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 unless we are the root
578 * node of a snapshot mounted under .zfs.
580 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
581 ZTOV(zp)->v_flag |= VROOT;
583 mutex_exit(&zp->z_lock);
588 zfs_znode_dmu_fini(znode_t *zp)
590 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
592 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
594 sa_handle_destroy(zp->z_sa_hdl);
599 zfs_vnode_forget(vnode_t *vp)
602 /* copied from insmntque_stddtr */
604 vp->v_op = &dead_vnodeops;
610 * Construct a new znode/vnode and intialize.
612 * This does not do a call to dmu_set_user() that is
613 * up to the caller to do, in case you don't want to
617 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
618 dmu_object_type_t obj_type, sa_handle_t *hdl)
624 sa_bulk_attr_t bulk[9];
628 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
630 KASSERT(curthread->td_vp_reserv > 0,
631 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
632 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
634 kmem_cache_free(znode_cache, zp);
640 ASSERT(zp->z_dirlocks == NULL);
641 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
645 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
646 * the zfs_znode_move() callback.
650 zp->z_atime_dirty = 0;
652 zp->z_id = db->db_object;
654 zp->z_seq = 0x7A4653;
659 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
661 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
662 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
663 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
665 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
667 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
669 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
670 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
672 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
674 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
677 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
679 sa_handle_destroy(zp->z_sa_hdl);
680 zfs_vnode_forget(vp);
682 kmem_cache_free(znode_cache, zp);
688 vp->v_type = IFTOVT((mode_t)mode);
690 switch (vp->v_type) {
692 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
699 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
700 &rdev, sizeof (rdev)) == 0);
702 vp->v_rdev = zfs_cmpldev(rdev);
711 vp->v_op = &zfs_fifoops;
714 if (parent == zfsvfs->z_shares_dir) {
715 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
716 vp->v_op = &zfs_shareops;
721 vn_setops(vp, zfs_symvnodeops);
724 vn_setops(vp, zfs_evnodeops);
729 mutex_enter(&zfsvfs->z_znodes_lock);
730 list_insert_tail(&zfsvfs->z_all_znodes, zp);
733 * Everything else must be valid before assigning z_zfsvfs makes the
734 * znode eligible for zfs_znode_move().
736 zp->z_zfsvfs = zfsvfs;
737 mutex_exit(&zfsvfs->z_znodes_lock);
740 * Acquire vnode lock before making it available to the world.
742 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
744 if (vp->v_type != VFIFO)
748 VFS_HOLD(zfsvfs->z_vfs);
753 static uint64_t empty_xattr;
754 static uint64_t pad[4];
755 static zfs_acl_phys_t acl_phys;
757 * Create a new DMU object to hold a zfs znode.
759 * IN: dzp - parent directory for new znode
760 * vap - file attributes for new znode
761 * tx - dmu transaction id for zap operations
762 * cr - credentials of caller
764 * IS_ROOT_NODE - new object will be root
765 * IS_XATTR - new object is an attribute
766 * bonuslen - length of bonus buffer
767 * setaclp - File/Dir initial ACL
768 * fuidp - Tracks fuid allocation.
770 * OUT: zpp - allocated znode
774 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
775 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
777 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
778 uint64_t mode, size, links, parent, pflags;
779 uint64_t dzp_pflags = 0;
781 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
788 dmu_object_type_t obj_type;
789 sa_bulk_attr_t sa_attrs[ZPL_END];
791 zfs_acl_locator_cb_t locate = { 0 };
793 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
795 if (zfsvfs->z_replay) {
796 obj = vap->va_nodeid;
797 now = vap->va_ctime; /* see zfs_replay_create() */
798 gen = vap->va_nblocks; /* ditto */
802 gen = dmu_tx_get_txg(tx);
805 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
806 bonuslen = (obj_type == DMU_OT_SA) ?
807 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
810 * Create a new DMU object.
813 * There's currently no mechanism for pre-reading the blocks that will
814 * be needed to allocate a new object, so we accept the small chance
815 * that there will be an i/o error and we will fail one of the
818 if (vap->va_type == VDIR) {
819 if (zfsvfs->z_replay) {
820 VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj,
821 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
822 obj_type, bonuslen, tx));
824 obj = zap_create_norm(zfsvfs->z_os,
825 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
826 obj_type, bonuslen, tx);
829 if (zfsvfs->z_replay) {
830 VERIFY0(dmu_object_claim(zfsvfs->z_os, obj,
831 DMU_OT_PLAIN_FILE_CONTENTS, 0,
832 obj_type, bonuslen, tx));
834 obj = dmu_object_alloc(zfsvfs->z_os,
835 DMU_OT_PLAIN_FILE_CONTENTS, 0,
836 obj_type, bonuslen, tx);
840 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
841 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
844 * If this is the root, fix up the half-initialized parent pointer
845 * to reference the just-allocated physical data area.
847 if (flag & IS_ROOT_NODE) {
850 dzp_pflags = dzp->z_pflags;
854 * If parent is an xattr, so am I.
856 if (dzp_pflags & ZFS_XATTR) {
860 if (zfsvfs->z_use_fuids)
861 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
865 if (vap->va_type == VDIR) {
866 size = 2; /* contents ("." and "..") */
867 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
872 if (vap->va_type == VBLK || vap->va_type == VCHR) {
873 rdev = zfs_expldev(vap->va_rdev);
877 mode = acl_ids->z_mode;
882 * No execs denied will be deterimed when zfs_mode_compute() is called.
884 pflags |= acl_ids->z_aclp->z_hints &
885 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
886 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
888 ZFS_TIME_ENCODE(&now, crtime);
889 ZFS_TIME_ENCODE(&now, ctime);
891 if (vap->va_mask & AT_ATIME) {
892 ZFS_TIME_ENCODE(&vap->va_atime, atime);
894 ZFS_TIME_ENCODE(&now, atime);
897 if (vap->va_mask & AT_MTIME) {
898 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
900 ZFS_TIME_ENCODE(&now, mtime);
903 /* Now add in all of the "SA" attributes */
904 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
908 * Setup the array of attributes to be replaced/set on the new file
910 * order for DMU_OT_ZNODE is critical since it needs to be constructed
911 * in the old znode_phys_t format. Don't change this ordering
914 if (obj_type == DMU_OT_ZNODE) {
915 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
917 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
919 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
921 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
923 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
925 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
927 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
929 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
932 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
939 &acl_ids->z_fuid, 8);
940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
941 &acl_ids->z_fgid, 8);
942 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
944 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
946 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
948 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
950 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
952 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
956 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
958 if (obj_type == DMU_OT_ZNODE) {
959 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
962 if (obj_type == DMU_OT_ZNODE ||
963 (vap->va_type == VBLK || vap->va_type == VCHR)) {
964 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
968 if (obj_type == DMU_OT_ZNODE) {
969 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
971 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
972 &acl_ids->z_fuid, 8);
973 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
974 &acl_ids->z_fgid, 8);
975 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
976 sizeof (uint64_t) * 4);
977 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
978 &acl_phys, sizeof (zfs_acl_phys_t));
979 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
980 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
981 &acl_ids->z_aclp->z_acl_count, 8);
982 locate.cb_aclp = acl_ids->z_aclp;
983 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
984 zfs_acl_data_locator, &locate,
985 acl_ids->z_aclp->z_acl_bytes);
986 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
987 acl_ids->z_fuid, acl_ids->z_fgid);
990 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
992 if (!(flag & IS_ROOT_NODE)) {
993 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
994 ASSERT(*zpp != NULL);
997 * If we are creating the root node, the "parent" we
998 * passed in is the znode for the root.
1002 (*zpp)->z_sa_hdl = sa_hdl;
1005 (*zpp)->z_pflags = pflags;
1006 (*zpp)->z_mode = mode;
1008 if (vap->va_mask & AT_XVATTR)
1009 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1011 if (obj_type == DMU_OT_ZNODE ||
1012 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1013 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1015 if (!(flag & IS_ROOT_NODE)) {
1019 vp->v_vflag |= VV_FORCEINSMQ;
1020 err = insmntque(vp, zfsvfs->z_vfs);
1021 vp->v_vflag &= ~VV_FORCEINSMQ;
1022 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1024 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1028 * Update in-core attributes. It is assumed the caller will be doing an
1029 * sa_bulk_update to push the changes out.
1032 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1036 xoap = xva_getxoptattr(xvap);
1039 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1041 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1042 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1043 ×, sizeof (times), tx);
1044 XVA_SET_RTN(xvap, XAT_CREATETIME);
1046 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1047 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1049 XVA_SET_RTN(xvap, XAT_READONLY);
1051 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1052 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1054 XVA_SET_RTN(xvap, XAT_HIDDEN);
1056 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1057 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1059 XVA_SET_RTN(xvap, XAT_SYSTEM);
1061 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1062 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1064 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1066 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1067 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1069 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1071 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1072 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1074 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1076 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1077 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1079 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1081 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1082 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1084 XVA_SET_RTN(xvap, XAT_NODUMP);
1086 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1087 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1089 XVA_SET_RTN(xvap, XAT_OPAQUE);
1091 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1092 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1093 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1094 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1096 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1097 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1099 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1101 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1102 zfs_sa_set_scanstamp(zp, xvap, tx);
1103 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1105 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1106 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1108 XVA_SET_RTN(xvap, XAT_REPARSE);
1110 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1111 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1113 XVA_SET_RTN(xvap, XAT_OFFLINE);
1115 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1116 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1118 XVA_SET_RTN(xvap, XAT_SPARSE);
1123 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1125 dmu_object_info_t doi;
1135 getnewvnode_reserve(1);
1138 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1140 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1142 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1143 getnewvnode_drop_reserve();
1147 dmu_object_info_from_db(db, &doi);
1148 if (doi.doi_bonus_type != DMU_OT_SA &&
1149 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1150 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1151 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1152 sa_buf_rele(db, NULL);
1153 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1155 getnewvnode_drop_reserve();
1157 return (SET_ERROR(EINVAL));
1160 hdl = dmu_buf_get_user(db);
1162 zp = sa_get_userdata(hdl);
1166 * Since "SA" does immediate eviction we
1167 * should never find a sa handle that doesn't
1168 * know about the znode.
1171 ASSERT3P(zp, !=, NULL);
1173 mutex_enter(&zp->z_lock);
1174 ASSERT3U(zp->z_id, ==, obj_num);
1175 if (zp->z_unlinked) {
1176 err = SET_ERROR(ENOENT);
1183 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */
1187 mutex_exit(&zp->z_lock);
1188 sa_buf_rele(db, NULL);
1189 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1192 locked = VOP_ISLOCKED(vp);
1194 if ((vp->v_iflag & VI_DOOMED) != 0 &&
1195 locked != LK_EXCLUSIVE) {
1197 * The vnode is doomed and this thread doesn't
1198 * hold the exclusive lock on it, so the vnode
1199 * must be being reclaimed by another thread.
1200 * Otherwise the doomed vnode is being reclaimed
1201 * by this thread and zfs_zget is called from
1210 getnewvnode_drop_reserve();
1215 * Not found create new znode/vnode
1216 * but only if file exists.
1218 * There is a small window where zfs_vget() could
1219 * find this object while a file create is still in
1220 * progress. This is checked for in zfs_znode_alloc()
1222 * if zfs_znode_alloc() fails it will drop the hold on the
1225 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1226 doi.doi_bonus_type, NULL);
1228 err = SET_ERROR(ENOENT);
1233 vnode_t *vp = ZTOV(zp);
1235 err = insmntque(vp, zfsvfs->z_vfs);
1237 vp->v_hash = obj_num;
1241 zfs_znode_dmu_fini(zp);
1246 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1247 getnewvnode_drop_reserve();
1252 zfs_rezget(znode_t *zp)
1254 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1255 dmu_object_info_t doi;
1258 uint64_t obj_num = zp->z_id;
1259 uint64_t mode, size;
1260 sa_bulk_attr_t bulk[8];
1265 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1267 mutex_enter(&zp->z_acl_lock);
1268 if (zp->z_acl_cached) {
1269 zfs_acl_free(zp->z_acl_cached);
1270 zp->z_acl_cached = NULL;
1273 mutex_exit(&zp->z_acl_lock);
1274 ASSERT(zp->z_sa_hdl == NULL);
1275 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1277 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1281 dmu_object_info_from_db(db, &doi);
1282 if (doi.doi_bonus_type != DMU_OT_SA &&
1283 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1284 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1285 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1286 sa_buf_rele(db, NULL);
1287 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1288 return (SET_ERROR(EINVAL));
1291 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1294 /* reload cached values */
1295 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1296 &gen, sizeof (gen));
1297 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1298 &zp->z_size, sizeof (zp->z_size));
1299 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1300 &zp->z_links, sizeof (zp->z_links));
1301 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1302 &zp->z_pflags, sizeof (zp->z_pflags));
1303 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1304 &zp->z_atime, sizeof (zp->z_atime));
1305 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1306 &zp->z_uid, sizeof (zp->z_uid));
1307 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1308 &zp->z_gid, sizeof (zp->z_gid));
1309 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1310 &mode, sizeof (mode));
1312 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1313 zfs_znode_dmu_fini(zp);
1314 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1315 return (SET_ERROR(EIO));
1320 if (gen != zp->z_gen) {
1321 zfs_znode_dmu_fini(zp);
1322 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1323 return (SET_ERROR(EIO));
1327 * It is highly improbable but still quite possible that two
1328 * objects in different datasets are created with the same
1329 * object numbers and in transaction groups with the same
1330 * numbers. znodes corresponding to those objects would
1331 * have the same z_id and z_gen, but their other attributes
1333 * zfs recv -F may replace one of such objects with the other.
1334 * As a result file properties recorded in the replaced
1335 * object's vnode may no longer match the received object's
1336 * properties. At present the only cached property is the
1337 * files type recorded in v_type.
1338 * So, handle this case by leaving the old vnode and znode
1339 * disassociated from the actual object. A new vnode and a
1340 * znode will be created if the object is accessed
1341 * (e.g. via a look-up). The old vnode and znode will be
1342 * recycled when the last vnode reference is dropped.
1345 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1346 zfs_znode_dmu_fini(zp);
1347 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1351 zp->z_unlinked = (zp->z_links == 0);
1352 zp->z_blksz = doi.doi_data_block_size;
1353 vn_pages_remove(vp, 0, 0);
1354 if (zp->z_size != size)
1355 vnode_pager_setsize(vp, zp->z_size);
1357 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1363 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1365 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1366 objset_t *os = zfsvfs->z_os;
1367 uint64_t obj = zp->z_id;
1368 uint64_t acl_obj = zfs_external_acl(zp);
1370 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1372 VERIFY(!zp->z_is_sa);
1373 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1375 VERIFY(0 == dmu_object_free(os, obj, tx));
1376 zfs_znode_dmu_fini(zp);
1377 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1382 zfs_zinactive(znode_t *zp)
1384 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1385 uint64_t z_id = zp->z_id;
1387 ASSERT(zp->z_sa_hdl);
1390 * Don't allow a zfs_zget() while were trying to release this znode
1392 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1394 mutex_enter(&zp->z_lock);
1397 * If this was the last reference to a file with no links,
1398 * remove the file from the file system.
1400 if (zp->z_unlinked) {
1401 mutex_exit(&zp->z_lock);
1402 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1407 mutex_exit(&zp->z_lock);
1408 zfs_znode_dmu_fini(zp);
1409 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1414 zfs_znode_free(znode_t *zp)
1416 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1418 ASSERT(zp->z_sa_hdl == NULL);
1420 mutex_enter(&zfsvfs->z_znodes_lock);
1421 POINTER_INVALIDATE(&zp->z_zfsvfs);
1422 list_remove(&zfsvfs->z_all_znodes, zp);
1423 mutex_exit(&zfsvfs->z_znodes_lock);
1425 if (zp->z_acl_cached) {
1426 zfs_acl_free(zp->z_acl_cached);
1427 zp->z_acl_cached = NULL;
1430 kmem_cache_free(znode_cache, zp);
1433 VFS_RELE(zfsvfs->z_vfs);
1438 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1439 uint64_t ctime[2], boolean_t have_tx)
1443 vfs_timestamp(&now);
1445 if (have_tx) { /* will sa_bulk_update happen really soon? */
1446 zp->z_atime_dirty = 0;
1449 zp->z_atime_dirty = 1;
1452 if (flag & AT_ATIME) {
1453 ZFS_TIME_ENCODE(&now, zp->z_atime);
1456 if (flag & AT_MTIME) {
1457 ZFS_TIME_ENCODE(&now, mtime);
1458 if (zp->z_zfsvfs->z_use_fuids) {
1459 zp->z_pflags |= (ZFS_ARCHIVE |
1464 if (flag & AT_CTIME) {
1465 ZFS_TIME_ENCODE(&now, ctime);
1466 if (zp->z_zfsvfs->z_use_fuids)
1467 zp->z_pflags |= ZFS_ARCHIVE;
1472 * Grow the block size for a file.
1474 * IN: zp - znode of file to free data in.
1475 * size - requested block size
1476 * tx - open transaction.
1478 * NOTE: this function assumes that the znode is write locked.
1481 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1486 if (size <= zp->z_blksz)
1489 * If the file size is already greater than the current blocksize,
1490 * we will not grow. If there is more than one block in a file,
1491 * the blocksize cannot change.
1493 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1496 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1499 if (error == ENOTSUP)
1503 /* What blocksize did we actually get? */
1504 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1509 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1510 * be calling back into the fs for a putpage(). E.g.: when truncating
1511 * a file, the pages being "thrown away* don't need to be written out.
1515 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1516 int flags, cred_t *cr)
1524 * Increase the file length
1526 * IN: zp - znode of file to free data in.
1527 * end - new end-of-file
1529 * RETURN: 0 on success, error code on failure
1532 zfs_extend(znode_t *zp, uint64_t end)
1534 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1541 * We will change zp_size, lock the whole file.
1543 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1546 * Nothing to do if file already at desired length.
1548 if (end <= zp->z_size) {
1549 zfs_range_unlock(rl);
1552 tx = dmu_tx_create(zfsvfs->z_os);
1553 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1554 zfs_sa_upgrade_txholds(tx, zp);
1555 if (end > zp->z_blksz &&
1556 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1558 * We are growing the file past the current block size.
1560 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1562 * File's blocksize is already larger than the
1563 * "recordsize" property. Only let it grow to
1564 * the next power of 2.
1566 ASSERT(!ISP2(zp->z_blksz));
1567 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1569 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1571 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1576 error = dmu_tx_assign(tx, TXG_WAIT);
1579 zfs_range_unlock(rl);
1584 zfs_grow_blocksize(zp, newblksz, tx);
1588 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1589 &zp->z_size, sizeof (zp->z_size), tx));
1591 vnode_pager_setsize(ZTOV(zp), end);
1593 zfs_range_unlock(rl);
1601 * Free space in a file.
1603 * IN: zp - znode of file to free data in.
1604 * off - start of section to free.
1605 * len - length of section to free.
1607 * RETURN: 0 on success, error code on failure
1610 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1612 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1617 * Lock the range being freed.
1619 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1622 * Nothing to do if file already at desired length.
1624 if (off >= zp->z_size) {
1625 zfs_range_unlock(rl);
1629 if (off + len > zp->z_size)
1630 len = zp->z_size - off;
1632 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1636 * In FreeBSD we cannot free block in the middle of a file,
1637 * but only at the end of a file, so this code path should
1640 vnode_pager_setsize(ZTOV(zp), off);
1643 zfs_range_unlock(rl);
1651 * IN: zp - znode of file to free data in.
1652 * end - new end-of-file.
1654 * RETURN: 0 on success, error code on failure
1657 zfs_trunc(znode_t *zp, uint64_t end)
1659 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1660 vnode_t *vp = ZTOV(zp);
1664 sa_bulk_attr_t bulk[2];
1668 * We will change zp_size, lock the whole file.
1670 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1673 * Nothing to do if file already at desired length.
1675 if (end >= zp->z_size) {
1676 zfs_range_unlock(rl);
1680 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1682 zfs_range_unlock(rl);
1685 tx = dmu_tx_create(zfsvfs->z_os);
1686 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1687 zfs_sa_upgrade_txholds(tx, zp);
1688 dmu_tx_mark_netfree(tx);
1689 error = dmu_tx_assign(tx, TXG_WAIT);
1692 zfs_range_unlock(rl);
1697 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1698 NULL, &zp->z_size, sizeof (zp->z_size));
1701 zp->z_pflags &= ~ZFS_SPARSE;
1702 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1703 NULL, &zp->z_pflags, 8);
1705 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1710 * Clear any mapped pages in the truncated region. This has to
1711 * happen outside of the transaction to avoid the possibility of
1712 * a deadlock with someone trying to push a page that we are
1713 * about to invalidate.
1715 vnode_pager_setsize(vp, end);
1717 zfs_range_unlock(rl);
1723 * Free space in a file
1725 * IN: zp - znode of file to free data in.
1726 * off - start of range
1727 * len - end of range (0 => EOF)
1728 * flag - current file open mode flags.
1729 * log - TRUE if this action should be logged
1731 * RETURN: 0 on success, error code on failure
1734 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1736 vnode_t *vp = ZTOV(zp);
1738 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1739 zilog_t *zilog = zfsvfs->z_log;
1741 uint64_t mtime[2], ctime[2];
1742 sa_bulk_attr_t bulk[3];
1746 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1747 sizeof (mode))) != 0)
1750 if (off > zp->z_size) {
1751 error = zfs_extend(zp, off+len);
1752 if (error == 0 && log)
1759 * Check for any locks in the region to be freed.
1762 if (MANDLOCK(vp, (mode_t)mode)) {
1763 uint64_t length = (len ? len : zp->z_size - off);
1764 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1769 error = zfs_trunc(zp, off);
1771 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1772 off + len > zp->z_size)
1773 error = zfs_extend(zp, off+len);
1778 tx = dmu_tx_create(zfsvfs->z_os);
1779 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1780 zfs_sa_upgrade_txholds(tx, zp);
1781 error = dmu_tx_assign(tx, TXG_WAIT);
1787 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1788 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1789 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1790 NULL, &zp->z_pflags, 8);
1791 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1792 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1795 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1802 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1804 uint64_t moid, obj, sa_obj, version;
1805 uint64_t sense = ZFS_CASE_SENSITIVE;
1810 znode_t *rootzp = NULL;
1814 zfs_acl_ids_t acl_ids;
1817 * First attempt to create master node.
1820 * In an empty objset, there are no blocks to read and thus
1821 * there can be no i/o errors (which we assert below).
1823 moid = MASTER_NODE_OBJ;
1824 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1825 DMU_OT_NONE, 0, tx);
1829 * Set starting attributes.
1831 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1833 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1834 /* For the moment we expect all zpl props to be uint64_ts */
1838 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1839 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1840 name = nvpair_name(elem);
1841 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1845 error = zap_update(os, moid, name, 8, 1, &val, tx);
1848 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1850 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1853 ASSERT(version != 0);
1854 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1857 * Create zap object used for SA attribute registration
1860 if (version >= ZPL_VERSION_SA) {
1861 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1862 DMU_OT_NONE, 0, tx);
1863 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1869 * Create a delete queue.
1871 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1873 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1877 * Create root znode. Create minimal znode/vnode/zfsvfs
1878 * to allow zfs_mknode to work.
1881 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1882 vattr.va_type = VDIR;
1883 vattr.va_mode = S_IFDIR|0755;
1884 vattr.va_uid = crgetuid(cr);
1885 vattr.va_gid = crgetgid(cr);
1887 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
1889 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1890 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1891 rootzp->z_moved = 0;
1892 rootzp->z_unlinked = 0;
1893 rootzp->z_atime_dirty = 0;
1894 rootzp->z_is_sa = USE_SA(version, os);
1897 zfsvfs->z_parent = zfsvfs;
1898 zfsvfs->z_version = version;
1899 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
1900 zfsvfs->z_use_sa = USE_SA(version, os);
1901 zfsvfs->z_norm = norm;
1903 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1904 &zfsvfs->z_attr_table);
1909 * Fold case on file systems that are always or sometimes case
1912 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1913 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
1915 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1916 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
1917 offsetof(znode_t, z_link_node));
1919 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1920 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1922 rootzp->z_zfsvfs = zfsvfs;
1923 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1924 cr, NULL, &acl_ids));
1925 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1926 ASSERT3P(zp, ==, rootzp);
1927 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1929 zfs_acl_ids_free(&acl_ids);
1930 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1932 sa_handle_destroy(rootzp->z_sa_hdl);
1933 kmem_cache_free(znode_cache, rootzp);
1936 * Create shares directory
1939 error = zfs_create_share_dir(zfsvfs, tx);
1943 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1944 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1945 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1948 #endif /* _KERNEL */
1951 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1953 uint64_t sa_obj = 0;
1956 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1957 if (error != 0 && error != ENOENT)
1960 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1965 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1966 dmu_buf_t **db, void *tag)
1968 dmu_object_info_t doi;
1971 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1974 dmu_object_info_from_db(*db, &doi);
1975 if ((doi.doi_bonus_type != DMU_OT_SA &&
1976 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1977 doi.doi_bonus_type == DMU_OT_ZNODE &&
1978 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1979 sa_buf_rele(*db, tag);
1980 return (SET_ERROR(ENOTSUP));
1983 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1985 sa_buf_rele(*db, tag);
1993 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1995 sa_handle_destroy(hdl);
1996 sa_buf_rele(db, tag);
2000 * Given an object number, return its parent object number and whether
2001 * or not the object is an extended attribute directory.
2004 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
2005 uint64_t *pobjp, int *is_xattrdir)
2010 uint64_t parent_mode;
2011 sa_bulk_attr_t bulk[3];
2012 sa_handle_t *sa_hdl;
2017 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2018 &parent, sizeof (parent));
2019 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2020 &pflags, sizeof (pflags));
2021 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2022 &mode, sizeof (mode));
2024 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2028 * When a link is removed its parent pointer is not changed and will
2029 * be invalid. There are two cases where a link is removed but the
2030 * file stays around, when it goes to the delete queue and when there
2031 * are additional links.
2033 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2037 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2038 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2042 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2045 * Extended attributes can be applied to files, directories, etc.
2046 * Otherwise the parent must be a directory.
2048 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2049 return (SET_ERROR(EINVAL));
2057 * Given an object number, return some zpl level statistics
2060 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2063 sa_bulk_attr_t bulk[4];
2066 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2067 &sb->zs_mode, sizeof (sb->zs_mode));
2068 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2069 &sb->zs_gen, sizeof (sb->zs_gen));
2070 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2071 &sb->zs_links, sizeof (sb->zs_links));
2072 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2073 &sb->zs_ctime, sizeof (sb->zs_ctime));
2075 return (sa_bulk_lookup(hdl, bulk, count));
2079 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2080 sa_attr_type_t *sa_table, char *buf, int len)
2082 sa_handle_t *sa_hdl;
2083 sa_handle_t *prevhdl = NULL;
2084 dmu_buf_t *prevdb = NULL;
2085 dmu_buf_t *sa_db = NULL;
2086 char *path = buf + len - 1;
2094 char component[MAXNAMELEN + 2];
2099 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2101 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2102 &is_xattrdir)) != 0)
2113 (void) sprintf(component + 1, "<xattrdir>");
2115 error = zap_value_search(osp, pobj, obj,
2116 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2121 complen = strlen(component);
2123 ASSERT(path >= buf);
2124 bcopy(component, path, complen);
2127 if (sa_hdl != hdl) {
2131 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2139 if (sa_hdl != NULL && sa_hdl != hdl) {
2140 ASSERT(sa_db != NULL);
2141 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2145 (void) memmove(buf, path, buf + len - path);
2151 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2153 sa_attr_type_t *sa_table;
2158 error = zfs_sa_setup(osp, &sa_table);
2162 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2166 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2168 zfs_release_sa_handle(hdl, db, FTAG);
2173 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2176 char *path = buf + len - 1;
2177 sa_attr_type_t *sa_table;
2184 error = zfs_sa_setup(osp, &sa_table);
2188 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2192 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2194 zfs_release_sa_handle(hdl, db, FTAG);
2198 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2200 zfs_release_sa_handle(hdl, db, FTAG);