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.
26 /* Portions Copyright 2007 Jeremy Teo */
27 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
30 #include <sys/types.h>
31 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
36 #include <sys/mntent.h>
37 #include <sys/u8_textprep.h>
38 #include <sys/dsl_dataset.h>
40 #include <sys/vnode.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
45 #include <sys/atomic.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/zfs_ioctl.h>
49 #include <sys/zfs_rlock.h>
50 #include <sys/zfs_fuid.h>
51 #include <sys/dnode.h>
52 #include <sys/fs/zfs.h>
53 #include <sys/kidmap.h>
57 #include <sys/dmu_objset.h>
58 #include <sys/refcount.h>
61 #include <sys/zfs_znode.h>
63 #include <sys/zfs_sa.h>
64 #include <sys/zfs_stat.h>
65 #include <sys/refcount.h>
68 #include "zfs_comutil.h"
70 /* Used by fstat(1). */
71 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
72 SYSCTL_NULL_INT_PTR, sizeof(znode_t), "sizeof(znode_t)");
75 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
76 * turned on when DEBUG is also defined.
83 #define ZNODE_STAT_ADD(stat) ((stat)++)
85 #define ZNODE_STAT_ADD(stat) /* nothing */
86 #endif /* ZNODE_STATS */
89 * Functions needed for userland (ie: libzpool) are not put under
90 * #ifdef_KERNEL; the rest of the functions have dependencies
91 * (such as VFS logic) that will not compile easily in userland.
95 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
96 * be freed before it can be safely accessed.
98 krwlock_t zfsvfs_lock;
100 static kmem_cache_t *znode_cache = NULL;
104 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
107 * We should never drop all dbuf refs without first clearing
108 * the eviction callback.
110 panic("evicting znode %p\n", user_ptr);
113 extern struct vop_vector zfs_vnodeops;
114 extern struct vop_vector zfs_fifoops;
115 extern struct vop_vector zfs_shareops;
118 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
122 POINTER_INVALIDATE(&zp->z_zfsvfs);
124 list_link_init(&zp->z_link_node);
126 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
127 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
128 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
129 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
131 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
132 avl_create(&zp->z_range_avl, zfs_range_compare,
133 sizeof (rl_t), offsetof(rl_t, r_node));
135 zp->z_dirlocks = NULL;
136 zp->z_acl_cached = NULL;
144 zfs_znode_cache_destructor(void *buf, void *arg)
148 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
149 ASSERT(ZTOV(zp) == NULL);
151 ASSERT(!list_link_active(&zp->z_link_node));
152 mutex_destroy(&zp->z_lock);
153 rw_destroy(&zp->z_parent_lock);
154 rw_destroy(&zp->z_name_lock);
155 mutex_destroy(&zp->z_acl_lock);
156 avl_destroy(&zp->z_range_avl);
157 mutex_destroy(&zp->z_range_lock);
159 ASSERT(zp->z_dirlocks == NULL);
160 ASSERT(zp->z_acl_cached == NULL);
165 uint64_t zms_zfsvfs_invalid;
166 uint64_t zms_zfsvfs_recheck1;
167 uint64_t zms_zfsvfs_unmounted;
168 uint64_t zms_zfsvfs_recheck2;
169 uint64_t zms_obj_held;
170 uint64_t zms_vnode_locked;
171 uint64_t zms_not_only_dnlc;
173 #endif /* ZNODE_STATS */
177 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
182 nzp->z_zfsvfs = ozp->z_zfsvfs;
186 nzp->z_vnode = ozp->z_vnode;
187 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
188 ZTOV(ozp)->v_data = ozp;
189 ZTOV(nzp)->v_data = nzp;
191 nzp->z_id = ozp->z_id;
192 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
193 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
194 nzp->z_unlinked = ozp->z_unlinked;
195 nzp->z_atime_dirty = ozp->z_atime_dirty;
196 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
197 nzp->z_blksz = ozp->z_blksz;
198 nzp->z_seq = ozp->z_seq;
199 nzp->z_mapcnt = ozp->z_mapcnt;
200 nzp->z_gen = ozp->z_gen;
201 nzp->z_sync_cnt = ozp->z_sync_cnt;
202 nzp->z_is_sa = ozp->z_is_sa;
203 nzp->z_sa_hdl = ozp->z_sa_hdl;
204 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
205 nzp->z_links = ozp->z_links;
206 nzp->z_size = ozp->z_size;
207 nzp->z_pflags = ozp->z_pflags;
208 nzp->z_uid = ozp->z_uid;
209 nzp->z_gid = ozp->z_gid;
210 nzp->z_mode = ozp->z_mode;
213 * Since this is just an idle znode and kmem is already dealing with
214 * memory pressure, release any cached ACL.
216 if (ozp->z_acl_cached) {
217 zfs_acl_free(ozp->z_acl_cached);
218 ozp->z_acl_cached = NULL;
221 sa_set_userp(nzp->z_sa_hdl, nzp);
224 * Invalidate the original znode by clearing fields that provide a
225 * pointer back to the znode. Set the low bit of the vfs pointer to
226 * ensure that zfs_znode_move() recognizes the znode as invalid in any
227 * subsequent callback.
229 ozp->z_sa_hdl = NULL;
230 POINTER_INVALIDATE(&ozp->z_zfsvfs);
236 ozp->z_moved = (uint8_t)-1;
241 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
243 znode_t *ozp = buf, *nzp = newbuf;
248 * The znode is on the file system's list of known znodes if the vfs
249 * pointer is valid. We set the low bit of the vfs pointer when freeing
250 * the znode to invalidate it, and the memory patterns written by kmem
251 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
252 * created znode sets the vfs pointer last of all to indicate that the
253 * znode is known and in a valid state to be moved by this function.
255 zfsvfs = ozp->z_zfsvfs;
256 if (!POINTER_IS_VALID(zfsvfs)) {
257 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
258 return (KMEM_CBRC_DONT_KNOW);
262 * Close a small window in which it's possible that the filesystem could
263 * be unmounted and freed, and zfsvfs, though valid in the previous
264 * statement, could point to unrelated memory by the time we try to
265 * prevent the filesystem from being unmounted.
267 rw_enter(&zfsvfs_lock, RW_WRITER);
268 if (zfsvfs != ozp->z_zfsvfs) {
269 rw_exit(&zfsvfs_lock);
270 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
271 return (KMEM_CBRC_DONT_KNOW);
275 * If the znode is still valid, then so is the file system. We know that
276 * no valid file system can be freed while we hold zfsvfs_lock, so we
277 * can safely ensure that the filesystem is not and will not be
278 * unmounted. The next statement is equivalent to ZFS_ENTER().
280 rrm_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
281 if (zfsvfs->z_unmounted) {
283 rw_exit(&zfsvfs_lock);
284 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
285 return (KMEM_CBRC_DONT_KNOW);
287 rw_exit(&zfsvfs_lock);
289 mutex_enter(&zfsvfs->z_znodes_lock);
291 * Recheck the vfs pointer in case the znode was removed just before
292 * acquiring the lock.
294 if (zfsvfs != ozp->z_zfsvfs) {
295 mutex_exit(&zfsvfs->z_znodes_lock);
297 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
298 return (KMEM_CBRC_DONT_KNOW);
302 * At this point we know that as long as we hold z_znodes_lock, the
303 * znode cannot be freed and fields within the znode can be safely
304 * accessed. Now, prevent a race with zfs_zget().
306 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
307 mutex_exit(&zfsvfs->z_znodes_lock);
309 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
310 return (KMEM_CBRC_LATER);
314 if (mutex_tryenter(&vp->v_lock) == 0) {
315 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
316 mutex_exit(&zfsvfs->z_znodes_lock);
318 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
319 return (KMEM_CBRC_LATER);
322 /* Only move znodes that are referenced _only_ by the DNLC. */
323 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
324 mutex_exit(&vp->v_lock);
325 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
326 mutex_exit(&zfsvfs->z_znodes_lock);
328 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
329 return (KMEM_CBRC_LATER);
333 * The znode is known and in a valid state to move. We're holding the
334 * locks needed to execute the critical section.
336 zfs_znode_move_impl(ozp, nzp);
337 mutex_exit(&vp->v_lock);
338 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
340 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
341 mutex_exit(&zfsvfs->z_znodes_lock);
344 return (KMEM_CBRC_YES);
354 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
355 ASSERT(znode_cache == NULL);
356 znode_cache = kmem_cache_create("zfs_znode_cache",
357 sizeof (znode_t), 0, zfs_znode_cache_constructor,
358 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
359 kmem_cache_set_move(znode_cache, zfs_znode_move);
367 * Cleanup vfs & vnode ops
369 zfs_remove_op_tables();
376 kmem_cache_destroy(znode_cache);
378 rw_destroy(&zfsvfs_lock);
382 struct vnodeops *zfs_dvnodeops;
383 struct vnodeops *zfs_fvnodeops;
384 struct vnodeops *zfs_symvnodeops;
385 struct vnodeops *zfs_xdvnodeops;
386 struct vnodeops *zfs_evnodeops;
387 struct vnodeops *zfs_sharevnodeops;
390 zfs_remove_op_tables()
396 (void) vfs_freevfsops_by_type(zfsfstype);
403 vn_freevnodeops(zfs_dvnodeops);
405 vn_freevnodeops(zfs_fvnodeops);
407 vn_freevnodeops(zfs_symvnodeops);
409 vn_freevnodeops(zfs_xdvnodeops);
411 vn_freevnodeops(zfs_evnodeops);
412 if (zfs_sharevnodeops)
413 vn_freevnodeops(zfs_sharevnodeops);
415 zfs_dvnodeops = NULL;
416 zfs_fvnodeops = NULL;
417 zfs_symvnodeops = NULL;
418 zfs_xdvnodeops = NULL;
419 zfs_evnodeops = NULL;
420 zfs_sharevnodeops = NULL;
423 extern const fs_operation_def_t zfs_dvnodeops_template[];
424 extern const fs_operation_def_t zfs_fvnodeops_template[];
425 extern const fs_operation_def_t zfs_xdvnodeops_template[];
426 extern const fs_operation_def_t zfs_symvnodeops_template[];
427 extern const fs_operation_def_t zfs_evnodeops_template[];
428 extern const fs_operation_def_t zfs_sharevnodeops_template[];
431 zfs_create_op_tables()
436 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
437 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
438 * In this case we just return as the ops vectors are already set up.
443 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
448 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
453 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
458 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
463 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
468 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
476 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
478 zfs_acl_ids_t acl_ids;
484 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
485 vattr.va_type = VDIR;
486 vattr.va_mode = S_IFDIR|0555;
487 vattr.va_uid = crgetuid(kcred);
488 vattr.va_gid = crgetgid(kcred);
490 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
491 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
492 sharezp->z_moved = 0;
493 sharezp->z_unlinked = 0;
494 sharezp->z_atime_dirty = 0;
495 sharezp->z_zfsvfs = zfsvfs;
496 sharezp->z_is_sa = zfsvfs->z_use_sa;
498 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
499 kcred, NULL, &acl_ids));
500 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
501 ASSERT3P(zp, ==, sharezp);
502 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
503 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
504 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
505 zfsvfs->z_shares_dir = sharezp->z_id;
507 zfs_acl_ids_free(&acl_ids);
508 sa_handle_destroy(sharezp->z_sa_hdl);
509 kmem_cache_free(znode_cache, sharezp);
515 * define a couple of values we need available
516 * for both 64 and 32 bit environments.
519 #define NBITSMINOR64 32
522 #define MAXMAJ64 0xffffffffUL
525 #define MAXMIN64 0xffffffffUL
529 * Create special expldev for ZFS private use.
530 * Can't use standard expldev since it doesn't do
531 * what we want. The standard expldev() takes a
532 * dev32_t in LP64 and expands it to a long dev_t.
533 * We need an interface that takes a dev32_t in ILP32
534 * and expands it to a long dev_t.
537 zfs_expldev(dev_t dev)
539 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
542 * Special cmpldev for ZFS private use.
543 * Can't use standard cmpldev since it takes
544 * a long dev_t and compresses it to dev32_t in
545 * LP64. We need to do a compaction of a long dev_t
546 * to a dev32_t in ILP32.
549 zfs_cmpldev(uint64_t dev)
551 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
555 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
556 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
558 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
559 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
561 mutex_enter(&zp->z_lock);
563 ASSERT(zp->z_sa_hdl == NULL);
564 ASSERT(zp->z_acl_cached == NULL);
565 if (sa_hdl == NULL) {
566 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
567 SA_HDL_SHARED, &zp->z_sa_hdl));
569 zp->z_sa_hdl = sa_hdl;
570 sa_set_userp(sa_hdl, zp);
573 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
576 * Slap on VROOT if we are the root znode
578 if (zp->z_id == zfsvfs->z_root)
579 ZTOV(zp)->v_flag |= VROOT;
581 mutex_exit(&zp->z_lock);
586 zfs_znode_dmu_fini(znode_t *zp)
588 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
590 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
592 sa_handle_destroy(zp->z_sa_hdl);
597 zfs_vnode_forget(vnode_t *vp)
600 /* copied from insmntque_stddtr */
602 vp->v_op = &dead_vnodeops;
608 * Construct a new znode/vnode and intialize.
610 * This does not do a call to dmu_set_user() that is
611 * up to the caller to do, in case you don't want to
615 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
616 dmu_object_type_t obj_type, sa_handle_t *hdl)
622 sa_bulk_attr_t bulk[9];
626 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
628 KASSERT(curthread->td_vp_reserv > 0,
629 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
630 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
632 kmem_cache_free(znode_cache, zp);
638 ASSERT(zp->z_dirlocks == NULL);
639 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
643 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
644 * the zfs_znode_move() callback.
648 zp->z_atime_dirty = 0;
650 zp->z_id = db->db_object;
652 zp->z_seq = 0x7A4653;
657 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
659 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
660 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
661 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
663 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
665 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
667 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
670 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
672 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
675 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
677 sa_handle_destroy(zp->z_sa_hdl);
678 zfs_vnode_forget(vp);
680 kmem_cache_free(znode_cache, zp);
686 vp->v_type = IFTOVT((mode_t)mode);
688 switch (vp->v_type) {
690 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
697 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
698 &rdev, sizeof (rdev)) == 0);
700 vp->v_rdev = zfs_cmpldev(rdev);
709 vp->v_op = &zfs_fifoops;
712 if (parent == zfsvfs->z_shares_dir) {
713 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
714 vp->v_op = &zfs_shareops;
719 vn_setops(vp, zfs_symvnodeops);
722 vn_setops(vp, zfs_evnodeops);
727 mutex_enter(&zfsvfs->z_znodes_lock);
728 list_insert_tail(&zfsvfs->z_all_znodes, zp);
731 * Everything else must be valid before assigning z_zfsvfs makes the
732 * znode eligible for zfs_znode_move().
734 zp->z_zfsvfs = zfsvfs;
735 mutex_exit(&zfsvfs->z_znodes_lock);
738 * Acquire vnode lock before making it available to the world.
740 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
742 if (vp->v_type != VFIFO)
745 VFS_HOLD(zfsvfs->z_vfs);
749 static uint64_t empty_xattr;
750 static uint64_t pad[4];
751 static zfs_acl_phys_t acl_phys;
753 * Create a new DMU object to hold a zfs znode.
755 * IN: dzp - parent directory for new znode
756 * vap - file attributes for new znode
757 * tx - dmu transaction id for zap operations
758 * cr - credentials of caller
760 * IS_ROOT_NODE - new object will be root
761 * IS_XATTR - new object is an attribute
762 * bonuslen - length of bonus buffer
763 * setaclp - File/Dir initial ACL
764 * fuidp - Tracks fuid allocation.
766 * OUT: zpp - allocated znode
770 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
771 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
773 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
774 uint64_t mode, size, links, parent, pflags;
775 uint64_t dzp_pflags = 0;
777 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
784 dmu_object_type_t obj_type;
785 sa_bulk_attr_t sa_attrs[ZPL_END];
787 zfs_acl_locator_cb_t locate = { 0 };
789 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
791 if (zfsvfs->z_replay) {
792 obj = vap->va_nodeid;
793 now = vap->va_ctime; /* see zfs_replay_create() */
794 gen = vap->va_nblocks; /* ditto */
798 gen = dmu_tx_get_txg(tx);
801 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
802 bonuslen = (obj_type == DMU_OT_SA) ?
803 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
806 * Create a new DMU object.
809 * There's currently no mechanism for pre-reading the blocks that will
810 * be needed to allocate a new object, so we accept the small chance
811 * that there will be an i/o error and we will fail one of the
814 if (vap->va_type == VDIR) {
815 if (zfsvfs->z_replay) {
816 VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj,
817 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
818 obj_type, bonuslen, tx));
820 obj = zap_create_norm(zfsvfs->z_os,
821 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
822 obj_type, bonuslen, tx);
825 if (zfsvfs->z_replay) {
826 VERIFY0(dmu_object_claim(zfsvfs->z_os, obj,
827 DMU_OT_PLAIN_FILE_CONTENTS, 0,
828 obj_type, bonuslen, tx));
830 obj = dmu_object_alloc(zfsvfs->z_os,
831 DMU_OT_PLAIN_FILE_CONTENTS, 0,
832 obj_type, bonuslen, tx);
836 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
837 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
840 * If this is the root, fix up the half-initialized parent pointer
841 * to reference the just-allocated physical data area.
843 if (flag & IS_ROOT_NODE) {
846 dzp_pflags = dzp->z_pflags;
850 * If parent is an xattr, so am I.
852 if (dzp_pflags & ZFS_XATTR) {
856 if (zfsvfs->z_use_fuids)
857 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
861 if (vap->va_type == VDIR) {
862 size = 2; /* contents ("." and "..") */
863 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
868 if (vap->va_type == VBLK || vap->va_type == VCHR) {
869 rdev = zfs_expldev(vap->va_rdev);
873 mode = acl_ids->z_mode;
878 * No execs denied will be deterimed when zfs_mode_compute() is called.
880 pflags |= acl_ids->z_aclp->z_hints &
881 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
882 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
884 ZFS_TIME_ENCODE(&now, crtime);
885 ZFS_TIME_ENCODE(&now, ctime);
887 if (vap->va_mask & AT_ATIME) {
888 ZFS_TIME_ENCODE(&vap->va_atime, atime);
890 ZFS_TIME_ENCODE(&now, atime);
893 if (vap->va_mask & AT_MTIME) {
894 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
896 ZFS_TIME_ENCODE(&now, mtime);
899 /* Now add in all of the "SA" attributes */
900 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
904 * Setup the array of attributes to be replaced/set on the new file
906 * order for DMU_OT_ZNODE is critical since it needs to be constructed
907 * in the old znode_phys_t format. Don't change this ordering
910 if (obj_type == DMU_OT_ZNODE) {
911 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
913 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
915 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
917 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
919 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
921 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
923 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
925 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
928 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
930 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
932 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
935 &acl_ids->z_fuid, 8);
936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
937 &acl_ids->z_fgid, 8);
938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
942 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
944 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
946 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
948 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
952 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
954 if (obj_type == DMU_OT_ZNODE) {
955 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
958 if (obj_type == DMU_OT_ZNODE ||
959 (vap->va_type == VBLK || vap->va_type == VCHR)) {
960 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
964 if (obj_type == DMU_OT_ZNODE) {
965 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
967 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
968 &acl_ids->z_fuid, 8);
969 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
970 &acl_ids->z_fgid, 8);
971 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
972 sizeof (uint64_t) * 4);
973 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
974 &acl_phys, sizeof (zfs_acl_phys_t));
975 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
976 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
977 &acl_ids->z_aclp->z_acl_count, 8);
978 locate.cb_aclp = acl_ids->z_aclp;
979 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
980 zfs_acl_data_locator, &locate,
981 acl_ids->z_aclp->z_acl_bytes);
982 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
983 acl_ids->z_fuid, acl_ids->z_fgid);
986 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
988 if (!(flag & IS_ROOT_NODE)) {
989 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
990 ASSERT(*zpp != NULL);
993 * If we are creating the root node, the "parent" we
994 * passed in is the znode for the root.
998 (*zpp)->z_sa_hdl = sa_hdl;
1001 (*zpp)->z_pflags = pflags;
1002 (*zpp)->z_mode = mode;
1004 if (vap->va_mask & AT_XVATTR)
1005 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1007 if (obj_type == DMU_OT_ZNODE ||
1008 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1009 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1011 if (!(flag & IS_ROOT_NODE)) {
1015 vp->v_vflag |= VV_FORCEINSMQ;
1016 err = insmntque(vp, zfsvfs->z_vfs);
1017 vp->v_vflag &= ~VV_FORCEINSMQ;
1018 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1020 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1024 * Update in-core attributes. It is assumed the caller will be doing an
1025 * sa_bulk_update to push the changes out.
1028 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1032 xoap = xva_getxoptattr(xvap);
1035 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1037 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1038 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1039 ×, sizeof (times), tx);
1040 XVA_SET_RTN(xvap, XAT_CREATETIME);
1042 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1043 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1045 XVA_SET_RTN(xvap, XAT_READONLY);
1047 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1048 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1050 XVA_SET_RTN(xvap, XAT_HIDDEN);
1052 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1053 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1055 XVA_SET_RTN(xvap, XAT_SYSTEM);
1057 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1058 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1060 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1062 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1063 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1065 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1067 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1068 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1070 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1072 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1073 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1075 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1077 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1078 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1080 XVA_SET_RTN(xvap, XAT_NODUMP);
1082 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1083 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1085 XVA_SET_RTN(xvap, XAT_OPAQUE);
1087 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1088 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1089 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1090 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1092 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1093 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1095 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1097 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1098 zfs_sa_set_scanstamp(zp, xvap, tx);
1099 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1101 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1102 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1104 XVA_SET_RTN(xvap, XAT_REPARSE);
1106 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1107 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1109 XVA_SET_RTN(xvap, XAT_OFFLINE);
1111 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1112 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1114 XVA_SET_RTN(xvap, XAT_SPARSE);
1119 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1121 dmu_object_info_t doi;
1131 getnewvnode_reserve(1);
1134 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1136 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1138 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1139 getnewvnode_drop_reserve();
1143 dmu_object_info_from_db(db, &doi);
1144 if (doi.doi_bonus_type != DMU_OT_SA &&
1145 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1146 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1147 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1148 sa_buf_rele(db, NULL);
1149 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1151 getnewvnode_drop_reserve();
1153 return (SET_ERROR(EINVAL));
1156 hdl = dmu_buf_get_user(db);
1158 zp = sa_get_userdata(hdl);
1162 * Since "SA" does immediate eviction we
1163 * should never find a sa handle that doesn't
1164 * know about the znode.
1167 ASSERT3P(zp, !=, NULL);
1169 mutex_enter(&zp->z_lock);
1170 ASSERT3U(zp->z_id, ==, obj_num);
1171 if (zp->z_unlinked) {
1172 err = SET_ERROR(ENOENT);
1178 sa_buf_rele(db, NULL);
1180 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */
1184 mutex_exit(&zp->z_lock);
1185 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1188 locked = VOP_ISLOCKED(vp);
1190 if ((vp->v_iflag & VI_DOOMED) != 0 &&
1191 locked != LK_EXCLUSIVE) {
1193 * The vnode is doomed and this thread doesn't
1194 * hold the exclusive lock on it, so the vnode
1195 * must be being reclaimed by another thread.
1196 * Otherwise the doomed vnode is being reclaimed
1197 * by this thread and zfs_zget is called from
1206 getnewvnode_drop_reserve();
1211 * Not found create new znode/vnode
1212 * but only if file exists.
1214 * There is a small window where zfs_vget() could
1215 * find this object while a file create is still in
1216 * progress. This is checked for in zfs_znode_alloc()
1218 * if zfs_znode_alloc() fails it will drop the hold on the
1221 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1222 doi.doi_bonus_type, NULL);
1224 err = SET_ERROR(ENOENT);
1229 vnode_t *vp = ZTOV(zp);
1231 err = insmntque(vp, zfsvfs->z_vfs);
1233 vp->v_hash = obj_num;
1237 zfs_znode_dmu_fini(zp);
1242 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1243 getnewvnode_drop_reserve();
1248 zfs_rezget(znode_t *zp)
1250 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1251 dmu_object_info_t doi;
1254 uint64_t obj_num = zp->z_id;
1255 uint64_t mode, size;
1256 sa_bulk_attr_t bulk[8];
1261 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1263 mutex_enter(&zp->z_acl_lock);
1264 if (zp->z_acl_cached) {
1265 zfs_acl_free(zp->z_acl_cached);
1266 zp->z_acl_cached = NULL;
1269 mutex_exit(&zp->z_acl_lock);
1270 ASSERT(zp->z_sa_hdl == NULL);
1271 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1273 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1277 dmu_object_info_from_db(db, &doi);
1278 if (doi.doi_bonus_type != DMU_OT_SA &&
1279 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1280 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1281 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1282 sa_buf_rele(db, NULL);
1283 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1284 return (SET_ERROR(EINVAL));
1287 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1290 /* reload cached values */
1291 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1292 &gen, sizeof (gen));
1293 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1294 &zp->z_size, sizeof (zp->z_size));
1295 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1296 &zp->z_links, sizeof (zp->z_links));
1297 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1298 &zp->z_pflags, sizeof (zp->z_pflags));
1299 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1300 &zp->z_atime, sizeof (zp->z_atime));
1301 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1302 &zp->z_uid, sizeof (zp->z_uid));
1303 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1304 &zp->z_gid, sizeof (zp->z_gid));
1305 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1306 &mode, sizeof (mode));
1308 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1309 zfs_znode_dmu_fini(zp);
1310 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1311 return (SET_ERROR(EIO));
1316 if (gen != zp->z_gen) {
1317 zfs_znode_dmu_fini(zp);
1318 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1319 return (SET_ERROR(EIO));
1323 * It is highly improbable but still quite possible that two
1324 * objects in different datasets are created with the same
1325 * object numbers and in transaction groups with the same
1326 * numbers. znodes corresponding to those objects would
1327 * have the same z_id and z_gen, but their other attributes
1329 * zfs recv -F may replace one of such objects with the other.
1330 * As a result file properties recorded in the replaced
1331 * object's vnode may no longer match the received object's
1332 * properties. At present the only cached property is the
1333 * files type recorded in v_type.
1334 * So, handle this case by leaving the old vnode and znode
1335 * disassociated from the actual object. A new vnode and a
1336 * znode will be created if the object is accessed
1337 * (e.g. via a look-up). The old vnode and znode will be
1338 * recycled when the last vnode reference is dropped.
1341 if (vp != NULL && vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1342 zfs_znode_dmu_fini(zp);
1343 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1347 zp->z_unlinked = (zp->z_links == 0);
1348 zp->z_blksz = doi.doi_data_block_size;
1350 vn_pages_remove(vp, 0, 0);
1351 if (zp->z_size != size)
1352 vnode_pager_setsize(vp, zp->z_size);
1355 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1361 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1363 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1364 objset_t *os = zfsvfs->z_os;
1365 uint64_t obj = zp->z_id;
1366 uint64_t acl_obj = zfs_external_acl(zp);
1368 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1370 VERIFY(!zp->z_is_sa);
1371 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1373 VERIFY(0 == dmu_object_free(os, obj, tx));
1374 zfs_znode_dmu_fini(zp);
1375 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1380 zfs_zinactive(znode_t *zp)
1382 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1383 uint64_t z_id = zp->z_id;
1385 ASSERT(zp->z_sa_hdl);
1388 * Don't allow a zfs_zget() while were trying to release this znode
1390 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1392 mutex_enter(&zp->z_lock);
1395 * If this was the last reference to a file with no links,
1396 * remove the file from the file system.
1398 if (zp->z_unlinked) {
1399 mutex_exit(&zp->z_lock);
1400 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1405 mutex_exit(&zp->z_lock);
1406 zfs_znode_dmu_fini(zp);
1407 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1412 zfs_znode_free(znode_t *zp)
1414 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1416 ASSERT(zp->z_sa_hdl == NULL);
1418 mutex_enter(&zfsvfs->z_znodes_lock);
1419 POINTER_INVALIDATE(&zp->z_zfsvfs);
1420 list_remove(&zfsvfs->z_all_znodes, zp);
1421 mutex_exit(&zfsvfs->z_znodes_lock);
1423 if (zp->z_acl_cached) {
1424 zfs_acl_free(zp->z_acl_cached);
1425 zp->z_acl_cached = NULL;
1428 kmem_cache_free(znode_cache, zp);
1430 VFS_RELE(zfsvfs->z_vfs);
1434 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1435 uint64_t ctime[2], boolean_t have_tx)
1441 if (have_tx) { /* will sa_bulk_update happen really soon? */
1442 zp->z_atime_dirty = 0;
1445 zp->z_atime_dirty = 1;
1448 if (flag & AT_ATIME) {
1449 ZFS_TIME_ENCODE(&now, zp->z_atime);
1452 if (flag & AT_MTIME) {
1453 ZFS_TIME_ENCODE(&now, mtime);
1454 if (zp->z_zfsvfs->z_use_fuids) {
1455 zp->z_pflags |= (ZFS_ARCHIVE |
1460 if (flag & AT_CTIME) {
1461 ZFS_TIME_ENCODE(&now, ctime);
1462 if (zp->z_zfsvfs->z_use_fuids)
1463 zp->z_pflags |= ZFS_ARCHIVE;
1468 * Grow the block size for a file.
1470 * IN: zp - znode of file to free data in.
1471 * size - requested block size
1472 * tx - open transaction.
1474 * NOTE: this function assumes that the znode is write locked.
1477 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1482 if (size <= zp->z_blksz)
1485 * If the file size is already greater than the current blocksize,
1486 * we will not grow. If there is more than one block in a file,
1487 * the blocksize cannot change.
1489 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1492 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1495 if (error == ENOTSUP)
1499 /* What blocksize did we actually get? */
1500 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1505 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1506 * be calling back into the fs for a putpage(). E.g.: when truncating
1507 * a file, the pages being "thrown away* don't need to be written out.
1511 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1512 int flags, cred_t *cr)
1520 * Increase the file length
1522 * IN: zp - znode of file to free data in.
1523 * end - new end-of-file
1525 * RETURN: 0 on success, error code on failure
1528 zfs_extend(znode_t *zp, uint64_t end)
1530 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1537 * We will change zp_size, lock the whole file.
1539 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1542 * Nothing to do if file already at desired length.
1544 if (end <= zp->z_size) {
1545 zfs_range_unlock(rl);
1548 tx = dmu_tx_create(zfsvfs->z_os);
1549 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1550 zfs_sa_upgrade_txholds(tx, zp);
1551 if (end > zp->z_blksz &&
1552 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1554 * We are growing the file past the current block size.
1556 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1558 * File's blocksize is already larger than the
1559 * "recordsize" property. Only let it grow to
1560 * the next power of 2.
1562 ASSERT(!ISP2(zp->z_blksz));
1563 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1565 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1567 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1572 error = dmu_tx_assign(tx, TXG_WAIT);
1575 zfs_range_unlock(rl);
1580 zfs_grow_blocksize(zp, newblksz, tx);
1584 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1585 &zp->z_size, sizeof (zp->z_size), tx));
1587 vnode_pager_setsize(ZTOV(zp), end);
1589 zfs_range_unlock(rl);
1597 * Free space in a file.
1599 * IN: zp - znode of file to free data in.
1600 * off - start of section to free.
1601 * len - length of section to free.
1603 * RETURN: 0 on success, error code on failure
1606 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1608 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1613 * Lock the range being freed.
1615 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1618 * Nothing to do if file already at desired length.
1620 if (off >= zp->z_size) {
1621 zfs_range_unlock(rl);
1625 if (off + len > zp->z_size)
1626 len = zp->z_size - off;
1628 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1632 * In FreeBSD we cannot free block in the middle of a file,
1633 * but only at the end of a file, so this code path should
1636 vnode_pager_setsize(ZTOV(zp), off);
1639 zfs_range_unlock(rl);
1647 * IN: zp - znode of file to free data in.
1648 * end - new end-of-file.
1650 * RETURN: 0 on success, error code on failure
1653 zfs_trunc(znode_t *zp, uint64_t end)
1655 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1656 vnode_t *vp = ZTOV(zp);
1660 sa_bulk_attr_t bulk[2];
1664 * We will change zp_size, lock the whole file.
1666 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1669 * Nothing to do if file already at desired length.
1671 if (end >= zp->z_size) {
1672 zfs_range_unlock(rl);
1676 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1678 zfs_range_unlock(rl);
1681 tx = dmu_tx_create(zfsvfs->z_os);
1682 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1683 zfs_sa_upgrade_txholds(tx, zp);
1684 dmu_tx_mark_netfree(tx);
1685 error = dmu_tx_assign(tx, TXG_WAIT);
1688 zfs_range_unlock(rl);
1693 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1694 NULL, &zp->z_size, sizeof (zp->z_size));
1697 zp->z_pflags &= ~ZFS_SPARSE;
1698 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1699 NULL, &zp->z_pflags, 8);
1701 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1706 * Clear any mapped pages in the truncated region. This has to
1707 * happen outside of the transaction to avoid the possibility of
1708 * a deadlock with someone trying to push a page that we are
1709 * about to invalidate.
1711 vnode_pager_setsize(vp, end);
1713 zfs_range_unlock(rl);
1719 * Free space in a file
1721 * IN: zp - znode of file to free data in.
1722 * off - start of range
1723 * len - end of range (0 => EOF)
1724 * flag - current file open mode flags.
1725 * log - TRUE if this action should be logged
1727 * RETURN: 0 on success, error code on failure
1730 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1732 vnode_t *vp = ZTOV(zp);
1734 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1735 zilog_t *zilog = zfsvfs->z_log;
1737 uint64_t mtime[2], ctime[2];
1738 sa_bulk_attr_t bulk[3];
1742 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1743 sizeof (mode))) != 0)
1746 if (off > zp->z_size) {
1747 error = zfs_extend(zp, off+len);
1748 if (error == 0 && log)
1755 * Check for any locks in the region to be freed.
1758 if (MANDLOCK(vp, (mode_t)mode)) {
1759 uint64_t length = (len ? len : zp->z_size - off);
1760 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1765 error = zfs_trunc(zp, off);
1767 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1768 off + len > zp->z_size)
1769 error = zfs_extend(zp, off+len);
1774 tx = dmu_tx_create(zfsvfs->z_os);
1775 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1776 zfs_sa_upgrade_txholds(tx, zp);
1777 error = dmu_tx_assign(tx, TXG_WAIT);
1783 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1784 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1785 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1786 NULL, &zp->z_pflags, 8);
1787 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1788 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1791 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1798 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1800 uint64_t moid, obj, sa_obj, version;
1801 uint64_t sense = ZFS_CASE_SENSITIVE;
1806 znode_t *rootzp = NULL;
1810 zfs_acl_ids_t acl_ids;
1813 * First attempt to create master node.
1816 * In an empty objset, there are no blocks to read and thus
1817 * there can be no i/o errors (which we assert below).
1819 moid = MASTER_NODE_OBJ;
1820 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1821 DMU_OT_NONE, 0, tx);
1825 * Set starting attributes.
1827 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1829 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1830 /* For the moment we expect all zpl props to be uint64_ts */
1834 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1835 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1836 name = nvpair_name(elem);
1837 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1841 error = zap_update(os, moid, name, 8, 1, &val, tx);
1844 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1846 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1849 ASSERT(version != 0);
1850 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1853 * Create zap object used for SA attribute registration
1856 if (version >= ZPL_VERSION_SA) {
1857 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1858 DMU_OT_NONE, 0, tx);
1859 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1865 * Create a delete queue.
1867 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1869 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1873 * Create root znode. Create minimal znode/vnode/zfsvfs
1874 * to allow zfs_mknode to work.
1877 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1878 vattr.va_type = VDIR;
1879 vattr.va_mode = S_IFDIR|0755;
1880 vattr.va_uid = crgetuid(cr);
1881 vattr.va_gid = crgetgid(cr);
1883 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
1885 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1886 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1887 rootzp->z_moved = 0;
1888 rootzp->z_unlinked = 0;
1889 rootzp->z_atime_dirty = 0;
1890 rootzp->z_is_sa = USE_SA(version, os);
1893 zfsvfs->z_parent = zfsvfs;
1894 zfsvfs->z_version = version;
1895 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
1896 zfsvfs->z_use_sa = USE_SA(version, os);
1897 zfsvfs->z_norm = norm;
1899 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1900 &zfsvfs->z_attr_table);
1905 * Fold case on file systems that are always or sometimes case
1908 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1909 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
1911 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1912 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
1913 offsetof(znode_t, z_link_node));
1915 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1916 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1918 rootzp->z_zfsvfs = zfsvfs;
1919 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1920 cr, NULL, &acl_ids));
1921 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1922 ASSERT3P(zp, ==, rootzp);
1923 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1925 zfs_acl_ids_free(&acl_ids);
1926 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1928 sa_handle_destroy(rootzp->z_sa_hdl);
1929 kmem_cache_free(znode_cache, rootzp);
1932 * Create shares directory
1935 error = zfs_create_share_dir(zfsvfs, tx);
1939 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1940 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1941 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1944 #endif /* _KERNEL */
1947 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1949 uint64_t sa_obj = 0;
1952 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1953 if (error != 0 && error != ENOENT)
1956 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1961 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1962 dmu_buf_t **db, void *tag)
1964 dmu_object_info_t doi;
1967 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1970 dmu_object_info_from_db(*db, &doi);
1971 if ((doi.doi_bonus_type != DMU_OT_SA &&
1972 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1973 doi.doi_bonus_type == DMU_OT_ZNODE &&
1974 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1975 sa_buf_rele(*db, tag);
1976 return (SET_ERROR(ENOTSUP));
1979 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1981 sa_buf_rele(*db, tag);
1989 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1991 sa_handle_destroy(hdl);
1992 sa_buf_rele(db, tag);
1996 * Given an object number, return its parent object number and whether
1997 * or not the object is an extended attribute directory.
2000 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
2001 uint64_t *pobjp, int *is_xattrdir)
2006 uint64_t parent_mode;
2007 sa_bulk_attr_t bulk[3];
2008 sa_handle_t *sa_hdl;
2013 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2014 &parent, sizeof (parent));
2015 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2016 &pflags, sizeof (pflags));
2017 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2018 &mode, sizeof (mode));
2020 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2024 * When a link is removed its parent pointer is not changed and will
2025 * be invalid. There are two cases where a link is removed but the
2026 * file stays around, when it goes to the delete queue and when there
2027 * are additional links.
2029 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2033 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2034 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2038 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2041 * Extended attributes can be applied to files, directories, etc.
2042 * Otherwise the parent must be a directory.
2044 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2045 return (SET_ERROR(EINVAL));
2053 * Given an object number, return some zpl level statistics
2056 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2059 sa_bulk_attr_t bulk[4];
2062 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2063 &sb->zs_mode, sizeof (sb->zs_mode));
2064 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2065 &sb->zs_gen, sizeof (sb->zs_gen));
2066 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2067 &sb->zs_links, sizeof (sb->zs_links));
2068 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2069 &sb->zs_ctime, sizeof (sb->zs_ctime));
2071 return (sa_bulk_lookup(hdl, bulk, count));
2075 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2076 sa_attr_type_t *sa_table, char *buf, int len)
2078 sa_handle_t *sa_hdl;
2079 sa_handle_t *prevhdl = NULL;
2080 dmu_buf_t *prevdb = NULL;
2081 dmu_buf_t *sa_db = NULL;
2082 char *path = buf + len - 1;
2090 char component[MAXNAMELEN + 2];
2095 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2097 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2098 &is_xattrdir)) != 0)
2109 (void) sprintf(component + 1, "<xattrdir>");
2111 error = zap_value_search(osp, pobj, obj,
2112 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2117 complen = strlen(component);
2119 ASSERT(path >= buf);
2120 bcopy(component, path, complen);
2123 if (sa_hdl != hdl) {
2127 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2135 if (sa_hdl != NULL && sa_hdl != hdl) {
2136 ASSERT(sa_db != NULL);
2137 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2141 (void) memmove(buf, path, buf + len - path);
2147 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2149 sa_attr_type_t *sa_table;
2154 error = zfs_sa_setup(osp, &sa_table);
2158 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2162 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2164 zfs_release_sa_handle(hdl, db, FTAG);
2169 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2172 char *path = buf + len - 1;
2173 sa_attr_type_t *sa_table;
2180 error = zfs_sa_setup(osp, &sa_table);
2184 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2188 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2190 zfs_release_sa_handle(hdl, db, FTAG);
2194 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2196 zfs_release_sa_handle(hdl, db, FTAG);