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 */
26 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
29 #include <sys/types.h>
30 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/sysmacros.h>
34 #include <sys/resource.h>
35 #include <sys/mntent.h>
36 #include <sys/u8_textprep.h>
37 #include <sys/dsl_dataset.h>
39 #include <sys/vnode.h>
42 #include <sys/errno.h>
43 #include <sys/unistd.h>
44 #include <sys/atomic.h>
45 #include <sys/zfs_dir.h>
46 #include <sys/zfs_acl.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/zfs_rlock.h>
49 #include <sys/zfs_fuid.h>
50 #include <sys/dnode.h>
51 #include <sys/fs/zfs.h>
52 #include <sys/kidmap.h>
56 #include <sys/refcount.h>
59 #include <sys/zfs_znode.h>
61 #include <sys/zfs_sa.h>
62 #include <sys/zfs_stat.h>
63 #include <sys/refcount.h>
66 #include "zfs_comutil.h"
68 /* Used by fstat(1). */
69 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 0, sizeof(znode_t),
73 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
74 * turned on when DEBUG is also defined.
81 #define ZNODE_STAT_ADD(stat) ((stat)++)
83 #define ZNODE_STAT_ADD(stat) /* nothing */
84 #endif /* ZNODE_STATS */
87 * Functions needed for userland (ie: libzpool) are not put under
88 * #ifdef_KERNEL; the rest of the functions have dependencies
89 * (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);
111 extern struct vop_vector zfs_vnodeops;
112 extern struct vop_vector zfs_fifoops;
113 extern struct vop_vector zfs_shareops;
116 * XXX: We cannot use this function as a cache constructor, because
117 * there is one global cache for all file systems and we need
118 * to pass vfsp here, which is not possible, because argument
119 * 'cdrarg' is defined at kmem_cache_create() time.
123 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
130 POINTER_INVALIDATE(&zp->z_zfsvfs);
131 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
134 error = getnewvnode("zfs", vfsp, &zfs_vnodeops, &vp);
135 if (error != 0 && (kmflags & KM_NOSLEEP))
138 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
140 vp->v_data = (caddr_t)zp;
147 list_link_init(&zp->z_link_node);
149 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
150 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
151 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
152 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
154 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
155 avl_create(&zp->z_range_avl, zfs_range_compare,
156 sizeof (rl_t), offsetof(rl_t, r_node));
158 zp->z_dirlocks = NULL;
159 zp->z_acl_cached = NULL;
166 zfs_znode_cache_destructor(void *buf, void *arg)
170 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
171 ASSERT(ZTOV(zp) == NULL);
173 ASSERT(!list_link_active(&zp->z_link_node));
174 mutex_destroy(&zp->z_lock);
175 rw_destroy(&zp->z_parent_lock);
176 rw_destroy(&zp->z_name_lock);
177 mutex_destroy(&zp->z_acl_lock);
178 avl_destroy(&zp->z_range_avl);
179 mutex_destroy(&zp->z_range_lock);
181 ASSERT(zp->z_dirlocks == NULL);
182 ASSERT(zp->z_acl_cached == NULL);
187 uint64_t zms_zfsvfs_invalid;
188 uint64_t zms_zfsvfs_recheck1;
189 uint64_t zms_zfsvfs_unmounted;
190 uint64_t zms_zfsvfs_recheck2;
191 uint64_t zms_obj_held;
192 uint64_t zms_vnode_locked;
193 uint64_t zms_not_only_dnlc;
195 #endif /* ZNODE_STATS */
199 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
204 nzp->z_zfsvfs = ozp->z_zfsvfs;
208 nzp->z_vnode = ozp->z_vnode;
209 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
210 ZTOV(ozp)->v_data = ozp;
211 ZTOV(nzp)->v_data = nzp;
213 nzp->z_id = ozp->z_id;
214 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
215 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
216 nzp->z_unlinked = ozp->z_unlinked;
217 nzp->z_atime_dirty = ozp->z_atime_dirty;
218 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
219 nzp->z_blksz = ozp->z_blksz;
220 nzp->z_seq = ozp->z_seq;
221 nzp->z_mapcnt = ozp->z_mapcnt;
222 nzp->z_gen = ozp->z_gen;
223 nzp->z_sync_cnt = ozp->z_sync_cnt;
224 nzp->z_is_sa = ozp->z_is_sa;
225 nzp->z_sa_hdl = ozp->z_sa_hdl;
226 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
227 nzp->z_links = ozp->z_links;
228 nzp->z_size = ozp->z_size;
229 nzp->z_pflags = ozp->z_pflags;
230 nzp->z_uid = ozp->z_uid;
231 nzp->z_gid = ozp->z_gid;
232 nzp->z_mode = ozp->z_mode;
235 * Since this is just an idle znode and kmem is already dealing with
236 * memory pressure, release any cached ACL.
238 if (ozp->z_acl_cached) {
239 zfs_acl_free(ozp->z_acl_cached);
240 ozp->z_acl_cached = NULL;
243 sa_set_userp(nzp->z_sa_hdl, nzp);
246 * Invalidate the original znode by clearing fields that provide a
247 * pointer back to the znode. Set the low bit of the vfs pointer to
248 * ensure that zfs_znode_move() recognizes the znode as invalid in any
249 * subsequent callback.
251 ozp->z_sa_hdl = NULL;
252 POINTER_INVALIDATE(&ozp->z_zfsvfs);
258 ozp->z_moved = (uint8_t)-1;
263 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
265 znode_t *ozp = buf, *nzp = newbuf;
270 * The znode is on the file system's list of known znodes if the vfs
271 * pointer is valid. We set the low bit of the vfs pointer when freeing
272 * the znode to invalidate it, and the memory patterns written by kmem
273 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
274 * created znode sets the vfs pointer last of all to indicate that the
275 * znode is known and in a valid state to be moved by this function.
277 zfsvfs = ozp->z_zfsvfs;
278 if (!POINTER_IS_VALID(zfsvfs)) {
279 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
280 return (KMEM_CBRC_DONT_KNOW);
284 * Close a small window in which it's possible that the filesystem could
285 * be unmounted and freed, and zfsvfs, though valid in the previous
286 * statement, could point to unrelated memory by the time we try to
287 * prevent the filesystem from being unmounted.
289 rw_enter(&zfsvfs_lock, RW_WRITER);
290 if (zfsvfs != ozp->z_zfsvfs) {
291 rw_exit(&zfsvfs_lock);
292 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
293 return (KMEM_CBRC_DONT_KNOW);
297 * If the znode is still valid, then so is the file system. We know that
298 * no valid file system can be freed while we hold zfsvfs_lock, so we
299 * can safely ensure that the filesystem is not and will not be
300 * unmounted. The next statement is equivalent to ZFS_ENTER().
302 rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
303 if (zfsvfs->z_unmounted) {
305 rw_exit(&zfsvfs_lock);
306 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
307 return (KMEM_CBRC_DONT_KNOW);
309 rw_exit(&zfsvfs_lock);
311 mutex_enter(&zfsvfs->z_znodes_lock);
313 * Recheck the vfs pointer in case the znode was removed just before
314 * acquiring the lock.
316 if (zfsvfs != ozp->z_zfsvfs) {
317 mutex_exit(&zfsvfs->z_znodes_lock);
319 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
320 return (KMEM_CBRC_DONT_KNOW);
324 * At this point we know that as long as we hold z_znodes_lock, the
325 * znode cannot be freed and fields within the znode can be safely
326 * accessed. Now, prevent a race with zfs_zget().
328 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
329 mutex_exit(&zfsvfs->z_znodes_lock);
331 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
332 return (KMEM_CBRC_LATER);
336 if (mutex_tryenter(&vp->v_lock) == 0) {
337 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
338 mutex_exit(&zfsvfs->z_znodes_lock);
340 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
341 return (KMEM_CBRC_LATER);
344 /* Only move znodes that are referenced _only_ by the DNLC. */
345 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
346 mutex_exit(&vp->v_lock);
347 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
348 mutex_exit(&zfsvfs->z_znodes_lock);
350 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
351 return (KMEM_CBRC_LATER);
355 * The znode is known and in a valid state to move. We're holding the
356 * locks needed to execute the critical section.
358 zfs_znode_move_impl(ozp, nzp);
359 mutex_exit(&vp->v_lock);
360 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
362 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
363 mutex_exit(&zfsvfs->z_znodes_lock);
366 return (KMEM_CBRC_YES);
376 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
377 ASSERT(znode_cache == NULL);
378 znode_cache = kmem_cache_create("zfs_znode_cache",
379 sizeof (znode_t), 0, /* zfs_znode_cache_constructor */ NULL,
380 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
381 kmem_cache_set_move(znode_cache, zfs_znode_move);
389 * Cleanup vfs & vnode ops
391 zfs_remove_op_tables();
398 kmem_cache_destroy(znode_cache);
400 rw_destroy(&zfsvfs_lock);
404 struct vnodeops *zfs_dvnodeops;
405 struct vnodeops *zfs_fvnodeops;
406 struct vnodeops *zfs_symvnodeops;
407 struct vnodeops *zfs_xdvnodeops;
408 struct vnodeops *zfs_evnodeops;
409 struct vnodeops *zfs_sharevnodeops;
412 zfs_remove_op_tables()
418 (void) vfs_freevfsops_by_type(zfsfstype);
425 vn_freevnodeops(zfs_dvnodeops);
427 vn_freevnodeops(zfs_fvnodeops);
429 vn_freevnodeops(zfs_symvnodeops);
431 vn_freevnodeops(zfs_xdvnodeops);
433 vn_freevnodeops(zfs_evnodeops);
434 if (zfs_sharevnodeops)
435 vn_freevnodeops(zfs_sharevnodeops);
437 zfs_dvnodeops = NULL;
438 zfs_fvnodeops = NULL;
439 zfs_symvnodeops = NULL;
440 zfs_xdvnodeops = NULL;
441 zfs_evnodeops = NULL;
442 zfs_sharevnodeops = NULL;
445 extern const fs_operation_def_t zfs_dvnodeops_template[];
446 extern const fs_operation_def_t zfs_fvnodeops_template[];
447 extern const fs_operation_def_t zfs_xdvnodeops_template[];
448 extern const fs_operation_def_t zfs_symvnodeops_template[];
449 extern const fs_operation_def_t zfs_evnodeops_template[];
450 extern const fs_operation_def_t zfs_sharevnodeops_template[];
453 zfs_create_op_tables()
458 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
459 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
460 * In this case we just return as the ops vectors are already set up.
465 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
470 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
475 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
480 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
485 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
490 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
498 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
500 zfs_acl_ids_t acl_ids;
507 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
508 vattr.va_type = VDIR;
509 vattr.va_mode = S_IFDIR|0555;
510 vattr.va_uid = crgetuid(kcred);
511 vattr.va_gid = crgetgid(kcred);
513 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
514 zfs_znode_cache_constructor(sharezp, zfsvfs->z_parent->z_vfs, 0);
515 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
516 sharezp->z_moved = 0;
517 sharezp->z_unlinked = 0;
518 sharezp->z_atime_dirty = 0;
519 sharezp->z_zfsvfs = zfsvfs;
520 sharezp->z_is_sa = zfsvfs->z_use_sa;
522 sharezp->z_vnode = &vnode;
523 vnode.v_data = sharezp;
528 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
529 kcred, NULL, &acl_ids));
530 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
531 ASSERT3P(zp, ==, sharezp);
532 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
533 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
534 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
535 zfsvfs->z_shares_dir = sharezp->z_id;
537 zfs_acl_ids_free(&acl_ids);
538 ZTOV(sharezp)->v_data = NULL;
539 ZTOV(sharezp)->v_count = 0;
540 ZTOV(sharezp)->v_holdcnt = 0;
542 sa_handle_destroy(sharezp->z_sa_hdl);
543 sharezp->z_vnode = NULL;
544 kmem_cache_free(znode_cache, sharezp);
550 * define a couple of values we need available
551 * for both 64 and 32 bit environments.
554 #define NBITSMINOR64 32
557 #define MAXMAJ64 0xffffffffUL
560 #define MAXMIN64 0xffffffffUL
564 * Create special expldev for ZFS private use.
565 * Can't use standard expldev since it doesn't do
566 * what we want. The standard expldev() takes a
567 * dev32_t in LP64 and expands it to a long dev_t.
568 * We need an interface that takes a dev32_t in ILP32
569 * and expands it to a long dev_t.
572 zfs_expldev(dev_t dev)
574 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
577 * Special cmpldev for ZFS private use.
578 * Can't use standard cmpldev since it takes
579 * a long dev_t and compresses it to dev32_t in
580 * LP64. We need to do a compaction of a long dev_t
581 * to a dev32_t in ILP32.
584 zfs_cmpldev(uint64_t dev)
586 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
590 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
591 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
593 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
594 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
596 mutex_enter(&zp->z_lock);
598 ASSERT(zp->z_sa_hdl == NULL);
599 ASSERT(zp->z_acl_cached == NULL);
600 if (sa_hdl == NULL) {
601 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
602 SA_HDL_SHARED, &zp->z_sa_hdl));
604 zp->z_sa_hdl = sa_hdl;
605 sa_set_userp(sa_hdl, zp);
608 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
611 * Slap on VROOT if we are the root znode
613 if (zp->z_id == zfsvfs->z_root)
614 ZTOV(zp)->v_flag |= VROOT;
616 mutex_exit(&zp->z_lock);
621 zfs_znode_dmu_fini(znode_t *zp)
623 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
625 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
627 sa_handle_destroy(zp->z_sa_hdl);
632 zfs_vnode_forget(vnode_t *vp)
638 vp->v_iflag |= VI_DOOMED;
644 * Construct a new znode/vnode and intialize.
646 * This does not do a call to dmu_set_user() that is
647 * up to the caller to do, in case you don't want to
651 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
652 dmu_object_type_t obj_type, sa_handle_t *hdl)
658 sa_bulk_attr_t bulk[9];
661 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
662 zfs_znode_cache_constructor(zp, zfsvfs->z_parent->z_vfs, 0);
664 ASSERT(zp->z_dirlocks == NULL);
665 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
669 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
670 * the zfs_znode_move() callback.
674 zp->z_atime_dirty = 0;
676 zp->z_id = db->db_object;
678 zp->z_seq = 0x7A4653;
683 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
685 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
686 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
687 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
689 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
691 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
693 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
694 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
696 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
698 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
701 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
703 sa_handle_destroy(zp->z_sa_hdl);
704 zfs_vnode_forget(vp);
706 kmem_cache_free(znode_cache, zp);
712 vp->v_type = IFTOVT((mode_t)mode);
714 switch (vp->v_type) {
716 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
723 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
724 &rdev, sizeof (rdev)) == 0);
726 vp->v_rdev = zfs_cmpldev(rdev);
735 vp->v_op = &zfs_fifoops;
738 if (parent == zfsvfs->z_shares_dir) {
739 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
740 vp->v_op = &zfs_shareops;
745 vn_setops(vp, zfs_symvnodeops);
748 vn_setops(vp, zfs_evnodeops);
752 if (vp->v_type != VFIFO)
755 mutex_enter(&zfsvfs->z_znodes_lock);
756 list_insert_tail(&zfsvfs->z_all_znodes, zp);
759 * Everything else must be valid before assigning z_zfsvfs makes the
760 * znode eligible for zfs_znode_move().
762 zp->z_zfsvfs = zfsvfs;
763 mutex_exit(&zfsvfs->z_znodes_lock);
765 VFS_HOLD(zfsvfs->z_vfs);
769 static uint64_t empty_xattr;
770 static uint64_t pad[4];
771 static zfs_acl_phys_t acl_phys;
773 * Create a new DMU object to hold a zfs znode.
775 * IN: dzp - parent directory for new znode
776 * vap - file attributes for new znode
777 * tx - dmu transaction id for zap operations
778 * cr - credentials of caller
780 * IS_ROOT_NODE - new object will be root
781 * IS_XATTR - new object is an attribute
782 * bonuslen - length of bonus buffer
783 * setaclp - File/Dir initial ACL
784 * fuidp - Tracks fuid allocation.
786 * OUT: zpp - allocated znode
790 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
791 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
793 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
794 uint64_t mode, size, links, parent, pflags;
795 uint64_t dzp_pflags = 0;
797 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
804 dmu_object_type_t obj_type;
805 sa_bulk_attr_t sa_attrs[ZPL_END];
807 zfs_acl_locator_cb_t locate = { 0 };
809 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
811 if (zfsvfs->z_replay) {
812 obj = vap->va_nodeid;
813 now = vap->va_ctime; /* see zfs_replay_create() */
814 gen = vap->va_nblocks; /* ditto */
818 gen = dmu_tx_get_txg(tx);
821 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
822 bonuslen = (obj_type == DMU_OT_SA) ?
823 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
826 * Create a new DMU object.
829 * There's currently no mechanism for pre-reading the blocks that will
830 * be needed to allocate a new object, so we accept the small chance
831 * that there will be an i/o error and we will fail one of the
834 if (vap->va_type == VDIR) {
835 if (zfsvfs->z_replay) {
836 err = zap_create_claim_norm(zfsvfs->z_os, obj,
837 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
838 obj_type, bonuslen, tx);
839 ASSERT3U(err, ==, 0);
841 obj = zap_create_norm(zfsvfs->z_os,
842 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
843 obj_type, bonuslen, tx);
846 if (zfsvfs->z_replay) {
847 err = dmu_object_claim(zfsvfs->z_os, obj,
848 DMU_OT_PLAIN_FILE_CONTENTS, 0,
849 obj_type, bonuslen, tx);
850 ASSERT3U(err, ==, 0);
852 obj = dmu_object_alloc(zfsvfs->z_os,
853 DMU_OT_PLAIN_FILE_CONTENTS, 0,
854 obj_type, bonuslen, tx);
858 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
859 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
862 * If this is the root, fix up the half-initialized parent pointer
863 * to reference the just-allocated physical data area.
865 if (flag & IS_ROOT_NODE) {
868 dzp_pflags = dzp->z_pflags;
872 * If parent is an xattr, so am I.
874 if (dzp_pflags & ZFS_XATTR) {
878 if (zfsvfs->z_use_fuids)
879 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
883 if (vap->va_type == VDIR) {
884 size = 2; /* contents ("." and "..") */
885 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
890 if (vap->va_type == VBLK || vap->va_type == VCHR) {
891 rdev = zfs_expldev(vap->va_rdev);
895 mode = acl_ids->z_mode;
900 * No execs denied will be deterimed when zfs_mode_compute() is called.
902 pflags |= acl_ids->z_aclp->z_hints &
903 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
904 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
906 ZFS_TIME_ENCODE(&now, crtime);
907 ZFS_TIME_ENCODE(&now, ctime);
909 if (vap->va_mask & AT_ATIME) {
910 ZFS_TIME_ENCODE(&vap->va_atime, atime);
912 ZFS_TIME_ENCODE(&now, atime);
915 if (vap->va_mask & AT_MTIME) {
916 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
918 ZFS_TIME_ENCODE(&now, mtime);
921 /* Now add in all of the "SA" attributes */
922 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
926 * Setup the array of attributes to be replaced/set on the new file
928 * order for DMU_OT_ZNODE is critical since it needs to be constructed
929 * in the old znode_phys_t format. Don't change this ordering
932 if (obj_type == DMU_OT_ZNODE) {
933 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
935 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
937 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
939 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
941 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
943 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
945 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
947 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
950 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
952 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
954 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
956 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
957 &acl_ids->z_fuid, 8);
958 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
959 &acl_ids->z_fgid, 8);
960 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
962 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
964 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
966 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
968 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
970 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
974 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
976 if (obj_type == DMU_OT_ZNODE) {
977 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
980 if (obj_type == DMU_OT_ZNODE ||
981 (vap->va_type == VBLK || vap->va_type == VCHR)) {
982 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
986 if (obj_type == DMU_OT_ZNODE) {
987 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
989 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
990 &acl_ids->z_fuid, 8);
991 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
992 &acl_ids->z_fgid, 8);
993 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
994 sizeof (uint64_t) * 4);
995 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
996 &acl_phys, sizeof (zfs_acl_phys_t));
997 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
998 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
999 &acl_ids->z_aclp->z_acl_count, 8);
1000 locate.cb_aclp = acl_ids->z_aclp;
1001 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
1002 zfs_acl_data_locator, &locate,
1003 acl_ids->z_aclp->z_acl_bytes);
1004 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
1005 acl_ids->z_fuid, acl_ids->z_fgid);
1008 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
1010 if (!(flag & IS_ROOT_NODE)) {
1011 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
1012 ASSERT(*zpp != NULL);
1015 * If we are creating the root node, the "parent" we
1016 * passed in is the znode for the root.
1020 (*zpp)->z_sa_hdl = sa_hdl;
1023 (*zpp)->z_pflags = pflags;
1024 (*zpp)->z_mode = mode;
1026 if (vap->va_mask & AT_XVATTR)
1027 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1029 if (obj_type == DMU_OT_ZNODE ||
1030 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1031 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
1032 ASSERT3P(err, ==, 0);
1034 if (!(flag & IS_ROOT_NODE)) {
1038 vp->v_vflag |= VV_FORCEINSMQ;
1039 err = insmntque(vp, zfsvfs->z_vfs);
1040 vp->v_vflag &= ~VV_FORCEINSMQ;
1041 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1043 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1047 * zfs_xvattr_set only updates the in-core attributes
1048 * it is assumed the caller will be doing an sa_bulk_update
1049 * to push the changes out
1052 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1056 xoap = xva_getxoptattr(xvap);
1059 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1061 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1062 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1063 ×, sizeof (times), tx);
1064 XVA_SET_RTN(xvap, XAT_CREATETIME);
1066 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1067 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1069 XVA_SET_RTN(xvap, XAT_READONLY);
1071 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1072 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1074 XVA_SET_RTN(xvap, XAT_HIDDEN);
1076 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1077 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1079 XVA_SET_RTN(xvap, XAT_SYSTEM);
1081 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1082 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1084 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1086 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1087 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1089 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1091 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1092 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1094 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1096 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1097 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1099 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1101 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1102 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1104 XVA_SET_RTN(xvap, XAT_NODUMP);
1106 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1107 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1109 XVA_SET_RTN(xvap, XAT_OPAQUE);
1111 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1112 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1113 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1114 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1116 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1117 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1119 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1121 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1122 zfs_sa_set_scanstamp(zp, xvap, tx);
1123 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1125 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1126 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1128 XVA_SET_RTN(xvap, XAT_REPARSE);
1130 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1131 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1133 XVA_SET_RTN(xvap, XAT_OFFLINE);
1135 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1136 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1138 XVA_SET_RTN(xvap, XAT_SPARSE);
1143 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1145 dmu_object_info_t doi;
1155 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1157 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1159 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1163 dmu_object_info_from_db(db, &doi);
1164 if (doi.doi_bonus_type != DMU_OT_SA &&
1165 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1166 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1167 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1168 sa_buf_rele(db, NULL);
1169 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1173 hdl = dmu_buf_get_user(db);
1175 zp = sa_get_userdata(hdl);
1179 * Since "SA" does immediate eviction we
1180 * should never find a sa handle that doesn't
1181 * know about the znode.
1184 ASSERT3P(zp, !=, NULL);
1186 mutex_enter(&zp->z_lock);
1187 ASSERT3U(zp->z_id, ==, obj_num);
1188 if (zp->z_unlinked) {
1199 if ((vp->v_iflag & VI_DOOMED) != 0) {
1202 * Don't VN_RELE() vnode here, because
1203 * it can call vn_lock() which creates
1204 * LOR between vnode lock and znode
1205 * lock. We will VN_RELE() the vnode
1206 * after droping znode lock.
1212 ZFS_LOG(1, "dying znode detected (zp=%p)", zp);
1216 * znode is dying so we can't reuse it, we must
1217 * wait until destruction is completed.
1219 sa_buf_rele(db, NULL);
1220 mutex_exit(&zp->z_lock);
1221 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1224 tsleep(zp, 0, "zcollide", 1);
1230 sa_buf_rele(db, NULL);
1231 mutex_exit(&zp->z_lock);
1232 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1237 * Not found create new znode/vnode
1238 * but only if file exists.
1240 * There is a small window where zfs_vget() could
1241 * find this object while a file create is still in
1242 * progress. This is checked for in zfs_znode_alloc()
1244 * if zfs_znode_alloc() fails it will drop the hold on the
1247 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1248 doi.doi_bonus_type, NULL);
1255 vnode_t *vp = ZTOV(zp);
1257 err = insmntque(vp, zfsvfs->z_vfs);
1262 zfs_znode_dmu_fini(zp);
1267 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1272 zfs_rezget(znode_t *zp)
1274 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1275 dmu_object_info_t doi;
1278 uint64_t obj_num = zp->z_id;
1279 uint64_t mode, size;
1280 sa_bulk_attr_t bulk[8];
1285 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1287 mutex_enter(&zp->z_acl_lock);
1288 if (zp->z_acl_cached) {
1289 zfs_acl_free(zp->z_acl_cached);
1290 zp->z_acl_cached = NULL;
1293 mutex_exit(&zp->z_acl_lock);
1294 ASSERT(zp->z_sa_hdl == NULL);
1295 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1297 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1301 dmu_object_info_from_db(db, &doi);
1302 if (doi.doi_bonus_type != DMU_OT_SA &&
1303 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1304 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1305 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1306 sa_buf_rele(db, NULL);
1307 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1311 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1314 /* reload cached values */
1315 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1316 &gen, sizeof (gen));
1317 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1318 &zp->z_size, sizeof (zp->z_size));
1319 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1320 &zp->z_links, sizeof (zp->z_links));
1321 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1322 &zp->z_pflags, sizeof (zp->z_pflags));
1323 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1324 &zp->z_atime, sizeof (zp->z_atime));
1325 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1326 &zp->z_uid, sizeof (zp->z_uid));
1327 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1328 &zp->z_gid, sizeof (zp->z_gid));
1329 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1330 &mode, sizeof (mode));
1332 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1333 zfs_znode_dmu_fini(zp);
1334 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1340 if (gen != zp->z_gen) {
1341 zfs_znode_dmu_fini(zp);
1342 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1347 * XXXPJD: Not sure how is that possible, but under heavy
1348 * zfs recv -F load it happens that z_gen is the same, but
1349 * vnode type is different than znode type. This would mean
1350 * that for example regular file was replaced with directory
1351 * which has the same object number.
1355 vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1356 zfs_znode_dmu_fini(zp);
1357 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1361 zp->z_unlinked = (zp->z_links == 0);
1362 zp->z_blksz = doi.doi_data_block_size;
1364 vn_pages_remove(vp, 0, 0);
1365 if (zp->z_size != size)
1366 vnode_pager_setsize(vp, zp->z_size);
1369 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1375 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1377 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1378 objset_t *os = zfsvfs->z_os;
1379 uint64_t obj = zp->z_id;
1380 uint64_t acl_obj = zfs_external_acl(zp);
1382 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1384 VERIFY(!zp->z_is_sa);
1385 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1387 VERIFY(0 == dmu_object_free(os, obj, tx));
1388 zfs_znode_dmu_fini(zp);
1389 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1394 zfs_zinactive(znode_t *zp)
1396 vnode_t *vp = ZTOV(zp);
1397 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1398 uint64_t z_id = zp->z_id;
1401 ASSERT(zp->z_sa_hdl);
1404 * Don't allow a zfs_zget() while were trying to release this znode
1406 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1408 mutex_enter(&zp->z_lock);
1410 if (vp->v_count > 0) {
1412 * If the hold count is greater than zero, somebody has
1413 * obtained a new reference on this znode while we were
1414 * processing it here, so we are done.
1417 mutex_exit(&zp->z_lock);
1418 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1424 * If this was the last reference to a file with no links,
1425 * remove the file from the file system.
1427 if (zp->z_unlinked) {
1428 mutex_exit(&zp->z_lock);
1429 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1430 ASSERT(vp->v_count == 0);
1431 vrecycle(vp, curthread);
1432 vfslocked = VFS_LOCK_GIANT(zfsvfs->z_vfs);
1434 VFS_UNLOCK_GIANT(vfslocked);
1438 mutex_exit(&zp->z_lock);
1439 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1443 zfs_znode_free(znode_t *zp)
1445 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1447 ASSERT(ZTOV(zp) == NULL);
1448 ASSERT(zp->z_sa_hdl == NULL);
1449 mutex_enter(&zfsvfs->z_znodes_lock);
1450 POINTER_INVALIDATE(&zp->z_zfsvfs);
1451 list_remove(&zfsvfs->z_all_znodes, zp);
1452 mutex_exit(&zfsvfs->z_znodes_lock);
1454 if (zp->z_acl_cached) {
1455 zfs_acl_free(zp->z_acl_cached);
1456 zp->z_acl_cached = NULL;
1459 kmem_cache_free(znode_cache, zp);
1461 VFS_RELE(zfsvfs->z_vfs);
1465 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1466 uint64_t ctime[2], boolean_t have_tx)
1472 if (have_tx) { /* will sa_bulk_update happen really soon? */
1473 zp->z_atime_dirty = 0;
1476 zp->z_atime_dirty = 1;
1479 if (flag & AT_ATIME) {
1480 ZFS_TIME_ENCODE(&now, zp->z_atime);
1483 if (flag & AT_MTIME) {
1484 ZFS_TIME_ENCODE(&now, mtime);
1485 if (zp->z_zfsvfs->z_use_fuids) {
1486 zp->z_pflags |= (ZFS_ARCHIVE |
1491 if (flag & AT_CTIME) {
1492 ZFS_TIME_ENCODE(&now, ctime);
1493 if (zp->z_zfsvfs->z_use_fuids)
1494 zp->z_pflags |= ZFS_ARCHIVE;
1499 * Grow the block size for a file.
1501 * IN: zp - znode of file to free data in.
1502 * size - requested block size
1503 * tx - open transaction.
1505 * NOTE: this function assumes that the znode is write locked.
1508 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1513 if (size <= zp->z_blksz)
1516 * If the file size is already greater than the current blocksize,
1517 * we will not grow. If there is more than one block in a file,
1518 * the blocksize cannot change.
1520 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1523 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1526 if (error == ENOTSUP)
1528 ASSERT3U(error, ==, 0);
1530 /* What blocksize did we actually get? */
1531 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1536 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1537 * be calling back into the fs for a putpage(). E.g.: when truncating
1538 * a file, the pages being "thrown away* don't need to be written out.
1542 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1543 int flags, cred_t *cr)
1551 * Increase the file length
1553 * IN: zp - znode of file to free data in.
1554 * end - new end-of-file
1556 * RETURN: 0 if success
1557 * error code if failure
1560 zfs_extend(znode_t *zp, uint64_t end)
1562 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1569 * We will change zp_size, lock the whole file.
1571 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1574 * Nothing to do if file already at desired length.
1576 if (end <= zp->z_size) {
1577 zfs_range_unlock(rl);
1581 tx = dmu_tx_create(zfsvfs->z_os);
1582 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1583 zfs_sa_upgrade_txholds(tx, zp);
1584 if (end > zp->z_blksz &&
1585 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1587 * We are growing the file past the current block size.
1589 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1590 ASSERT(!ISP2(zp->z_blksz));
1591 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1593 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1595 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1600 error = dmu_tx_assign(tx, TXG_NOWAIT);
1602 if (error == ERESTART) {
1608 zfs_range_unlock(rl);
1613 zfs_grow_blocksize(zp, newblksz, tx);
1617 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1618 &zp->z_size, sizeof (zp->z_size), tx));
1620 vnode_pager_setsize(ZTOV(zp), end);
1622 zfs_range_unlock(rl);
1630 * Free space in a file.
1632 * IN: zp - znode of file to free data in.
1633 * off - start of section to free.
1634 * len - length of section to free.
1636 * RETURN: 0 if success
1637 * error code if failure
1640 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1642 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1647 * Lock the range being freed.
1649 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1652 * Nothing to do if file already at desired length.
1654 if (off >= zp->z_size) {
1655 zfs_range_unlock(rl);
1659 if (off + len > zp->z_size)
1660 len = zp->z_size - off;
1662 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1666 * In FreeBSD we cannot free block in the middle of a file,
1667 * but only at the end of a file, so this code path should
1670 vnode_pager_setsize(ZTOV(zp), off);
1673 zfs_range_unlock(rl);
1681 * IN: zp - znode of file to free data in.
1682 * end - new end-of-file.
1684 * RETURN: 0 if success
1685 * error code if failure
1688 zfs_trunc(znode_t *zp, uint64_t end)
1690 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1691 vnode_t *vp = ZTOV(zp);
1695 sa_bulk_attr_t bulk[2];
1699 * We will change zp_size, lock the whole file.
1701 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1704 * Nothing to do if file already at desired length.
1706 if (end >= zp->z_size) {
1707 zfs_range_unlock(rl);
1711 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1713 zfs_range_unlock(rl);
1717 tx = dmu_tx_create(zfsvfs->z_os);
1718 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1719 zfs_sa_upgrade_txholds(tx, zp);
1720 error = dmu_tx_assign(tx, TXG_NOWAIT);
1722 if (error == ERESTART) {
1728 zfs_range_unlock(rl);
1733 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1734 NULL, &zp->z_size, sizeof (zp->z_size));
1737 zp->z_pflags &= ~ZFS_SPARSE;
1738 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1739 NULL, &zp->z_pflags, 8);
1741 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1746 * Clear any mapped pages in the truncated region. This has to
1747 * happen outside of the transaction to avoid the possibility of
1748 * a deadlock with someone trying to push a page that we are
1749 * about to invalidate.
1751 vnode_pager_setsize(vp, end);
1753 zfs_range_unlock(rl);
1759 * Free space in a file
1761 * IN: zp - znode of file to free data in.
1762 * off - start of range
1763 * len - end of range (0 => EOF)
1764 * flag - current file open mode flags.
1765 * log - TRUE if this action should be logged
1767 * RETURN: 0 if success
1768 * error code if failure
1771 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1773 vnode_t *vp = ZTOV(zp);
1775 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1776 zilog_t *zilog = zfsvfs->z_log;
1778 uint64_t mtime[2], ctime[2];
1779 sa_bulk_attr_t bulk[3];
1783 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1784 sizeof (mode))) != 0)
1787 if (off > zp->z_size) {
1788 error = zfs_extend(zp, off+len);
1789 if (error == 0 && log)
1796 * Check for any locks in the region to be freed.
1799 if (MANDLOCK(vp, (mode_t)mode)) {
1800 uint64_t length = (len ? len : zp->z_size - off);
1801 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1806 error = zfs_trunc(zp, off);
1808 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1809 off + len > zp->z_size)
1810 error = zfs_extend(zp, off+len);
1815 tx = dmu_tx_create(zfsvfs->z_os);
1816 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1817 zfs_sa_upgrade_txholds(tx, zp);
1818 error = dmu_tx_assign(tx, TXG_NOWAIT);
1820 if (error == ERESTART) {
1829 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1830 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1831 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1832 NULL, &zp->z_pflags, 8);
1833 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1834 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1837 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1844 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1847 uint64_t moid, obj, sa_obj, version;
1848 uint64_t sense = ZFS_CASE_SENSITIVE;
1853 znode_t *rootzp = NULL;
1857 zfs_acl_ids_t acl_ids;
1860 * First attempt to create master node.
1863 * In an empty objset, there are no blocks to read and thus
1864 * there can be no i/o errors (which we assert below).
1866 moid = MASTER_NODE_OBJ;
1867 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1868 DMU_OT_NONE, 0, tx);
1872 * Set starting attributes.
1874 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1876 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1877 /* For the moment we expect all zpl props to be uint64_ts */
1881 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1882 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1883 name = nvpair_name(elem);
1884 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1888 error = zap_update(os, moid, name, 8, 1, &val, tx);
1891 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1893 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1896 ASSERT(version != 0);
1897 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1900 * Create zap object used for SA attribute registration
1903 if (version >= ZPL_VERSION_SA) {
1904 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1905 DMU_OT_NONE, 0, tx);
1906 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1912 * Create a delete queue.
1914 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1916 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1920 * Create root znode. Create minimal znode/vnode/zfsvfs
1921 * to allow zfs_mknode to work.
1924 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1925 vattr.va_type = VDIR;
1926 vattr.va_mode = S_IFDIR|0755;
1927 vattr.va_uid = crgetuid(cr);
1928 vattr.va_gid = crgetgid(cr);
1930 bzero(&zfsvfs, sizeof (zfsvfs_t));
1932 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1933 zfs_znode_cache_constructor(rootzp, NULL, 0);
1934 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1935 rootzp->z_moved = 0;
1936 rootzp->z_unlinked = 0;
1937 rootzp->z_atime_dirty = 0;
1938 rootzp->z_is_sa = USE_SA(version, os);
1940 vnode.v_type = VDIR;
1941 vnode.v_data = rootzp;
1942 rootzp->z_vnode = &vnode;
1945 zfsvfs.z_parent = &zfsvfs;
1946 zfsvfs.z_version = version;
1947 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1948 zfsvfs.z_use_sa = USE_SA(version, os);
1949 zfsvfs.z_norm = norm;
1951 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1952 &zfsvfs.z_attr_table);
1957 * Fold case on file systems that are always or sometimes case
1960 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1961 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1963 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1964 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1965 offsetof(znode_t, z_link_node));
1967 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1968 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1970 rootzp->z_zfsvfs = &zfsvfs;
1971 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1972 cr, NULL, &acl_ids));
1973 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1974 ASSERT3P(zp, ==, rootzp);
1975 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1977 zfs_acl_ids_free(&acl_ids);
1978 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1980 sa_handle_destroy(rootzp->z_sa_hdl);
1981 rootzp->z_vnode = NULL;
1982 kmem_cache_free(znode_cache, rootzp);
1985 * Create shares directory
1988 error = zfs_create_share_dir(&zfsvfs, tx);
1992 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1993 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1996 #endif /* _KERNEL */
1999 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
2001 uint64_t sa_obj = 0;
2004 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
2005 if (error != 0 && error != ENOENT)
2008 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
2013 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
2014 dmu_buf_t **db, void *tag)
2016 dmu_object_info_t doi;
2019 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
2022 dmu_object_info_from_db(*db, &doi);
2023 if ((doi.doi_bonus_type != DMU_OT_SA &&
2024 doi.doi_bonus_type != DMU_OT_ZNODE) ||
2025 doi.doi_bonus_type == DMU_OT_ZNODE &&
2026 doi.doi_bonus_size < sizeof (znode_phys_t)) {
2027 sa_buf_rele(*db, tag);
2031 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
2033 sa_buf_rele(*db, tag);
2041 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
2043 sa_handle_destroy(hdl);
2044 sa_buf_rele(db, tag);
2048 * Given an object number, return its parent object number and whether
2049 * or not the object is an extended attribute directory.
2052 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
2058 sa_bulk_attr_t bulk[3];
2062 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2063 &parent, sizeof (parent));
2064 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2065 &pflags, sizeof (pflags));
2066 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2067 &mode, sizeof (mode));
2069 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2073 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2079 * Given an object number, return some zpl level statistics
2082 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2085 sa_bulk_attr_t bulk[4];
2088 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2089 &sb->zs_mode, sizeof (sb->zs_mode));
2090 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2091 &sb->zs_gen, sizeof (sb->zs_gen));
2092 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2093 &sb->zs_links, sizeof (sb->zs_links));
2094 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2095 &sb->zs_ctime, sizeof (sb->zs_ctime));
2097 return (sa_bulk_lookup(hdl, bulk, count));
2101 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2102 sa_attr_type_t *sa_table, char *buf, int len)
2104 sa_handle_t *sa_hdl;
2105 sa_handle_t *prevhdl = NULL;
2106 dmu_buf_t *prevdb = NULL;
2107 dmu_buf_t *sa_db = NULL;
2108 char *path = buf + len - 1;
2116 char component[MAXNAMELEN + 2];
2121 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2123 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
2124 &is_xattrdir)) != 0)
2135 (void) sprintf(component + 1, "<xattrdir>");
2137 error = zap_value_search(osp, pobj, obj,
2138 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2143 complen = strlen(component);
2145 ASSERT(path >= buf);
2146 bcopy(component, path, complen);
2149 if (sa_hdl != hdl) {
2153 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2161 if (sa_hdl != NULL && sa_hdl != hdl) {
2162 ASSERT(sa_db != NULL);
2163 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2167 (void) memmove(buf, path, buf + len - path);
2173 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2175 sa_attr_type_t *sa_table;
2180 error = zfs_sa_setup(osp, &sa_table);
2184 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2188 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2190 zfs_release_sa_handle(hdl, db, FTAG);
2195 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2198 char *path = buf + len - 1;
2199 sa_attr_type_t *sa_table;
2206 error = zfs_sa_setup(osp, &sa_table);
2210 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2214 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2216 zfs_release_sa_handle(hdl, db, FTAG);
2220 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2222 zfs_release_sa_handle(hdl, db, FTAG);