4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
28 #include <sys/types.h>
29 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
34 #include <sys/mntent.h>
35 #include <sys/mkdev.h>
36 #include <sys/u8_textprep.h>
37 #include <sys/dsl_dataset.h>
39 #include <sys/vfs_opreg.h>
40 #include <sys/vnode.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
46 #include <sys/atomic.h>
48 #include "fs/fs_subr.h"
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_acl.h>
51 #include <sys/zfs_ioctl.h>
52 #include <sys/zfs_rlock.h>
53 #include <sys/zfs_fuid.h>
54 #include <sys/dnode.h>
55 #include <sys/fs/zfs.h>
56 #include <sys/kidmap.h>
60 #include <sys/refcount.h>
63 #include <sys/zfs_znode.h>
65 #include <sys/zfs_sa.h>
66 #include <sys/zfs_stat.h>
69 #include "zfs_comutil.h"
72 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
73 * turned on when DEBUG is also defined.
80 #define ZNODE_STAT_ADD(stat) ((stat)++)
82 #define ZNODE_STAT_ADD(stat) /* nothing */
83 #endif /* ZNODE_STATS */
86 * Functions needed for userland (ie: libzpool) are not put under
87 * #ifdef_KERNEL; the rest of the functions have dependencies
88 * (such as VFS logic) that will not compile easily in userland.
93 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
94 * be freed before it can be safely accessed.
96 krwlock_t zfsvfs_lock;
98 static kmem_cache_t *znode_cache = NULL;
102 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
105 * We should never drop all dbuf refs without first clearing
106 * the eviction callback.
108 panic("evicting znode %p\n", user_ptr);
113 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
117 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
119 zp->z_vnode = vn_alloc(kmflags);
120 if (zp->z_vnode == NULL) {
123 ZTOV(zp)->v_data = zp;
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;
144 zfs_znode_cache_destructor(void *buf, void *arg)
148 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
149 ASSERT(ZTOV(zp)->v_data == zp);
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 */
176 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
181 nzp->z_zfsvfs = ozp->z_zfsvfs;
185 nzp->z_vnode = ozp->z_vnode;
186 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
187 ZTOV(ozp)->v_data = ozp;
188 ZTOV(nzp)->v_data = nzp;
190 nzp->z_id = ozp->z_id;
191 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
192 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
193 nzp->z_unlinked = ozp->z_unlinked;
194 nzp->z_atime_dirty = ozp->z_atime_dirty;
195 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
196 nzp->z_blksz = ozp->z_blksz;
197 nzp->z_seq = ozp->z_seq;
198 nzp->z_mapcnt = ozp->z_mapcnt;
199 nzp->z_gen = ozp->z_gen;
200 nzp->z_sync_cnt = ozp->z_sync_cnt;
201 nzp->z_is_sa = ozp->z_is_sa;
202 nzp->z_sa_hdl = ozp->z_sa_hdl;
203 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
204 nzp->z_links = ozp->z_links;
205 nzp->z_size = ozp->z_size;
206 nzp->z_pflags = ozp->z_pflags;
207 nzp->z_uid = ozp->z_uid;
208 nzp->z_gid = ozp->z_gid;
209 nzp->z_mode = ozp->z_mode;
212 * Since this is just an idle znode and kmem is already dealing with
213 * memory pressure, release any cached ACL.
215 if (ozp->z_acl_cached) {
216 zfs_acl_free(ozp->z_acl_cached);
217 ozp->z_acl_cached = NULL;
220 sa_set_userp(nzp->z_sa_hdl, nzp);
223 * Invalidate the original znode by clearing fields that provide a
224 * pointer back to the znode. Set the low bit of the vfs pointer to
225 * ensure that zfs_znode_move() recognizes the znode as invalid in any
226 * subsequent callback.
228 ozp->z_sa_hdl = NULL;
229 POINTER_INVALIDATE(&ozp->z_zfsvfs);
235 ozp->z_moved = (uint8_t)-1;
240 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
242 znode_t *ozp = buf, *nzp = newbuf;
247 * The znode is on the file system's list of known znodes if the vfs
248 * pointer is valid. We set the low bit of the vfs pointer when freeing
249 * the znode to invalidate it, and the memory patterns written by kmem
250 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
251 * created znode sets the vfs pointer last of all to indicate that the
252 * znode is known and in a valid state to be moved by this function.
254 zfsvfs = ozp->z_zfsvfs;
255 if (!POINTER_IS_VALID(zfsvfs)) {
256 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
257 return (KMEM_CBRC_DONT_KNOW);
261 * Close a small window in which it's possible that the filesystem could
262 * be unmounted and freed, and zfsvfs, though valid in the previous
263 * statement, could point to unrelated memory by the time we try to
264 * prevent the filesystem from being unmounted.
266 rw_enter(&zfsvfs_lock, RW_WRITER);
267 if (zfsvfs != ozp->z_zfsvfs) {
268 rw_exit(&zfsvfs_lock);
269 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
270 return (KMEM_CBRC_DONT_KNOW);
274 * If the znode is still valid, then so is the file system. We know that
275 * no valid file system can be freed while we hold zfsvfs_lock, so we
276 * can safely ensure that the filesystem is not and will not be
277 * unmounted. The next statement is equivalent to ZFS_ENTER().
279 rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
280 if (zfsvfs->z_unmounted) {
282 rw_exit(&zfsvfs_lock);
283 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
284 return (KMEM_CBRC_DONT_KNOW);
286 rw_exit(&zfsvfs_lock);
288 mutex_enter(&zfsvfs->z_znodes_lock);
290 * Recheck the vfs pointer in case the znode was removed just before
291 * acquiring the lock.
293 if (zfsvfs != ozp->z_zfsvfs) {
294 mutex_exit(&zfsvfs->z_znodes_lock);
296 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
297 return (KMEM_CBRC_DONT_KNOW);
301 * At this point we know that as long as we hold z_znodes_lock, the
302 * znode cannot be freed and fields within the znode can be safely
303 * accessed. Now, prevent a race with zfs_zget().
305 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
306 mutex_exit(&zfsvfs->z_znodes_lock);
308 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
309 return (KMEM_CBRC_LATER);
313 if (mutex_tryenter(&vp->v_lock) == 0) {
314 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
315 mutex_exit(&zfsvfs->z_znodes_lock);
317 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
318 return (KMEM_CBRC_LATER);
321 /* Only move znodes that are referenced _only_ by the DNLC. */
322 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
323 mutex_exit(&vp->v_lock);
324 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
325 mutex_exit(&zfsvfs->z_znodes_lock);
327 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
328 return (KMEM_CBRC_LATER);
332 * The znode is known and in a valid state to move. We're holding the
333 * locks needed to execute the critical section.
335 zfs_znode_move_impl(ozp, nzp);
336 mutex_exit(&vp->v_lock);
337 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
339 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
340 mutex_exit(&zfsvfs->z_znodes_lock);
343 return (KMEM_CBRC_YES);
352 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
353 ASSERT(znode_cache == NULL);
354 znode_cache = kmem_cache_create("zfs_znode_cache",
355 sizeof (znode_t), 0, zfs_znode_cache_constructor,
356 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
357 kmem_cache_set_move(znode_cache, zfs_znode_move);
364 * Cleanup vfs & vnode ops
366 zfs_remove_op_tables();
372 kmem_cache_destroy(znode_cache);
374 rw_destroy(&zfsvfs_lock);
377 struct vnodeops *zfs_dvnodeops;
378 struct vnodeops *zfs_fvnodeops;
379 struct vnodeops *zfs_symvnodeops;
380 struct vnodeops *zfs_xdvnodeops;
381 struct vnodeops *zfs_evnodeops;
382 struct vnodeops *zfs_sharevnodeops;
385 zfs_remove_op_tables()
391 (void) vfs_freevfsops_by_type(zfsfstype);
398 vn_freevnodeops(zfs_dvnodeops);
400 vn_freevnodeops(zfs_fvnodeops);
402 vn_freevnodeops(zfs_symvnodeops);
404 vn_freevnodeops(zfs_xdvnodeops);
406 vn_freevnodeops(zfs_evnodeops);
407 if (zfs_sharevnodeops)
408 vn_freevnodeops(zfs_sharevnodeops);
410 zfs_dvnodeops = NULL;
411 zfs_fvnodeops = NULL;
412 zfs_symvnodeops = NULL;
413 zfs_xdvnodeops = NULL;
414 zfs_evnodeops = NULL;
415 zfs_sharevnodeops = NULL;
418 extern const fs_operation_def_t zfs_dvnodeops_template[];
419 extern const fs_operation_def_t zfs_fvnodeops_template[];
420 extern const fs_operation_def_t zfs_xdvnodeops_template[];
421 extern const fs_operation_def_t zfs_symvnodeops_template[];
422 extern const fs_operation_def_t zfs_evnodeops_template[];
423 extern const fs_operation_def_t zfs_sharevnodeops_template[];
426 zfs_create_op_tables()
431 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
432 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
433 * In this case we just return as the ops vectors are already set up.
438 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
443 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
448 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
453 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
458 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
463 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
470 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
472 zfs_acl_ids_t acl_ids;
479 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
480 vattr.va_type = VDIR;
481 vattr.va_mode = S_IFDIR|0555;
482 vattr.va_uid = crgetuid(kcred);
483 vattr.va_gid = crgetgid(kcred);
485 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
486 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
487 sharezp->z_moved = 0;
488 sharezp->z_unlinked = 0;
489 sharezp->z_atime_dirty = 0;
490 sharezp->z_zfsvfs = zfsvfs;
491 sharezp->z_is_sa = zfsvfs->z_use_sa;
497 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
498 kcred, NULL, &acl_ids));
499 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
500 ASSERT3P(zp, ==, sharezp);
501 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
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 ZTOV(sharezp)->v_count = 0;
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)
541 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
542 return (((uint64_t)major << NBITSMINOR64) |
543 ((minor_t)dev & MAXMIN32));
550 * Special cmpldev for ZFS private use.
551 * Can't use standard cmpldev since it takes
552 * a long dev_t and compresses it to dev32_t in
553 * LP64. We need to do a compaction of a long dev_t
554 * to a dev32_t in ILP32.
557 zfs_cmpldev(uint64_t dev)
560 minor_t minor = (minor_t)dev & MAXMIN64;
561 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
563 if (major > MAXMAJ32 || minor > MAXMIN32)
566 return (((dev32_t)major << NBITSMINOR32) | minor);
573 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
574 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
576 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
577 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
579 mutex_enter(&zp->z_lock);
581 ASSERT(zp->z_sa_hdl == NULL);
582 ASSERT(zp->z_acl_cached == NULL);
583 if (sa_hdl == NULL) {
584 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
585 SA_HDL_SHARED, &zp->z_sa_hdl));
587 zp->z_sa_hdl = sa_hdl;
588 sa_set_userp(sa_hdl, zp);
591 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
594 * Slap on VROOT if we are the root znode
596 if (zp->z_id == zfsvfs->z_root)
597 ZTOV(zp)->v_flag |= VROOT;
599 mutex_exit(&zp->z_lock);
604 zfs_znode_dmu_fini(znode_t *zp)
606 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
608 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
610 sa_handle_destroy(zp->z_sa_hdl);
615 * Construct a new znode/vnode and intialize.
617 * This does not do a call to dmu_set_user() that is
618 * up to the caller to do, in case you don't want to
622 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
623 dmu_object_type_t obj_type, sa_handle_t *hdl)
629 sa_bulk_attr_t bulk[9];
632 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
634 ASSERT(zp->z_dirlocks == NULL);
635 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
639 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
640 * the zfs_znode_move() callback.
644 zp->z_atime_dirty = 0;
646 zp->z_id = db->db_object;
648 zp->z_seq = 0x7A4653;
654 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
656 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
657 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
658 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
660 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
662 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
664 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
665 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
667 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
669 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
672 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
674 sa_handle_destroy(zp->z_sa_hdl);
675 kmem_cache_free(znode_cache, zp);
680 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
682 vp->v_type = IFTOVT((mode_t)mode);
684 switch (vp->v_type) {
686 if (zp->z_pflags & ZFS_XATTR) {
687 vn_setops(vp, zfs_xdvnodeops);
688 vp->v_flag |= V_XATTRDIR;
690 vn_setops(vp, zfs_dvnodeops);
692 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
698 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
699 &rdev, sizeof (rdev)) == 0);
701 vp->v_rdev = zfs_cmpldev(rdev);
707 vn_setops(vp, zfs_fvnodeops);
710 vp->v_flag |= VMODSORT;
711 if (parent == zfsvfs->z_shares_dir) {
712 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
713 vn_setops(vp, zfs_sharevnodeops);
715 vn_setops(vp, zfs_fvnodeops);
719 vn_setops(vp, zfs_symvnodeops);
722 vn_setops(vp, zfs_evnodeops);
726 mutex_enter(&zfsvfs->z_znodes_lock);
727 list_insert_tail(&zfsvfs->z_all_znodes, zp);
730 * Everything else must be valid before assigning z_zfsvfs makes the
731 * znode eligible for zfs_znode_move().
733 zp->z_zfsvfs = zfsvfs;
734 mutex_exit(&zfsvfs->z_znodes_lock);
736 VFS_HOLD(zfsvfs->z_vfs);
740 static uint64_t empty_xattr;
741 static uint64_t pad[4];
742 static zfs_acl_phys_t acl_phys;
744 * Create a new DMU object to hold a zfs znode.
746 * IN: dzp - parent directory for new znode
747 * vap - file attributes for new znode
748 * tx - dmu transaction id for zap operations
749 * cr - credentials of caller
751 * IS_ROOT_NODE - new object will be root
752 * IS_XATTR - new object is an attribute
753 * bonuslen - length of bonus buffer
754 * setaclp - File/Dir initial ACL
755 * fuidp - Tracks fuid allocation.
757 * OUT: zpp - allocated znode
761 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
762 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
764 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
765 uint64_t mode, size, links, parent, pflags;
766 uint64_t dzp_pflags = 0;
768 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
775 dmu_object_type_t obj_type;
776 sa_bulk_attr_t *sa_attrs;
778 zfs_acl_locator_cb_t locate = { 0 };
780 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
782 if (zfsvfs->z_replay) {
783 obj = vap->va_nodeid;
784 now = vap->va_ctime; /* see zfs_replay_create() */
785 gen = vap->va_nblocks; /* ditto */
789 gen = dmu_tx_get_txg(tx);
792 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
793 bonuslen = (obj_type == DMU_OT_SA) ?
794 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
797 * Create a new DMU object.
800 * There's currently no mechanism for pre-reading the blocks that will
801 * be needed to allocate a new object, so we accept the small chance
802 * that there will be an i/o error and we will fail one of the
805 if (vap->va_type == VDIR) {
806 if (zfsvfs->z_replay) {
807 err = zap_create_claim_norm(zfsvfs->z_os, obj,
808 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
809 obj_type, bonuslen, tx);
810 ASSERT3U(err, ==, 0);
812 obj = zap_create_norm(zfsvfs->z_os,
813 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
814 obj_type, bonuslen, tx);
817 if (zfsvfs->z_replay) {
818 err = dmu_object_claim(zfsvfs->z_os, obj,
819 DMU_OT_PLAIN_FILE_CONTENTS, 0,
820 obj_type, bonuslen, tx);
821 ASSERT3U(err, ==, 0);
823 obj = dmu_object_alloc(zfsvfs->z_os,
824 DMU_OT_PLAIN_FILE_CONTENTS, 0,
825 obj_type, bonuslen, tx);
829 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
830 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
833 * If this is the root, fix up the half-initialized parent pointer
834 * to reference the just-allocated physical data area.
836 if (flag & IS_ROOT_NODE) {
839 dzp_pflags = dzp->z_pflags;
843 * If parent is an xattr, so am I.
845 if (dzp_pflags & ZFS_XATTR) {
849 if (zfsvfs->z_use_fuids)
850 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
854 if (vap->va_type == VDIR) {
855 size = 2; /* contents ("." and "..") */
856 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
861 if (vap->va_type == VBLK || vap->va_type == VCHR) {
862 rdev = zfs_expldev(vap->va_rdev);
866 mode = acl_ids->z_mode;
871 * No execs denied will be deterimed when zfs_mode_compute() is called.
873 pflags |= acl_ids->z_aclp->z_hints &
874 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
875 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
877 ZFS_TIME_ENCODE(&now, crtime);
878 ZFS_TIME_ENCODE(&now, ctime);
880 if (vap->va_mask & AT_ATIME) {
881 ZFS_TIME_ENCODE(&vap->va_atime, atime);
883 ZFS_TIME_ENCODE(&now, atime);
886 if (vap->va_mask & AT_MTIME) {
887 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
889 ZFS_TIME_ENCODE(&now, mtime);
892 /* Now add in all of the "SA" attributes */
893 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
897 * Setup the array of attributes to be replaced/set on the new file
899 * order for DMU_OT_ZNODE is critical since it needs to be constructed
900 * in the old znode_phys_t format. Don't change this ordering
902 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
904 if (obj_type == DMU_OT_ZNODE) {
905 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
907 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
909 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
911 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
913 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
915 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
917 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
919 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
922 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
924 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
926 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
928 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
929 &acl_ids->z_fuid, 8);
930 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
931 &acl_ids->z_fgid, 8);
932 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
942 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
946 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
948 if (obj_type == DMU_OT_ZNODE) {
949 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
952 if (obj_type == DMU_OT_ZNODE ||
953 (vap->va_type == VBLK || vap->va_type == VCHR)) {
954 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
958 if (obj_type == DMU_OT_ZNODE) {
959 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
961 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
962 &acl_ids->z_fuid, 8);
963 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
964 &acl_ids->z_fgid, 8);
965 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
966 sizeof (uint64_t) * 4);
967 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
968 &acl_phys, sizeof (zfs_acl_phys_t));
969 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
970 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
971 &acl_ids->z_aclp->z_acl_count, 8);
972 locate.cb_aclp = acl_ids->z_aclp;
973 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
974 zfs_acl_data_locator, &locate,
975 acl_ids->z_aclp->z_acl_bytes);
976 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
977 acl_ids->z_fuid, acl_ids->z_fgid);
980 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
982 if (!(flag & IS_ROOT_NODE)) {
983 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
984 ASSERT(*zpp != NULL);
987 * If we are creating the root node, the "parent" we
988 * passed in is the znode for the root.
992 (*zpp)->z_sa_hdl = sa_hdl;
995 (*zpp)->z_pflags = pflags;
996 (*zpp)->z_mode = mode;
998 if (vap->va_mask & AT_XVATTR)
999 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1001 if (obj_type == DMU_OT_ZNODE ||
1002 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1003 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
1004 ASSERT3S(err, ==, 0);
1006 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
1007 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1011 * zfs_xvattr_set only updates the in-core attributes
1012 * it is assumed the caller will be doing an sa_bulk_update
1013 * to push the changes out
1016 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1020 xoap = xva_getxoptattr(xvap);
1023 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1025 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1026 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1027 ×, sizeof (times), tx);
1028 XVA_SET_RTN(xvap, XAT_CREATETIME);
1030 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1031 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1033 XVA_SET_RTN(xvap, XAT_READONLY);
1035 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1036 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1038 XVA_SET_RTN(xvap, XAT_HIDDEN);
1040 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1041 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1043 XVA_SET_RTN(xvap, XAT_SYSTEM);
1045 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1046 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1048 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1050 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1051 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1053 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1055 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1056 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1058 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1060 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1061 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1063 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1065 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1066 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1068 XVA_SET_RTN(xvap, XAT_NODUMP);
1070 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1071 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1073 XVA_SET_RTN(xvap, XAT_OPAQUE);
1075 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1076 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1077 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1078 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1080 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1081 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1083 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1085 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1086 zfs_sa_set_scanstamp(zp, xvap, tx);
1087 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1089 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1090 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1092 XVA_SET_RTN(xvap, XAT_REPARSE);
1094 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1095 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1097 XVA_SET_RTN(xvap, XAT_OFFLINE);
1099 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1100 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1102 XVA_SET_RTN(xvap, XAT_SPARSE);
1107 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1109 dmu_object_info_t doi;
1117 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1119 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1121 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1125 dmu_object_info_from_db(db, &doi);
1126 if (doi.doi_bonus_type != DMU_OT_SA &&
1127 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1128 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1129 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1130 sa_buf_rele(db, NULL);
1131 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1135 hdl = dmu_buf_get_user(db);
1137 zp = sa_get_userdata(hdl);
1141 * Since "SA" does immediate eviction we
1142 * should never find a sa handle that doesn't
1143 * know about the znode.
1146 ASSERT3P(zp, !=, NULL);
1148 mutex_enter(&zp->z_lock);
1149 ASSERT3U(zp->z_id, ==, obj_num);
1150 if (zp->z_unlinked) {
1157 sa_buf_rele(db, NULL);
1158 mutex_exit(&zp->z_lock);
1159 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1164 * Not found create new znode/vnode
1165 * but only if file exists.
1167 * There is a small window where zfs_vget() could
1168 * find this object while a file create is still in
1169 * progress. This is checked for in zfs_znode_alloc()
1171 * if zfs_znode_alloc() fails it will drop the hold on the
1174 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1175 doi.doi_bonus_type, NULL);
1181 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1186 zfs_rezget(znode_t *zp)
1188 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1189 dmu_object_info_t doi;
1191 uint64_t obj_num = zp->z_id;
1193 sa_bulk_attr_t bulk[8];
1198 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1200 mutex_enter(&zp->z_acl_lock);
1201 if (zp->z_acl_cached) {
1202 zfs_acl_free(zp->z_acl_cached);
1203 zp->z_acl_cached = NULL;
1206 mutex_exit(&zp->z_acl_lock);
1207 ASSERT(zp->z_sa_hdl == NULL);
1208 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1210 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1214 dmu_object_info_from_db(db, &doi);
1215 if (doi.doi_bonus_type != DMU_OT_SA &&
1216 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1217 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1218 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1219 sa_buf_rele(db, NULL);
1220 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1224 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1226 /* reload cached values */
1227 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1228 &gen, sizeof (gen));
1229 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1230 &zp->z_size, sizeof (zp->z_size));
1231 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1232 &zp->z_links, sizeof (zp->z_links));
1233 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1234 &zp->z_pflags, sizeof (zp->z_pflags));
1235 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1236 &zp->z_atime, sizeof (zp->z_atime));
1237 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1238 &zp->z_uid, sizeof (zp->z_uid));
1239 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1240 &zp->z_gid, sizeof (zp->z_gid));
1241 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1242 &mode, sizeof (mode));
1244 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1245 zfs_znode_dmu_fini(zp);
1246 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1252 if (gen != zp->z_gen) {
1253 zfs_znode_dmu_fini(zp);
1254 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1258 zp->z_unlinked = (zp->z_links == 0);
1259 zp->z_blksz = doi.doi_data_block_size;
1261 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1267 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1269 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1270 objset_t *os = zfsvfs->z_os;
1271 uint64_t obj = zp->z_id;
1272 uint64_t acl_obj = zfs_external_acl(zp);
1274 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1276 VERIFY(!zp->z_is_sa);
1277 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1279 VERIFY(0 == dmu_object_free(os, obj, tx));
1280 zfs_znode_dmu_fini(zp);
1281 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1286 zfs_zinactive(znode_t *zp)
1288 vnode_t *vp = ZTOV(zp);
1289 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1290 uint64_t z_id = zp->z_id;
1292 ASSERT(zp->z_sa_hdl);
1295 * Don't allow a zfs_zget() while were trying to release this znode
1297 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1299 mutex_enter(&zp->z_lock);
1300 mutex_enter(&vp->v_lock);
1302 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1304 * If the hold count is greater than zero, somebody has
1305 * obtained a new reference on this znode while we were
1306 * processing it here, so we are done. If we still have
1307 * mapped pages then we are also done, since we don't
1308 * want to inactivate the znode until the pages get pushed.
1310 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1311 * this seems like it would leave the znode hanging with
1312 * no chance to go inactive...
1314 mutex_exit(&vp->v_lock);
1315 mutex_exit(&zp->z_lock);
1316 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1319 mutex_exit(&vp->v_lock);
1322 * If this was the last reference to a file with no links,
1323 * remove the file from the file system.
1325 if (zp->z_unlinked) {
1326 mutex_exit(&zp->z_lock);
1327 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1332 mutex_exit(&zp->z_lock);
1333 zfs_znode_dmu_fini(zp);
1334 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1339 zfs_znode_free(znode_t *zp)
1341 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1343 vn_invalid(ZTOV(zp));
1345 ASSERT(ZTOV(zp)->v_count == 0);
1347 mutex_enter(&zfsvfs->z_znodes_lock);
1348 POINTER_INVALIDATE(&zp->z_zfsvfs);
1349 list_remove(&zfsvfs->z_all_znodes, zp);
1350 mutex_exit(&zfsvfs->z_znodes_lock);
1352 if (zp->z_acl_cached) {
1353 zfs_acl_free(zp->z_acl_cached);
1354 zp->z_acl_cached = NULL;
1357 kmem_cache_free(znode_cache, zp);
1359 VFS_RELE(zfsvfs->z_vfs);
1363 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1364 uint64_t ctime[2], boolean_t have_tx)
1370 if (have_tx) { /* will sa_bulk_update happen really soon? */
1371 zp->z_atime_dirty = 0;
1374 zp->z_atime_dirty = 1;
1377 if (flag & AT_ATIME) {
1378 ZFS_TIME_ENCODE(&now, zp->z_atime);
1381 if (flag & AT_MTIME) {
1382 ZFS_TIME_ENCODE(&now, mtime);
1383 if (zp->z_zfsvfs->z_use_fuids) {
1384 zp->z_pflags |= (ZFS_ARCHIVE |
1389 if (flag & AT_CTIME) {
1390 ZFS_TIME_ENCODE(&now, ctime);
1391 if (zp->z_zfsvfs->z_use_fuids)
1392 zp->z_pflags |= ZFS_ARCHIVE;
1397 * Grow the block size for a file.
1399 * IN: zp - znode of file to free data in.
1400 * size - requested block size
1401 * tx - open transaction.
1403 * NOTE: this function assumes that the znode is write locked.
1406 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1411 if (size <= zp->z_blksz)
1414 * If the file size is already greater than the current blocksize,
1415 * we will not grow. If there is more than one block in a file,
1416 * the blocksize cannot change.
1418 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1421 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1424 if (error == ENOTSUP)
1426 ASSERT3U(error, ==, 0);
1428 /* What blocksize did we actually get? */
1429 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1433 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1434 * be calling back into the fs for a putpage(). E.g.: when truncating
1435 * a file, the pages being "thrown away* don't need to be written out.
1439 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1440 int flags, cred_t *cr)
1447 * Increase the file length
1449 * IN: zp - znode of file to free data in.
1450 * end - new end-of-file
1452 * RETURN: 0 if success
1453 * error code if failure
1456 zfs_extend(znode_t *zp, uint64_t end)
1458 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1465 * We will change zp_size, lock the whole file.
1467 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1470 * Nothing to do if file already at desired length.
1472 if (end <= zp->z_size) {
1473 zfs_range_unlock(rl);
1477 tx = dmu_tx_create(zfsvfs->z_os);
1478 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1479 zfs_sa_upgrade_txholds(tx, zp);
1480 if (end > zp->z_blksz &&
1481 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1483 * We are growing the file past the current block size.
1485 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1486 ASSERT(!ISP2(zp->z_blksz));
1487 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1489 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1491 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1496 error = dmu_tx_assign(tx, TXG_NOWAIT);
1498 if (error == ERESTART) {
1504 zfs_range_unlock(rl);
1509 zfs_grow_blocksize(zp, newblksz, tx);
1513 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1514 &zp->z_size, sizeof (zp->z_size), tx));
1516 zfs_range_unlock(rl);
1524 * Free space in a file.
1526 * IN: zp - znode of file to free data in.
1527 * off - start of section to free.
1528 * len - length of section to free.
1530 * RETURN: 0 if success
1531 * error code if failure
1534 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1536 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1541 * Lock the range being freed.
1543 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1546 * Nothing to do if file already at desired length.
1548 if (off >= zp->z_size) {
1549 zfs_range_unlock(rl);
1553 if (off + len > zp->z_size)
1554 len = zp->z_size - off;
1556 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1558 zfs_range_unlock(rl);
1566 * IN: zp - znode of file to free data in.
1567 * end - new end-of-file.
1569 * RETURN: 0 if success
1570 * error code if failure
1573 zfs_trunc(znode_t *zp, uint64_t end)
1575 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1576 vnode_t *vp = ZTOV(zp);
1580 sa_bulk_attr_t bulk[2];
1584 * We will change zp_size, lock the whole file.
1586 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1589 * Nothing to do if file already at desired length.
1591 if (end >= zp->z_size) {
1592 zfs_range_unlock(rl);
1596 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1598 zfs_range_unlock(rl);
1602 tx = dmu_tx_create(zfsvfs->z_os);
1603 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1604 zfs_sa_upgrade_txholds(tx, zp);
1605 error = dmu_tx_assign(tx, TXG_NOWAIT);
1607 if (error == ERESTART) {
1613 zfs_range_unlock(rl);
1618 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1619 NULL, &zp->z_size, sizeof (zp->z_size));
1622 zp->z_pflags &= ~ZFS_SPARSE;
1623 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1624 NULL, &zp->z_pflags, 8);
1626 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1631 * Clear any mapped pages in the truncated region. This has to
1632 * happen outside of the transaction to avoid the possibility of
1633 * a deadlock with someone trying to push a page that we are
1634 * about to invalidate.
1636 if (vn_has_cached_data(vp)) {
1638 uint64_t start = end & PAGEMASK;
1639 int poff = end & PAGEOFFSET;
1641 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1643 * We need to zero a partial page.
1645 pagezero(pp, poff, PAGESIZE - poff);
1649 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1650 B_INVAL | B_TRUNC, NULL);
1654 zfs_range_unlock(rl);
1660 * Free space in a file
1662 * IN: zp - znode of file to free data in.
1663 * off - start of range
1664 * len - end of range (0 => EOF)
1665 * flag - current file open mode flags.
1666 * log - TRUE if this action should be logged
1668 * RETURN: 0 if success
1669 * error code if failure
1672 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1674 vnode_t *vp = ZTOV(zp);
1676 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1677 zilog_t *zilog = zfsvfs->z_log;
1679 uint64_t mtime[2], ctime[2];
1680 sa_bulk_attr_t bulk[3];
1684 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1685 sizeof (mode))) != 0)
1688 if (off > zp->z_size) {
1689 error = zfs_extend(zp, off+len);
1690 if (error == 0 && log)
1697 * Check for any locks in the region to be freed.
1700 if (MANDLOCK(vp, (mode_t)mode)) {
1701 uint64_t length = (len ? len : zp->z_size - off);
1702 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1707 error = zfs_trunc(zp, off);
1709 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1710 off + len > zp->z_size)
1711 error = zfs_extend(zp, off+len);
1716 tx = dmu_tx_create(zfsvfs->z_os);
1717 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1718 zfs_sa_upgrade_txholds(tx, zp);
1719 error = dmu_tx_assign(tx, TXG_NOWAIT);
1721 if (error == ERESTART) {
1730 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1731 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1732 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1733 NULL, &zp->z_pflags, 8);
1734 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1735 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1738 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1743 #endif /* HAVE_ZPL */
1746 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1748 uint64_t moid, obj, sa_obj, version;
1749 uint64_t sense = ZFS_CASE_SENSITIVE;
1756 znode_t *rootzp = NULL;
1760 zfs_acl_ids_t acl_ids;
1765 #endif /* HAVE_ZPL */
1768 * First attempt to create master node.
1771 * In an empty objset, there are no blocks to read and thus
1772 * there can be no i/o errors (which we assert below).
1774 moid = MASTER_NODE_OBJ;
1775 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1776 DMU_OT_NONE, 0, tx);
1780 * Set starting attributes.
1782 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1784 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1785 /* For the moment we expect all zpl props to be uint64_ts */
1789 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1790 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1791 name = nvpair_name(elem);
1792 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1796 error = zap_update(os, moid, name, 8, 1, &val, tx);
1799 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1801 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1804 ASSERT(version != 0);
1805 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1808 * Create zap object used for SA attribute registration
1811 if (version >= ZPL_VERSION_SA) {
1812 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1813 DMU_OT_NONE, 0, tx);
1814 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1820 * Create a delete queue.
1822 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1824 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1829 * Create root znode. Create minimal znode/vnode/zfsvfs
1830 * to allow zfs_mknode to work.
1832 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1833 vattr.va_type = VDIR;
1834 vattr.va_mode = S_IFDIR|0755;
1835 vattr.va_uid = crgetuid(cr);
1836 vattr.va_gid = crgetgid(cr);
1838 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1839 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1840 rootzp->z_moved = 0;
1841 rootzp->z_unlinked = 0;
1842 rootzp->z_atime_dirty = 0;
1843 rootzp->z_is_sa = USE_SA(version, os);
1849 bzero(&zfsvfs, sizeof (zfsvfs_t));
1852 zfsvfs.z_parent = &zfsvfs;
1853 zfsvfs.z_version = version;
1854 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1855 zfsvfs.z_use_sa = USE_SA(version, os);
1856 zfsvfs.z_norm = norm;
1858 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1859 &zfsvfs.z_attr_table);
1864 * Fold case on file systems that are always or sometimes case
1867 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1868 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1870 /* XXX - This must be destroyed but I'm not quite sure yet so
1871 * I'm just annotating that fact when it's an issue. -Brian */
1872 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1873 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1874 offsetof(znode_t, z_link_node));
1876 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1877 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1879 rootzp->z_zfsvfs = &zfsvfs;
1880 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1881 cr, NULL, &acl_ids));
1882 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1883 ASSERT3P(zp, ==, rootzp);
1884 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1885 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1887 zfs_acl_ids_free(&acl_ids);
1888 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1890 ZTOV(rootzp)->v_count = 0;
1891 sa_handle_destroy(rootzp->z_sa_hdl);
1892 kmem_cache_free(znode_cache, rootzp);
1893 error = zfs_create_share_dir(&zfsvfs, tx);
1895 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1896 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1899 * Create root znode with code free of VFS dependencies
1901 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1902 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1904 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1905 dmu_buf_will_dirty(db, tx);
1908 * Initialize the znode physical data to zero.
1910 ASSERT(db->db_size >= sizeof (znode_phys_t));
1911 bzero(db->db_data, db->db_size);
1914 if (USE_FUIDS(version, os))
1915 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1917 pzp->zp_size = 2; /* "." and ".." */
1919 pzp->zp_parent = obj;
1920 pzp->zp_gen = dmu_tx_get_txg(tx);
1921 pzp->zp_mode = S_IFDIR | 0755;
1922 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1926 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1927 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1928 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1929 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1931 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1934 dmu_buf_rele(db, FTAG);
1935 #endif /* HAVE_ZPL */
1938 #endif /* _KERNEL */
1941 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1943 uint64_t sa_obj = 0;
1946 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1947 if (error != 0 && error != ENOENT)
1950 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1955 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1958 dmu_object_info_t doi;
1961 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1964 dmu_object_info_from_db(*db, &doi);
1965 if ((doi.doi_bonus_type != DMU_OT_SA &&
1966 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1967 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1968 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1969 sa_buf_rele(*db, FTAG);
1973 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1975 sa_buf_rele(*db, FTAG);
1983 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1985 sa_handle_destroy(hdl);
1986 sa_buf_rele(db, FTAG);
1990 * Given an object number, return its parent object number and whether
1991 * or not the object is an extended attribute directory.
1994 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
2000 sa_bulk_attr_t bulk[3];
2004 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2005 &parent, sizeof (parent));
2006 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2007 &pflags, sizeof (pflags));
2008 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2009 &mode, sizeof (mode));
2011 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2015 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2021 * Given an object number, return some zpl level statistics
2024 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2027 sa_bulk_attr_t bulk[4];
2030 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2031 &sb->zs_mode, sizeof (sb->zs_mode));
2032 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2033 &sb->zs_gen, sizeof (sb->zs_gen));
2034 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2035 &sb->zs_links, sizeof (sb->zs_links));
2036 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2037 &sb->zs_ctime, sizeof (sb->zs_ctime));
2039 return (sa_bulk_lookup(hdl, bulk, count));
2043 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2044 sa_attr_type_t *sa_table, char *buf, int len)
2046 sa_handle_t *sa_hdl;
2047 sa_handle_t *prevhdl = NULL;
2048 dmu_buf_t *prevdb = NULL;
2049 dmu_buf_t *sa_db = NULL;
2050 char *path = buf + len - 1;
2058 char component[MAXNAMELEN + 2];
2063 zfs_release_sa_handle(prevhdl, prevdb);
2065 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
2066 &is_xattrdir)) != 0)
2077 (void) sprintf(component + 1, "<xattrdir>");
2079 error = zap_value_search(osp, pobj, obj,
2080 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2085 complen = strlen(component);
2087 ASSERT(path >= buf);
2088 bcopy(component, path, complen);
2091 if (sa_hdl != hdl) {
2095 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
2103 if (sa_hdl != NULL && sa_hdl != hdl) {
2104 ASSERT(sa_db != NULL);
2105 zfs_release_sa_handle(sa_hdl, sa_db);
2109 (void) memmove(buf, path, buf + len - path);
2115 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2117 sa_attr_type_t *sa_table;
2122 error = zfs_sa_setup(osp, &sa_table);
2126 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
2130 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2132 zfs_release_sa_handle(hdl, db);
2137 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2140 char *path = buf + len - 1;
2141 sa_attr_type_t *sa_table;
2148 error = zfs_sa_setup(osp, &sa_table);
2152 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
2156 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2158 zfs_release_sa_handle(hdl, db);
2162 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2164 zfs_release_sa_handle(hdl, db);
2168 #if defined(_KERNEL) && defined(HAVE_SPL)
2169 EXPORT_SYMBOL(zfs_create_fs);
2170 EXPORT_SYMBOL(zfs_obj_to_path);