block size for a file.
1217 * IN: zp - znode of file to free data in.
1218 * size - requested block size
1219 * tx - open transaction.
1221 * NOTE: this function assumes that the znode is write locked.
1224 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1229 if (size <= zp->z_blksz)
1232 * If the file size is already greater than the current blocksize,
1233 * we will not grow. If there is more than one block in a file,
1234 * the blocksize cannot change.
1236 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1239 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1242 if (error == ENOTSUP)
1244 ASSERT3U(error, ==, 0);
1246 /* What blocksize did we actually get? */
1247 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1252 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1253 * be calling back into the fs for a putpage(). E.g.: when truncating
1254 * a file, the pages being "thrown away* don't need to be written out.
1258 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1259 int flags, cred_t *cr)
1264 #endif /* HAVE_ZPL */
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;
1397 vnode_t *vp = ZTOV(zp);
1398 #endif /* HAVE_ZPL */
1402 sa_bulk_attr_t bulk[2];
1406 * We will change zp_size, lock the whole file.
1408 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1411 * Nothing to do if file already at desired length.
1413 if (end >= zp->z_size) {
1414 zfs_range_unlock(rl);
1418 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1420 zfs_range_unlock(rl);
1424 tx = dmu_tx_create(zfsvfs->z_os);
1425 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1426 zfs_sa_upgrade_txholds(tx, zp);
1427 error = dmu_tx_assign(tx, TXG_NOWAIT);
1429 if (error == ERESTART) {
1435 zfs_range_unlock(rl);
1440 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1441 NULL, &zp->z_size, sizeof (zp->z_size));
1444 zp->z_pflags &= ~ZFS_SPARSE;
1445 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1446 NULL, &zp->z_pflags, 8);
1448 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1454 * Clear any mapped pages in the truncated region. This has to
1455 * happen outside of the transaction to avoid the possibility of
1456 * a deadlock with someone trying to push a page that we are
1457 * about to invalidate.
1459 if (vn_has_cached_data(vp)) {
1461 uint64_t start = end & PAGEMASK;
1462 int poff = end & PAGEOFFSET;
1464 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1466 * We need to zero a partial page.
1468 pagezero(pp, poff, PAGESIZE - poff);
1472 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1473 B_INVAL | B_TRUNC, NULL);
1476 #endif /* HAVE_ZPL */
1478 zfs_range_unlock(rl);
1484 * Free space in a file
1486 * IN: zp - znode of file to free data in.
1487 * off - start of range
1488 * len - end of range (0 => EOF)
1489 * flag - current file open mode flags.
1490 * log - TRUE if this action should be logged
1492 * RETURN: 0 if success
1493 * error code if failure
1496 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1499 vnode_t *vp = ZTOV(zp);
1500 #endif /* HAVE_ZPL */
1502 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1503 zilog_t *zilog = zfsvfs->z_log;
1505 uint64_t mtime[2], ctime[2];
1506 sa_bulk_attr_t bulk[3];
1510 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1511 sizeof (mode))) != 0)
1514 if (off > zp->z_size) {
1515 error = zfs_extend(zp, off+len);
1516 if (error == 0 && log)
1524 * Check for any locks in the region to be freed.
1527 if (MANDLOCK(vp, (mode_t)mode)) {
1528 uint64_t length = (len ? len : zp->z_size - off);
1529 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1532 #endif /* HAVE_ZPL */
1535 error = zfs_trunc(zp, off);
1537 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1538 off + len > zp->z_size)
1539 error = zfs_extend(zp, off+len);
1544 tx = dmu_tx_create(zfsvfs->z_os);
1545 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1546 zfs_sa_upgrade_txholds(tx, zp);
1547 error = dmu_tx_assign(tx, TXG_NOWAIT);
1549 if (error == ERESTART) {
1558 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1559 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1560 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1561 NULL, &zp->z_pflags, 8);
1562 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1563 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1566 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1573 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1575 uint64_t moid, obj, sa_obj, version;
1576 uint64_t sense = ZFS_CASE_SENSITIVE;
1583 znode_t *rootzp = NULL;
1587 zfs_acl_ids_t acl_ids;
1592 #endif /* HAVE_ZPL */
1595 * First attempt to create master node.
1598 * In an empty objset, there are no blocks to read and thus
1599 * there can be no i/o errors (which we assert below).
1601 moid = MASTER_NODE_OBJ;
1602 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1603 DMU_OT_NONE, 0, tx);
1607 * Set starting attributes.
1609 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1611 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1612 /* For the moment we expect all zpl props to be uint64_ts */
1616 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1617 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1618 name = nvpair_name(elem);
1619 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1623 error = zap_update(os, moid, name, 8, 1, &val, tx);
1626 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1628 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1631 ASSERT(version != 0);
1632 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1635 * Create zap object used for SA attribute registration
1638 if (version >= ZPL_VERSION_SA) {
1639 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1640 DMU_OT_NONE, 0, tx);
1641 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1647 * Create a delete queue.
1649 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1651 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1656 * Create root znode. Create minimal znode/vnode/zfsvfs
1657 * to allow zfs_mknode to work.
1659 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1660 vattr.va_type = VDIR;
1661 vattr.va_mode = S_IFDIR|0755;
1662 vattr.va_uid = crgetuid(cr);
1663 vattr.va_gid = crgetgid(cr);
1665 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1666 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1667 rootzp->z_moved = 0;
1668 rootzp->z_unlinked = 0;
1669 rootzp->z_atime_dirty = 0;
1670 rootzp->z_is_sa = USE_SA(version, os);
1676 bzero(&zfsvfs, sizeof (zfsvfs_t));
1679 zfsvfs.z_parent = &zfsvfs;
1680 zfsvfs.z_version = version;
1681 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1682 zfsvfs.z_use_sa = USE_SA(version, os);
1683 zfsvfs.z_norm = norm;
1685 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1686 &zfsvfs.z_attr_table);
1691 * Fold case on file systems that are always or sometimes case
1694 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1695 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1697 /* XXX - This must be destroyed but I'm not quite sure yet so
1698 * I'm just annotating that fact when it's an issue. -Brian */
1699 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1700 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1701 offsetof(znode_t, z_link_node));
1703 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1704 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1706 rootzp->z_zfsvfs = &zfsvfs;
1707 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1708 cr, NULL, &acl_ids));
1709 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1710 ASSERT3P(zp, ==, rootzp);
1711 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1712 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1714 zfs_acl_ids_free(&acl_ids);
1715 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1717 ZTOV(rootzp)->v_count = 0;
1718 sa_handle_destroy(rootzp->z_sa_hdl);
1719 kmem_cache_free(znode_cache, rootzp);
1720 error = zfs_create_share_dir(&zfsvfs, tx);
1722 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1723 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1726 * Create root znode with code free of VFS dependencies
1728 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1729 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1731 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1732 dmu_buf_will_dirty(db, tx);
1735 * Initialize the znode physical data to zero.
1737 ASSERT(db->db_size >= sizeof (znode_phys_t));
1738 bzero(db->db_data, db->db_size);
1741 if (USE_FUIDS(version, os))
1742 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1744 pzp->zp_size = 2; /* "." and ".." */
1746 pzp->zp_parent = obj;
1747 pzp->zp_gen = dmu_tx_get_txg(tx);
1748 pzp->zp_mode = S_IFDIR | 0755;
1749 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1753 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1754 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1755 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1756 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1758 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1761 dmu_buf_rele(db, FTAG);
1762 #endif /* HAVE_ZPL */
1765 #endif /* _KERNEL */
1768 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1770 uint64_t sa_obj = 0;
1773 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1774 if (error != 0 && error != ENOENT)
1777 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1782 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1785 dmu_object_info_t doi;
1788 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1791 dmu_object_info_from_db(*db, &doi);
1792 if ((doi.doi_bonus_type != DMU_OT_SA &&
1793 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1794 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1795 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1796 sa_buf_rele(*db, FTAG);
1800 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1802 sa_buf_rele(*db, FTAG);
1810 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1812 sa_handle_destroy(hdl);
1813 sa_buf_rele(db, FTAG);
1817 * Given an object number, return its parent object number and whether
1818 * or not the object is an extended attribute directory.
1821 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1827 sa_bulk_attr_t bulk[3];
1831 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1832 &parent, sizeof (parent));
1833 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1834 &pflags, sizeof (pflags));
1835 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1836 &mode, sizeof (mode));
1838 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1842 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1848 * Given an object number, return some zpl level statistics
1851 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1854 sa_bulk_attr_t bulk[4];
1857 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1858 &sb->zs_mode, sizeof (sb->zs_mode));
1859 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1860 &sb->zs_gen, sizeof (sb->zs_gen));
1861 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1862 &sb->zs_links, sizeof (sb->zs_links));
1863 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1864 &sb->zs_ctime, sizeof (sb->zs_ctime));
1866 return (sa_bulk_lookup(hdl, bulk, count));
1870 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1871 sa_attr_type_t *sa_table, char *buf, int len)
1873 sa_handle_t *sa_hdl;
1874 sa_handle_t *prevhdl = NULL;
1875 dmu_buf_t *prevdb = NULL;
1876 dmu_buf_t *sa_db = NULL;
1877 char *path = buf + len - 1;
1885 char component[MAXNAMELEN + 2];
1890 zfs_release_sa_handle(prevhdl, prevdb);
1892 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1893 &is_xattrdir)) != 0)
1904 (void) sprintf(component + 1, "<xattrdir>");
1906 error = zap_value_search(osp, pobj, obj,
1907 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1912 complen = strlen(component);
1914 ASSERT(path >= buf);
1915 bcopy(component, path, complen);
1918 if (sa_hdl != hdl) {
1922 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1930 if (sa_hdl != NULL && sa_hdl != hdl) {
1931 ASSERT(sa_db != NULL);
1932 zfs_release_sa_handle(sa_hdl, sa_db);
1936 (void) memmove(buf, path, buf + len - path);
1942 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1944 sa_attr_type_t *sa_table;
1949 error = zfs_sa_setup(osp, &sa_table);
1953 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1957 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1959 zfs_release_sa_handle(hdl, db);
1964 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1967 char *path = buf + len - 1;
1968 sa_attr_type_t *sa_table;
1975 error = zfs_sa_setup(osp, &sa_table);
1979 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1983 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1985 zfs_release_sa_handle(hdl, db);
1989 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1991 zfs_release_sa_handle(hdl, db);
1995 #if defined(_KERNEL) && defined(HAVE_SPL)
1996 EXPORT_SYMBOL(zfs_create_fs);
1997 EXPORT_SYMBOL(zfs_obj_to_path);