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.
23 * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
24 * Copyright (c) 2014 Integros [integros.com]
27 /* Portions Copyright 2007 Jeremy Teo */
28 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
31 #include <sys/types.h>
32 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/sysmacros.h>
36 #include <sys/resource.h>
37 #include <sys/mntent.h>
38 #include <sys/u8_textprep.h>
39 #include <sys/dsl_dataset.h>
41 #include <sys/vnode.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/atomic.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/zfs_ioctl.h>
50 #include <sys/zfs_rlock.h>
51 #include <sys/zfs_fuid.h>
52 #include <sys/dnode.h>
53 #include <sys/fs/zfs.h>
54 #include <sys/kidmap.h>
58 #include <sys/dmu_objset.h>
59 #include <sys/refcount.h>
62 #include <sys/zfs_znode.h>
64 #include <sys/zfs_sa.h>
65 #include <sys/zfs_stat.h>
66 #include <sys/refcount.h>
69 #include "zfs_comutil.h"
71 /* Used by fstat(1). */
72 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
73 SYSCTL_NULL_INT_PTR, sizeof(znode_t), "sizeof(znode_t)");
76 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
77 * turned on when DEBUG is also defined.
84 #define ZNODE_STAT_ADD(stat) ((stat)++)
86 #define ZNODE_STAT_ADD(stat) /* nothing */
87 #endif /* ZNODE_STATS */
90 * Functions needed for userland (ie: libzpool) are not put under
91 * #ifdef_KERNEL; the rest of the functions have dependencies
92 * (such as VFS logic) that will not compile easily in userland.
96 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
97 * be freed before it can be safely accessed.
99 krwlock_t zfsvfs_lock;
101 static kmem_cache_t *znode_cache = NULL;
105 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
108 * We should never drop all dbuf refs without first clearing
109 * the eviction callback.
111 panic("evicting znode %p\n", user_ptr);
114 extern struct vop_vector zfs_vnodeops;
115 extern struct vop_vector zfs_fifoops;
116 extern struct vop_vector zfs_shareops;
119 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
123 POINTER_INVALIDATE(&zp->z_zfsvfs);
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;
145 zfs_znode_cache_destructor(void *buf, void *arg)
149 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
150 ASSERT(ZTOV(zp) == NULL);
152 ASSERT(!list_link_active(&zp->z_link_node));
153 mutex_destroy(&zp->z_lock);
154 rw_destroy(&zp->z_parent_lock);
155 rw_destroy(&zp->z_name_lock);
156 mutex_destroy(&zp->z_acl_lock);
157 avl_destroy(&zp->z_range_avl);
158 mutex_destroy(&zp->z_range_lock);
160 ASSERT(zp->z_dirlocks == NULL);
161 ASSERT(zp->z_acl_cached == NULL);
166 uint64_t zms_zfsvfs_invalid;
167 uint64_t zms_zfsvfs_recheck1;
168 uint64_t zms_zfsvfs_unmounted;
169 uint64_t zms_zfsvfs_recheck2;
170 uint64_t zms_obj_held;
171 uint64_t zms_vnode_locked;
172 uint64_t zms_not_only_dnlc;
174 #endif /* ZNODE_STATS */
178 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
183 nzp->z_zfsvfs = ozp->z_zfsvfs;
187 nzp->z_vnode = ozp->z_vnode;
188 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
189 ZTOV(ozp)->v_data = ozp;
190 ZTOV(nzp)->v_data = nzp;
192 nzp->z_id = ozp->z_id;
193 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
194 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
195 nzp->z_unlinked = ozp->z_unlinked;
196 nzp->z_atime_dirty = ozp->z_atime_dirty;
197 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
198 nzp->z_blksz = ozp->z_blksz;
199 nzp->z_seq = ozp->z_seq;
200 nzp->z_mapcnt = ozp->z_mapcnt;
201 nzp->z_gen = ozp->z_gen;
202 nzp->z_sync_cnt = ozp->z_sync_cnt;
203 nzp->z_is_sa = ozp->z_is_sa;
204 nzp->z_sa_hdl = ozp->z_sa_hdl;
205 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
206 nzp->z_links = ozp->z_links;
207 nzp->z_size = ozp->z_size;
208 nzp->z_pflags = ozp->z_pflags;
209 nzp->z_uid = ozp->z_uid;
210 nzp->z_gid = ozp->z_gid;
211 nzp->z_mode = ozp->z_mode;
214 * Since this is just an idle znode and kmem is already dealing with
215 * memory pressure, release any cached ACL.
217 if (ozp->z_acl_cached) {
218 zfs_acl_free(ozp->z_acl_cached);
219 ozp->z_acl_cached = NULL;
222 sa_set_userp(nzp->z_sa_hdl, nzp);
225 * Invalidate the original znode by clearing fields that provide a
226 * pointer back to the znode. Set the low bit of the vfs pointer to
227 * ensure that zfs_znode_move() recognizes the znode as invalid in any
228 * subsequent callback.
230 ozp->z_sa_hdl = NULL;
231 POINTER_INVALIDATE(&ozp->z_zfsvfs);
237 ozp->z_moved = (uint8_t)-1;
242 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
244 znode_t *ozp = buf, *nzp = newbuf;
249 * The znode is on the file system's list of known znodes if the vfs
250 * pointer is valid. We set the low bit of the vfs pointer when freeing
251 * the znode to invalidate it, and the memory patterns written by kmem
252 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
253 * created znode sets the vfs pointer last of all to indicate that the
254 * znode is known and in a valid state to be moved by this function.
256 zfsvfs = ozp->z_zfsvfs;
257 if (!POINTER_IS_VALID(zfsvfs)) {
258 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
259 return (KMEM_CBRC_DONT_KNOW);
263 * Close a small window in which it's possible that the filesystem could
264 * be unmounted and freed, and zfsvfs, though valid in the previous
265 * statement, could point to unrelated memory by the time we try to
266 * prevent the filesystem from being unmounted.
268 rw_enter(&zfsvfs_lock, RW_WRITER);
269 if (zfsvfs != ozp->z_zfsvfs) {
270 rw_exit(&zfsvfs_lock);
271 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
272 return (KMEM_CBRC_DONT_KNOW);
276 * If the znode is still valid, then so is the file system. We know that
277 * no valid file system can be freed while we hold zfsvfs_lock, so we
278 * can safely ensure that the filesystem is not and will not be
279 * unmounted. The next statement is equivalent to ZFS_ENTER().
281 rrm_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
282 if (zfsvfs->z_unmounted) {
284 rw_exit(&zfsvfs_lock);
285 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
286 return (KMEM_CBRC_DONT_KNOW);
288 rw_exit(&zfsvfs_lock);
290 mutex_enter(&zfsvfs->z_znodes_lock);
292 * Recheck the vfs pointer in case the znode was removed just before
293 * acquiring the lock.
295 if (zfsvfs != ozp->z_zfsvfs) {
296 mutex_exit(&zfsvfs->z_znodes_lock);
298 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
299 return (KMEM_CBRC_DONT_KNOW);
303 * At this point we know that as long as we hold z_znodes_lock, the
304 * znode cannot be freed and fields within the znode can be safely
305 * accessed. Now, prevent a race with zfs_zget().
307 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
308 mutex_exit(&zfsvfs->z_znodes_lock);
310 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
311 return (KMEM_CBRC_LATER);
315 if (mutex_tryenter(&vp->v_lock) == 0) {
316 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
317 mutex_exit(&zfsvfs->z_znodes_lock);
319 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
320 return (KMEM_CBRC_LATER);
323 /* Only move znodes that are referenced _only_ by the DNLC. */
324 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
325 mutex_exit(&vp->v_lock);
326 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
327 mutex_exit(&zfsvfs->z_znodes_lock);
329 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
330 return (KMEM_CBRC_LATER);
334 * The znode is known and in a valid state to move. We're holding the
335 * locks needed to execute the critical section.
337 zfs_znode_move_impl(ozp, nzp);
338 mutex_exit(&vp->v_lock);
339 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
341 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
342 mutex_exit(&zfsvfs->z_znodes_lock);
345 return (KMEM_CBRC_YES);
355 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
356 ASSERT(znode_cache == NULL);
357 znode_cache = kmem_cache_create("zfs_znode_cache",
358 sizeof (znode_t), 0, zfs_znode_cache_constructor,
359 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
360 kmem_cache_set_move(znode_cache, zfs_znode_move);
368 * Cleanup vfs & vnode ops
370 zfs_remove_op_tables();
377 kmem_cache_destroy(znode_cache);
379 rw_destroy(&zfsvfs_lock);
383 struct vnodeops *zfs_dvnodeops;
384 struct vnodeops *zfs_fvnodeops;
385 struct vnodeops *zfs_symvnodeops;
386 struct vnodeops *zfs_xdvnodeops;
387 struct vnodeops *zfs_evnodeops;
388 struct vnodeops *zfs_sharevnodeops;
391 zfs_remove_op_tables()
397 (void) vfs_freevfsops_by_type(zfsfstype);
404 vn_freevnodeops(zfs_dvnodeops);
406 vn_freevnodeops(zfs_fvnodeops);
408 vn_freevnodeops(zfs_symvnodeops);
410 vn_freevnodeops(zfs_xdvnodeops);
412 vn_freevnodeops(zfs_evnodeops);
413 if (zfs_sharevnodeops)
414 vn_freevnodeops(zfs_sharevnodeops);
416 zfs_dvnodeops = NULL;
417 zfs_fvnodeops = NULL;
418 zfs_symvnodeops = NULL;
419 zfs_xdvnodeops = NULL;
420 zfs_evnodeops = NULL;
421 zfs_sharevnodeops = NULL;
424 extern const fs_operation_def_t zfs_dvnodeops_template[];
425 extern const fs_operation_def_t zfs_fvnodeops_template[];
426 extern const fs_operation_def_t zfs_xdvnodeops_template[];
427 extern const fs_operation_def_t zfs_symvnodeops_template[];
428 extern const fs_operation_def_t zfs_evnodeops_template[];
429 extern const fs_operation_def_t zfs_sharevnodeops_template[];
432 zfs_create_op_tables()
437 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
438 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
439 * In this case we just return as the ops vectors are already set up.
444 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
449 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
454 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
459 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
464 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
469 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
477 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
479 zfs_acl_ids_t acl_ids;
485 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
486 vattr.va_type = VDIR;
487 vattr.va_mode = S_IFDIR|0555;
488 vattr.va_uid = crgetuid(kcred);
489 vattr.va_gid = crgetgid(kcred);
491 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
492 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
493 sharezp->z_moved = 0;
494 sharezp->z_unlinked = 0;
495 sharezp->z_atime_dirty = 0;
496 sharezp->z_zfsvfs = zfsvfs;
497 sharezp->z_is_sa = zfsvfs->z_use_sa;
499 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
500 kcred, NULL, &acl_ids));
501 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
502 ASSERT3P(zp, ==, sharezp);
503 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
504 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
505 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
506 zfsvfs->z_shares_dir = sharezp->z_id;
508 zfs_acl_ids_free(&acl_ids);
509 sa_handle_destroy(sharezp->z_sa_hdl);
510 kmem_cache_free(znode_cache, sharezp);
516 * define a couple of values we need available
517 * for both 64 and 32 bit environments.
520 #define NBITSMINOR64 32
523 #define MAXMAJ64 0xffffffffUL
526 #define MAXMIN64 0xffffffffUL
530 * Create special expldev for ZFS private use.
531 * Can't use standard expldev since it doesn't do
532 * what we want. The standard expldev() takes a
533 * dev32_t in LP64 and expands it to a long dev_t.
534 * We need an interface that takes a dev32_t in ILP32
535 * and expands it to a long dev_t.
538 zfs_expldev(dev_t dev)
540 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
543 * Special cmpldev for ZFS private use.
544 * Can't use standard cmpldev since it takes
545 * a long dev_t and compresses it to dev32_t in
546 * LP64. We need to do a compaction of a long dev_t
547 * to a dev32_t in ILP32.
550 zfs_cmpldev(uint64_t dev)
552 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
556 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
557 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
559 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
560 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
562 mutex_enter(&zp->z_lock);
564 ASSERT(zp->z_sa_hdl == NULL);
565 ASSERT(zp->z_acl_cached == NULL);
566 if (sa_hdl == NULL) {
567 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
568 SA_HDL_SHARED, &zp->z_sa_hdl));
570 zp->z_sa_hdl = sa_hdl;
571 sa_set_userp(sa_hdl, zp);
574 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
577 * Slap on VROOT if we are the root znode
579 if (zp->z_id == zfsvfs->z_root)
580 ZTOV(zp)->v_flag |= VROOT;
582 mutex_exit(&zp->z_lock);
587 zfs_znode_dmu_fini(znode_t *zp)
589 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
591 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
593 sa_handle_destroy(zp->z_sa_hdl);
598 zfs_vnode_forget(vnode_t *vp)
601 /* copied from insmntque_stddtr */
603 vp->v_op = &dead_vnodeops;
609 * Construct a new znode/vnode and intialize.
611 * This does not do a call to dmu_set_user() that is
612 * up to the caller to do, in case you don't want to
616 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
617 dmu_object_type_t obj_type, sa_handle_t *hdl)
623 sa_bulk_attr_t bulk[9];
627 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
629 KASSERT(curthread->td_vp_reserv > 0,
630 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
631 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
633 kmem_cache_free(znode_cache, zp);
639 ASSERT(zp->z_dirlocks == NULL);
640 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
644 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
645 * the zfs_znode_move() callback.
649 zp->z_atime_dirty = 0;
651 zp->z_id = db->db_object;
653 zp->z_seq = 0x7A4653;
658 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
660 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
661 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
662 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
664 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
666 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
669 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
671 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
673 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
676 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
678 sa_handle_destroy(zp->z_sa_hdl);
679 zfs_vnode_forget(vp);
681 kmem_cache_free(znode_cache, zp);
687 vp->v_type = IFTOVT((mode_t)mode);
689 switch (vp->v_type) {
691 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
698 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
699 &rdev, sizeof (rdev)) == 0);
701 vp->v_rdev = zfs_cmpldev(rdev);
710 vp->v_op = &zfs_fifoops;
713 if (parent == zfsvfs->z_shares_dir) {
714 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
715 vp->v_op = &zfs_shareops;
720 vn_setops(vp, zfs_symvnodeops);
723 vn_setops(vp, zfs_evnodeops);
728 mutex_enter(&zfsvfs->z_znodes_lock);
729 list_insert_tail(&zfsvfs->z_all_znodes, zp);
732 * Everything else must be valid before assigning z_zfsvfs makes the
733 * znode eligible for zfs_znode_move().
735 zp->z_zfsvfs = zfsvfs;
736 mutex_exit(&zfsvfs->z_znodes_lock);
739 * Acquire vnode lock before making it available to the world.
741 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
743 if (vp->v_type != VFIFO)
747 VFS_HOLD(zfsvfs->z_vfs);
752 static uint64_t empty_xattr;
753 static uint64_t pad[4];
754 static zfs_acl_phys_t acl_phys;
756 * Create a new DMU object to hold a zfs znode.
758 * IN: dzp - parent directory for new znode
759 * vap - file attributes for new znode
760 * tx - dmu transaction id for zap operations
761 * cr - credentials of caller
763 * IS_ROOT_NODE - new object will be root
764 * IS_XATTR - new object is an attribute
765 * bonuslen - length of bonus buffer
766 * setaclp - File/Dir initial ACL
767 * fuidp - Tracks fuid allocation.
769 * OUT: zpp - allocated znode
773 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
774 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
776 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
777 uint64_t mode, size, links, parent, pflags;
778 uint64_t dzp_pflags = 0;
780 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
787 dmu_object_type_t obj_type;
788 sa_bulk_attr_t sa_attrs[ZPL_END];
790 zfs_acl_locator_cb_t locate = { 0 };
792 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
794 if (zfsvfs->z_replay) {
795 obj = vap->va_nodeid;
796 now = vap->va_ctime; /* see zfs_replay_create() */
797 gen = vap->va_nblocks; /* ditto */
801 gen = dmu_tx_get_txg(tx);
804 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
805 bonuslen = (obj_type == DMU_OT_SA) ?
806 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
809 * Create a new DMU object.
812 * There's currently no mechanism for pre-reading the blocks that will
813 * be needed to allocate a new object, so we accept the small chance
814 * that there will be an i/o error and we will fail one of the
817 if (vap->va_type == VDIR) {
818 if (zfsvfs->z_replay) {
819 VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj,
820 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
821 obj_type, bonuslen, tx));
823 obj = zap_create_norm(zfsvfs->z_os,
824 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
825 obj_type, bonuslen, tx);
828 if (zfsvfs->z_replay) {
829 VERIFY0(dmu_object_claim(zfsvfs->z_os, obj,
830 DMU_OT_PLAIN_FILE_CONTENTS, 0,
831 obj_type, bonuslen, tx));
833 obj = dmu_object_alloc(zfsvfs->z_os,
834 DMU_OT_PLAIN_FILE_CONTENTS, 0,
835 obj_type, bonuslen, tx);
839 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
840 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
843 * If this is the root, fix up the half-initialized parent pointer
844 * to reference the just-allocated physical data area.
846 if (flag & IS_ROOT_NODE) {
849 dzp_pflags = dzp->z_pflags;
853 * If parent is an xattr, so am I.
855 if (dzp_pflags & ZFS_XATTR) {
859 if (zfsvfs->z_use_fuids)
860 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
864 if (vap->va_type == VDIR) {
865 size = 2; /* contents ("." and "..") */
866 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
871 if (vap->va_type == VBLK || vap->va_type == VCHR) {
872 rdev = zfs_expldev(vap->va_rdev);
876 mode = acl_ids->z_mode;
881 * No execs denied will be deterimed when zfs_mode_compute() is called.
883 pflags |= acl_ids->z_aclp->z_hints &
884 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
885 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
887 ZFS_TIME_ENCODE(&now, crtime);
888 ZFS_TIME_ENCODE(&now, ctime);
890 if (vap->va_mask & AT_ATIME) {
891 ZFS_TIME_ENCODE(&vap->va_atime, atime);
893 ZFS_TIME_ENCODE(&now, atime);
896 if (vap->va_mask & AT_MTIME) {
897 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
899 ZFS_TIME_ENCODE(&now, mtime);
902 /* Now add in all of the "SA" attributes */
903 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
907 * Setup the array of attributes to be replaced/set on the new file
909 * order for DMU_OT_ZNODE is critical since it needs to be constructed
910 * in the old znode_phys_t format. Don't change this ordering
913 if (obj_type == DMU_OT_ZNODE) {
914 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
916 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
918 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
920 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
922 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
924 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
926 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
928 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
931 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
933 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
935 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
937 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
938 &acl_ids->z_fuid, 8);
939 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
940 &acl_ids->z_fgid, 8);
941 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
943 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
945 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
947 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
949 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
951 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
955 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
957 if (obj_type == DMU_OT_ZNODE) {
958 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
961 if (obj_type == DMU_OT_ZNODE ||
962 (vap->va_type == VBLK || vap->va_type == VCHR)) {
963 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
967 if (obj_type == DMU_OT_ZNODE) {
968 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
970 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
971 &acl_ids->z_fuid, 8);
972 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
973 &acl_ids->z_fgid, 8);
974 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
975 sizeof (uint64_t) * 4);
976 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
977 &acl_phys, sizeof (zfs_acl_phys_t));
978 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
979 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
980 &acl_ids->z_aclp->z_acl_count, 8);
981 locate.cb_aclp = acl_ids->z_aclp;
982 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
983 zfs_acl_data_locator, &locate,
984 acl_ids->z_aclp->z_acl_bytes);
985 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
986 acl_ids->z_fuid, acl_ids->z_fgid);
989 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
991 if (!(flag & IS_ROOT_NODE)) {
992 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
993 ASSERT(*zpp != NULL);
996 * If we are creating the root node, the "parent" we
997 * passed in is the znode for the root.
1001 (*zpp)->z_sa_hdl = sa_hdl;
1004 (*zpp)->z_pflags = pflags;
1005 (*zpp)->z_mode = mode;
1007 if (vap->va_mask & AT_XVATTR)
1008 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1010 if (obj_type == DMU_OT_ZNODE ||
1011 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1012 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1014 if (!(flag & IS_ROOT_NODE)) {
1018 vp->v_vflag |= VV_FORCEINSMQ;
1019 err = insmntque(vp, zfsvfs->z_vfs);
1020 vp->v_vflag &= ~VV_FORCEINSMQ;
1021 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1023 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1027 * Update in-core attributes. It is assumed the caller will be doing an
1028 * sa_bulk_update to push the changes out.
1031 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1035 xoap = xva_getxoptattr(xvap);
1038 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1040 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1041 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1042 ×, sizeof (times), tx);
1043 XVA_SET_RTN(xvap, XAT_CREATETIME);
1045 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1046 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1048 XVA_SET_RTN(xvap, XAT_READONLY);
1050 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1051 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1053 XVA_SET_RTN(xvap, XAT_HIDDEN);
1055 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1056 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1058 XVA_SET_RTN(xvap, XAT_SYSTEM);
1060 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1061 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1063 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1065 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1066 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1068 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1070 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1071 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1073 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1075 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1076 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1078 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1080 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1081 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1083 XVA_SET_RTN(xvap, XAT_NODUMP);
1085 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1086 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1088 XVA_SET_RTN(xvap, XAT_OPAQUE);
1090 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1091 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1092 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1093 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1095 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1096 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1098 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1100 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1101 zfs_sa_set_scanstamp(zp, xvap, tx);
1102 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1104 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1105 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1107 XVA_SET_RTN(xvap, XAT_REPARSE);
1109 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1110 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1112 XVA_SET_RTN(xvap, XAT_OFFLINE);
1114 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1115 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1117 XVA_SET_RTN(xvap, XAT_SPARSE);
1122 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1124 dmu_object_info_t doi;
1134 getnewvnode_reserve(1);
1137 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1139 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1141 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1142 getnewvnode_drop_reserve();
1146 dmu_object_info_from_db(db, &doi);
1147 if (doi.doi_bonus_type != DMU_OT_SA &&
1148 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1149 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1150 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1151 sa_buf_rele(db, NULL);
1152 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1154 getnewvnode_drop_reserve();
1156 return (SET_ERROR(EINVAL));
1159 hdl = dmu_buf_get_user(db);
1161 zp = sa_get_userdata(hdl);
1165 * Since "SA" does immediate eviction we
1166 * should never find a sa handle that doesn't
1167 * know about the znode.
1170 ASSERT3P(zp, !=, NULL);
1172 mutex_enter(&zp->z_lock);
1173 ASSERT3U(zp->z_id, ==, obj_num);
1174 if (zp->z_unlinked) {
1175 err = SET_ERROR(ENOENT);
1182 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */
1186 mutex_exit(&zp->z_lock);
1187 sa_buf_rele(db, NULL);
1188 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1191 locked = VOP_ISLOCKED(vp);
1193 if ((vp->v_iflag & VI_DOOMED) != 0 &&
1194 locked != LK_EXCLUSIVE) {
1196 * The vnode is doomed and this thread doesn't
1197 * hold the exclusive lock on it, so the vnode
1198 * must be being reclaimed by another thread.
1199 * Otherwise the doomed vnode is being reclaimed
1200 * by this thread and zfs_zget is called from
1209 getnewvnode_drop_reserve();
1214 * Not found create new znode/vnode
1215 * but only if file exists.
1217 * There is a small window where zfs_vget() could
1218 * find this object while a file create is still in
1219 * progress. This is checked for in zfs_znode_alloc()
1221 * if zfs_znode_alloc() fails it will drop the hold on the
1224 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1225 doi.doi_bonus_type, NULL);
1227 err = SET_ERROR(ENOENT);
1232 vnode_t *vp = ZTOV(zp);
1234 err = insmntque(vp, zfsvfs->z_vfs);
1236 vp->v_hash = obj_num;
1240 zfs_znode_dmu_fini(zp);
1245 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1246 getnewvnode_drop_reserve();
1251 zfs_rezget(znode_t *zp)
1253 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1254 dmu_object_info_t doi;
1257 uint64_t obj_num = zp->z_id;
1258 uint64_t mode, size;
1259 sa_bulk_attr_t bulk[8];
1264 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1266 mutex_enter(&zp->z_acl_lock);
1267 if (zp->z_acl_cached) {
1268 zfs_acl_free(zp->z_acl_cached);
1269 zp->z_acl_cached = NULL;
1272 mutex_exit(&zp->z_acl_lock);
1273 ASSERT(zp->z_sa_hdl == NULL);
1274 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1276 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1280 dmu_object_info_from_db(db, &doi);
1281 if (doi.doi_bonus_type != DMU_OT_SA &&
1282 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1283 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1284 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1285 sa_buf_rele(db, NULL);
1286 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1287 return (SET_ERROR(EINVAL));
1290 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1293 /* reload cached values */
1294 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1295 &gen, sizeof (gen));
1296 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1297 &zp->z_size, sizeof (zp->z_size));
1298 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1299 &zp->z_links, sizeof (zp->z_links));
1300 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1301 &zp->z_pflags, sizeof (zp->z_pflags));
1302 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1303 &zp->z_atime, sizeof (zp->z_atime));
1304 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1305 &zp->z_uid, sizeof (zp->z_uid));
1306 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1307 &zp->z_gid, sizeof (zp->z_gid));
1308 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1309 &mode, sizeof (mode));
1311 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1312 zfs_znode_dmu_fini(zp);
1313 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1314 return (SET_ERROR(EIO));
1319 if (gen != zp->z_gen) {
1320 zfs_znode_dmu_fini(zp);
1321 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1322 return (SET_ERROR(EIO));
1326 * It is highly improbable but still quite possible that two
1327 * objects in different datasets are created with the same
1328 * object numbers and in transaction groups with the same
1329 * numbers. znodes corresponding to those objects would
1330 * have the same z_id and z_gen, but their other attributes
1332 * zfs recv -F may replace one of such objects with the other.
1333 * As a result file properties recorded in the replaced
1334 * object's vnode may no longer match the received object's
1335 * properties. At present the only cached property is the
1336 * files type recorded in v_type.
1337 * So, handle this case by leaving the old vnode and znode
1338 * disassociated from the actual object. A new vnode and a
1339 * znode will be created if the object is accessed
1340 * (e.g. via a look-up). The old vnode and znode will be
1341 * recycled when the last vnode reference is dropped.
1344 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1345 zfs_znode_dmu_fini(zp);
1346 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1350 zp->z_unlinked = (zp->z_links == 0);
1351 zp->z_blksz = doi.doi_data_block_size;
1352 vn_pages_remove(vp, 0, 0);
1353 if (zp->z_size != size)
1354 vnode_pager_setsize(vp, zp->z_size);
1356 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1362 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1364 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1365 objset_t *os = zfsvfs->z_os;
1366 uint64_t obj = zp->z_id;
1367 uint64_t acl_obj = zfs_external_acl(zp);
1369 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1371 VERIFY(!zp->z_is_sa);
1372 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1374 VERIFY(0 == dmu_object_free(os, obj, tx));
1375 zfs_znode_dmu_fini(zp);
1376 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1381 zfs_zinactive(znode_t *zp)
1383 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1384 uint64_t z_id = zp->z_id;
1386 ASSERT(zp->z_sa_hdl);
1389 * Don't allow a zfs_zget() while were trying to release this znode
1391 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1393 mutex_enter(&zp->z_lock);
1396 * If this was the last reference to a file with no links,
1397 * remove the file from the file system.
1399 if (zp->z_unlinked) {
1400 mutex_exit(&zp->z_lock);
1401 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1406 mutex_exit(&zp->z_lock);
1407 zfs_znode_dmu_fini(zp);
1408 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1413 zfs_znode_free(znode_t *zp)
1415 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1417 ASSERT(zp->z_sa_hdl == NULL);
1419 mutex_enter(&zfsvfs->z_znodes_lock);
1420 POINTER_INVALIDATE(&zp->z_zfsvfs);
1421 list_remove(&zfsvfs->z_all_znodes, zp);
1422 mutex_exit(&zfsvfs->z_znodes_lock);
1424 if (zp->z_acl_cached) {
1425 zfs_acl_free(zp->z_acl_cached);
1426 zp->z_acl_cached = NULL;
1429 kmem_cache_free(znode_cache, zp);
1432 VFS_RELE(zfsvfs->z_vfs);
1437 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1438 uint64_t ctime[2], boolean_t have_tx)
1442 vfs_timestamp(&now);
1444 if (have_tx) { /* will sa_bulk_update happen really soon? */
1445 zp->z_atime_dirty = 0;
1448 zp->z_atime_dirty = 1;
1451 if (flag & AT_ATIME) {
1452 ZFS_TIME_ENCODE(&now, zp->z_atime);
1455 if (flag & AT_MTIME) {
1456 ZFS_TIME_ENCODE(&now, mtime);
1457 if (zp->z_zfsvfs->z_use_fuids) {
1458 zp->z_pflags |= (ZFS_ARCHIVE |
1463 if (flag & AT_CTIME) {
1464 ZFS_TIME_ENCODE(&now, ctime);
1465 if (zp->z_zfsvfs->z_use_fuids)
1466 zp->z_pflags |= ZFS_ARCHIVE;
1471 * Grow the block size for a file.
1473 * IN: zp - znode of file to free data in.
1474 * size - requested block size
1475 * tx - open transaction.
1477 * NOTE: this function assumes that the znode is write locked.
1480 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1485 if (size <= zp->z_blksz)
1488 * If the file size is already greater than the current blocksize,
1489 * we will not grow. If there is more than one block in a file,
1490 * the blocksize cannot change.
1492 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1495 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1498 if (error == ENOTSUP)
1502 /* What blocksize did we actually get? */
1503 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1508 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1509 * be calling back into the fs for a putpage(). E.g.: when truncating
1510 * a file, the pages being "thrown away* don't need to be written out.
1514 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1515 int flags, cred_t *cr)
1523 * Increase the file length
1525 * IN: zp - znode of file to free data in.
1526 * end - new end-of-file
1528 * RETURN: 0 on success, error code on failure
1531 zfs_extend(znode_t *zp, uint64_t end)
1533 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1540 * We will change zp_size, lock the whole file.
1542 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1545 * Nothing to do if file already at desired length.
1547 if (end <= zp->z_size) {
1548 zfs_range_unlock(rl);
1551 tx = dmu_tx_create(zfsvfs->z_os);
1552 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1553 zfs_sa_upgrade_txholds(tx, zp);
1554 if (end > zp->z_blksz &&
1555 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1557 * We are growing the file past the current block size.
1559 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1561 * File's blocksize is already larger than the
1562 * "recordsize" property. Only let it grow to
1563 * the next power of 2.
1565 ASSERT(!ISP2(zp->z_blksz));
1566 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1568 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1570 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1575 error = dmu_tx_assign(tx, TXG_WAIT);
1578 zfs_range_unlock(rl);
1583 zfs_grow_blocksize(zp, newblksz, tx);
1587 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1588 &zp->z_size, sizeof (zp->z_size), tx));
1590 vnode_pager_setsize(ZTOV(zp), end);
1592 zfs_range_unlock(rl);
1600 * Free space in a file.
1602 * IN: zp - znode of file to free data in.
1603 * off - start of section to free.
1604 * len - length of section to free.
1606 * RETURN: 0 on success, error code on failure
1609 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1611 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1616 * Lock the range being freed.
1618 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1621 * Nothing to do if file already at desired length.
1623 if (off >= zp->z_size) {
1624 zfs_range_unlock(rl);
1628 if (off + len > zp->z_size)
1629 len = zp->z_size - off;
1631 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1635 * In FreeBSD we cannot free block in the middle of a file,
1636 * but only at the end of a file, so this code path should
1639 vnode_pager_setsize(ZTOV(zp), off);
1642 zfs_range_unlock(rl);
1650 * IN: zp - znode of file to free data in.
1651 * end - new end-of-file.
1653 * RETURN: 0 on success, error code on failure
1656 zfs_trunc(znode_t *zp, uint64_t end)
1658 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1659 vnode_t *vp = ZTOV(zp);
1663 sa_bulk_attr_t bulk[2];
1667 * We will change zp_size, lock the whole file.
1669 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1672 * Nothing to do if file already at desired length.
1674 if (end >= zp->z_size) {
1675 zfs_range_unlock(rl);
1679 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1681 zfs_range_unlock(rl);
1684 tx = dmu_tx_create(zfsvfs->z_os);
1685 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1686 zfs_sa_upgrade_txholds(tx, zp);
1687 dmu_tx_mark_netfree(tx);
1688 error = dmu_tx_assign(tx, TXG_WAIT);
1691 zfs_range_unlock(rl);
1696 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1697 NULL, &zp->z_size, sizeof (zp->z_size));
1700 zp->z_pflags &= ~ZFS_SPARSE;
1701 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1702 NULL, &zp->z_pflags, 8);
1704 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1709 * Clear any mapped pages in the truncated region. This has to
1710 * happen outside of the transaction to avoid the possibility of
1711 * a deadlock with someone trying to push a page that we are
1712 * about to invalidate.
1714 vnode_pager_setsize(vp, end);
1716 zfs_range_unlock(rl);
1722 * Free space in a file
1724 * IN: zp - znode of file to free data in.
1725 * off - start of range
1726 * len - end of range (0 => EOF)
1727 * flag - current file open mode flags.
1728 * log - TRUE if this action should be logged
1730 * RETURN: 0 on success, error code on failure
1733 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1735 vnode_t *vp = ZTOV(zp);
1737 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1738 zilog_t *zilog = zfsvfs->z_log;
1740 uint64_t mtime[2], ctime[2];
1741 sa_bulk_attr_t bulk[3];
1745 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1746 sizeof (mode))) != 0)
1749 if (off > zp->z_size) {
1750 error = zfs_extend(zp, off+len);
1751 if (error == 0 && log)
1758 * Check for any locks in the region to be freed.
1761 if (MANDLOCK(vp, (mode_t)mode)) {
1762 uint64_t length = (len ? len : zp->z_size - off);
1763 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1768 error = zfs_trunc(zp, off);
1770 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1771 off + len > zp->z_size)
1772 error = zfs_extend(zp, off+len);
1777 tx = dmu_tx_create(zfsvfs->z_os);
1778 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1779 zfs_sa_upgrade_txholds(tx, zp);
1780 error = dmu_tx_assign(tx, TXG_WAIT);
1786 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1787 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1788 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1789 NULL, &zp->z_pflags, 8);
1790 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1791 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1794 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1801 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1803 uint64_t moid, obj, sa_obj, version;
1804 uint64_t sense = ZFS_CASE_SENSITIVE;
1809 znode_t *rootzp = NULL;
1813 zfs_acl_ids_t acl_ids;
1816 * First attempt to create master node.
1819 * In an empty objset, there are no blocks to read and thus
1820 * there can be no i/o errors (which we assert below).
1822 moid = MASTER_NODE_OBJ;
1823 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1824 DMU_OT_NONE, 0, tx);
1828 * Set starting attributes.
1830 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1832 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1833 /* For the moment we expect all zpl props to be uint64_ts */
1837 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1838 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1839 name = nvpair_name(elem);
1840 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1844 error = zap_update(os, moid, name, 8, 1, &val, tx);
1847 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1849 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1852 ASSERT(version != 0);
1853 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1856 * Create zap object used for SA attribute registration
1859 if (version >= ZPL_VERSION_SA) {
1860 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1861 DMU_OT_NONE, 0, tx);
1862 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1868 * Create a delete queue.
1870 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1872 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1876 * Create root znode. Create minimal znode/vnode/zfsvfs
1877 * to allow zfs_mknode to work.
1880 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1881 vattr.va_type = VDIR;
1882 vattr.va_mode = S_IFDIR|0755;
1883 vattr.va_uid = crgetuid(cr);
1884 vattr.va_gid = crgetgid(cr);
1886 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
1888 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1889 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1890 rootzp->z_moved = 0;
1891 rootzp->z_unlinked = 0;
1892 rootzp->z_atime_dirty = 0;
1893 rootzp->z_is_sa = USE_SA(version, os);
1896 zfsvfs->z_parent = zfsvfs;
1897 zfsvfs->z_version = version;
1898 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
1899 zfsvfs->z_use_sa = USE_SA(version, os);
1900 zfsvfs->z_norm = norm;
1902 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1903 &zfsvfs->z_attr_table);
1908 * Fold case on file systems that are always or sometimes case
1911 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1912 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
1914 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1915 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
1916 offsetof(znode_t, z_link_node));
1918 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1919 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1921 rootzp->z_zfsvfs = zfsvfs;
1922 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1923 cr, NULL, &acl_ids));
1924 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1925 ASSERT3P(zp, ==, rootzp);
1926 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1928 zfs_acl_ids_free(&acl_ids);
1929 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1931 sa_handle_destroy(rootzp->z_sa_hdl);
1932 kmem_cache_free(znode_cache, rootzp);
1935 * Create shares directory
1938 error = zfs_create_share_dir(zfsvfs, tx);
1942 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1943 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1944 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1947 #endif /* _KERNEL */
1950 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1952 uint64_t sa_obj = 0;
1955 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1956 if (error != 0 && error != ENOENT)
1959 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1964 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1965 dmu_buf_t **db, void *tag)
1967 dmu_object_info_t doi;
1970 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1973 dmu_object_info_from_db(*db, &doi);
1974 if ((doi.doi_bonus_type != DMU_OT_SA &&
1975 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1976 doi.doi_bonus_type == DMU_OT_ZNODE &&
1977 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1978 sa_buf_rele(*db, tag);
1979 return (SET_ERROR(ENOTSUP));
1982 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1984 sa_buf_rele(*db, tag);
1992 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1994 sa_handle_destroy(hdl);
1995 sa_buf_rele(db, tag);
1999 * Given an object number, return its parent object number and whether
2000 * or not the object is an extended attribute directory.
2003 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
2004 uint64_t *pobjp, int *is_xattrdir)
2009 uint64_t parent_mode;
2010 sa_bulk_attr_t bulk[3];
2011 sa_handle_t *sa_hdl;
2016 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2017 &parent, sizeof (parent));
2018 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2019 &pflags, sizeof (pflags));
2020 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2021 &mode, sizeof (mode));
2023 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2027 * When a link is removed its parent pointer is not changed and will
2028 * be invalid. There are two cases where a link is removed but the
2029 * file stays around, when it goes to the delete queue and when there
2030 * are additional links.
2032 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2036 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2037 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2041 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2044 * Extended attributes can be applied to files, directories, etc.
2045 * Otherwise the parent must be a directory.
2047 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2048 return (SET_ERROR(EINVAL));
2056 * Given an object number, return some zpl level statistics
2059 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2062 sa_bulk_attr_t bulk[4];
2065 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2066 &sb->zs_mode, sizeof (sb->zs_mode));
2067 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2068 &sb->zs_gen, sizeof (sb->zs_gen));
2069 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2070 &sb->zs_links, sizeof (sb->zs_links));
2071 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2072 &sb->zs_ctime, sizeof (sb->zs_ctime));
2074 return (sa_bulk_lookup(hdl, bulk, count));
2078 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2079 sa_attr_type_t *sa_table, char *buf, int len)
2081 sa_handle_t *sa_hdl;
2082 sa_handle_t *prevhdl = NULL;
2083 dmu_buf_t *prevdb = NULL;
2084 dmu_buf_t *sa_db = NULL;
2085 char *path = buf + len - 1;
2093 char component[MAXNAMELEN + 2];
2098 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2100 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2101 &is_xattrdir)) != 0)
2112 (void) sprintf(component + 1, "<xattrdir>");
2114 error = zap_value_search(osp, pobj, obj,
2115 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2120 complen = strlen(component);
2122 ASSERT(path >= buf);
2123 bcopy(component, path, complen);
2126 if (sa_hdl != hdl) {
2130 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2138 if (sa_hdl != NULL && sa_hdl != hdl) {
2139 ASSERT(sa_db != NULL);
2140 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2144 (void) memmove(buf, path, buf + len - path);
2150 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2152 sa_attr_type_t *sa_table;
2157 error = zfs_sa_setup(osp, &sa_table);
2161 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2165 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2167 zfs_release_sa_handle(hdl, db, FTAG);
2172 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2175 char *path = buf + len - 1;
2176 sa_attr_type_t *sa_table;
2183 error = zfs_sa_setup(osp, &sa_table);
2187 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2191 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2193 zfs_release_sa_handle(hdl, db, FTAG);
2197 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2199 zfs_release_sa_handle(hdl, db, FTAG);