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
[ZPL_END
];
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
903 if (obj_type
== DMU_OT_ZNODE
) {
904 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
906 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
908 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
910 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
912 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
914 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
916 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
918 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(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_GEN(zfsvfs
),
927 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
928 &acl_ids
->z_fuid
, 8);
929 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
930 &acl_ids
->z_fgid
, 8);
931 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
933 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
935 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
937 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
939 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
941 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
945 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zfsvfs
), NULL
, &links
, 8);
947 if (obj_type
== DMU_OT_ZNODE
) {
948 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zfsvfs
), NULL
,
951 if (obj_type
== DMU_OT_ZNODE
||
952 (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
)) {
953 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zfsvfs
),
957 if (obj_type
== DMU_OT_ZNODE
) {
958 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
960 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
961 &acl_ids
->z_fuid
, 8);
962 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
963 &acl_ids
->z_fgid
, 8);
964 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zfsvfs
), NULL
, pad
,
965 sizeof (uint64_t) * 4);
966 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zfsvfs
), NULL
,
967 &acl_phys
, sizeof (zfs_acl_phys_t
));
968 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
969 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zfsvfs
), NULL
,
970 &acl_ids
->z_aclp
->z_acl_count
, 8);
971 locate
.cb_aclp
= acl_ids
->z_aclp
;
972 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zfsvfs
),
973 zfs_acl_data_locator
, &locate
,
974 acl_ids
->z_aclp
->z_acl_bytes
);
975 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
976 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
979 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
981 if (!(flag
& IS_ROOT_NODE
)) {
982 *zpp
= zfs_znode_alloc(zfsvfs
, db
, 0, obj_type
, sa_hdl
);
983 ASSERT(*zpp
!= NULL
);
986 * If we are creating the root node, the "parent" we
987 * passed in is the znode for the root.
991 (*zpp
)->z_sa_hdl
= sa_hdl
;
994 (*zpp
)->z_pflags
= pflags
;
995 (*zpp
)->z_mode
= mode
;
997 if (vap
->va_mask
& AT_XVATTR
)
998 zfs_xvattr_set(*zpp
, (xvattr_t
*)vap
, tx
);
1000 if (obj_type
== DMU_OT_ZNODE
||
1001 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
1002 err
= zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
);
1003 ASSERT3P(err
, ==, 0);
1005 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1009 * zfs_xvattr_set only updates the in-core attributes
1010 * it is assumed the caller will be doing an sa_bulk_update
1011 * to push the changes out
1014 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
1018 xoap
= xva_getxoptattr(xvap
);
1021 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
1023 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
1024 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(zp
->z_zfsvfs
),
1025 ×
, sizeof (times
), tx
);
1026 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
1028 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
1029 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
1031 XVA_SET_RTN(xvap
, XAT_READONLY
);
1033 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
1034 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
1036 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
1038 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
1039 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
1041 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
1043 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
1044 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
1046 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
1048 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
1049 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
1051 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
1053 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
1054 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
1056 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
1058 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
1059 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
1061 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
1063 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
1064 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
1066 XVA_SET_RTN(xvap
, XAT_NODUMP
);
1068 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
1069 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
1071 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
1073 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
1074 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
1075 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
1076 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
1078 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
1079 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
1081 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
1083 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
1084 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
1085 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
1087 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
1088 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
1090 XVA_SET_RTN(xvap
, XAT_REPARSE
);
1092 if (XVA_ISSET_REQ(xvap
, XAT_OFFLINE
)) {
1093 ZFS_ATTR_SET(zp
, ZFS_OFFLINE
, xoap
->xoa_offline
,
1095 XVA_SET_RTN(xvap
, XAT_OFFLINE
);
1097 if (XVA_ISSET_REQ(xvap
, XAT_SPARSE
)) {
1098 ZFS_ATTR_SET(zp
, ZFS_SPARSE
, xoap
->xoa_sparse
,
1100 XVA_SET_RTN(xvap
, XAT_SPARSE
);
1105 zfs_zget(zfsvfs_t
*zfsvfs
, uint64_t obj_num
, znode_t
**zpp
)
1107 dmu_object_info_t doi
;
1115 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1117 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1119 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1123 dmu_object_info_from_db(db
, &doi
);
1124 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1125 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1126 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1127 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1128 sa_buf_rele(db
, NULL
);
1129 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1133 hdl
= dmu_buf_get_user(db
);
1135 zp
= sa_get_userdata(hdl
);
1139 * Since "SA" does immediate eviction we
1140 * should never find a sa handle that doesn't
1141 * know about the znode.
1144 ASSERT3P(zp
, !=, NULL
);
1146 mutex_enter(&zp
->z_lock
);
1147 ASSERT3U(zp
->z_id
, ==, obj_num
);
1148 if (zp
->z_unlinked
) {
1155 sa_buf_rele(db
, NULL
);
1156 mutex_exit(&zp
->z_lock
);
1157 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1162 * Not found create new znode/vnode
1163 * but only if file exists.
1165 * There is a small window where zfs_vget() could
1166 * find this object while a file create is still in
1167 * progress. This is checked for in zfs_znode_alloc()
1169 * if zfs_znode_alloc() fails it will drop the hold on the
1172 zp
= zfs_znode_alloc(zfsvfs
, db
, doi
.doi_data_block_size
,
1173 doi
.doi_bonus_type
, NULL
);
1179 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1184 zfs_rezget(znode_t
*zp
)
1186 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1187 dmu_object_info_t doi
;
1189 uint64_t obj_num
= zp
->z_id
;
1191 sa_bulk_attr_t bulk
[8];
1196 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1198 mutex_enter(&zp
->z_acl_lock
);
1199 if (zp
->z_acl_cached
) {
1200 zfs_acl_free(zp
->z_acl_cached
);
1201 zp
->z_acl_cached
= NULL
;
1204 mutex_exit(&zp
->z_acl_lock
);
1205 ASSERT(zp
->z_sa_hdl
== NULL
);
1206 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1208 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1212 dmu_object_info_from_db(db
, &doi
);
1213 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1214 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1215 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1216 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1217 sa_buf_rele(db
, NULL
);
1218 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1222 zfs_znode_sa_init(zfsvfs
, zp
, db
, doi
.doi_bonus_type
, NULL
);
1224 /* reload cached values */
1225 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
,
1226 &gen
, sizeof (gen
));
1227 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
1228 &zp
->z_size
, sizeof (zp
->z_size
));
1229 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
1230 &zp
->z_links
, sizeof (zp
->z_links
));
1231 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
1232 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1233 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
1234 &zp
->z_atime
, sizeof (zp
->z_atime
));
1235 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
1236 &zp
->z_uid
, sizeof (zp
->z_uid
));
1237 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
1238 &zp
->z_gid
, sizeof (zp
->z_gid
));
1239 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
,
1240 &mode
, sizeof (mode
));
1242 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
1243 zfs_znode_dmu_fini(zp
);
1244 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1250 if (gen
!= zp
->z_gen
) {
1251 zfs_znode_dmu_fini(zp
);
1252 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1256 zp
->z_unlinked
= (zp
->z_links
== 0);
1257 zp
->z_blksz
= doi
.doi_data_block_size
;
1259 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1265 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1267 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1268 objset_t
*os
= zfsvfs
->z_os
;
1269 uint64_t obj
= zp
->z_id
;
1270 uint64_t acl_obj
= zfs_external_acl(zp
);
1272 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
1274 VERIFY(!zp
->z_is_sa
);
1275 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1277 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1278 zfs_znode_dmu_fini(zp
);
1279 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1284 zfs_zinactive(znode_t
*zp
)
1286 vnode_t
*vp
= ZTOV(zp
);
1287 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1288 uint64_t z_id
= zp
->z_id
;
1290 ASSERT(zp
->z_sa_hdl
);
1293 * Don't allow a zfs_zget() while were trying to release this znode
1295 ZFS_OBJ_HOLD_ENTER(zfsvfs
, z_id
);
1297 mutex_enter(&zp
->z_lock
);
1298 mutex_enter(&vp
->v_lock
);
1300 if (vp
->v_count
> 0 || vn_has_cached_data(vp
)) {
1302 * If the hold count is greater than zero, somebody has
1303 * obtained a new reference on this znode while we were
1304 * processing it here, so we are done. If we still have
1305 * mapped pages then we are also done, since we don't
1306 * want to inactivate the znode until the pages get pushed.
1308 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1309 * this seems like it would leave the znode hanging with
1310 * no chance to go inactive...
1312 mutex_exit(&vp
->v_lock
);
1313 mutex_exit(&zp
->z_lock
);
1314 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1317 mutex_exit(&vp
->v_lock
);
1320 * If this was the last reference to a file with no links,
1321 * remove the file from the file system.
1323 if (zp
->z_unlinked
) {
1324 mutex_exit(&zp
->z_lock
);
1325 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1330 mutex_exit(&zp
->z_lock
);
1331 zfs_znode_dmu_fini(zp
);
1332 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1337 zfs_znode_free(znode_t
*zp
)
1339 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1341 vn_invalid(ZTOV(zp
));
1343 ASSERT(ZTOV(zp
)->v_count
== 0);
1345 mutex_enter(&zfsvfs
->z_znodes_lock
);
1346 POINTER_INVALIDATE(&zp
->z_zfsvfs
);
1347 list_remove(&zfsvfs
->z_all_znodes
, zp
);
1348 mutex_exit(&zfsvfs
->z_znodes_lock
);
1350 if (zp
->z_acl_cached
) {
1351 zfs_acl_free(zp
->z_acl_cached
);
1352 zp
->z_acl_cached
= NULL
;
1355 kmem_cache_free(znode_cache
, zp
);
1357 VFS_RELE(zfsvfs
->z_vfs
);
1361 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1362 uint64_t ctime
[2], boolean_t have_tx
)
1368 if (have_tx
) { /* will sa_bulk_update happen really soon? */
1369 zp
->z_atime_dirty
= 0;
1372 zp
->z_atime_dirty
= 1;
1375 if (flag
& AT_ATIME
) {
1376 ZFS_TIME_ENCODE(&now
, zp
->z_atime
);
1379 if (flag
& AT_MTIME
) {
1380 ZFS_TIME_ENCODE(&now
, mtime
);
1381 if (zp
->z_zfsvfs
->z_use_fuids
) {
1382 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1387 if (flag
& AT_CTIME
) {
1388 ZFS_TIME_ENCODE(&now
, ctime
);
1389 if (zp
->z_zfsvfs
->z_use_fuids
)
1390 zp
->z_pflags
|= ZFS_ARCHIVE
;
1395 * Grow the block size for a file.
1397 * IN: zp - znode of file to free data in.
1398 * size - requested block size
1399 * tx - open transaction.
1401 * NOTE: this function assumes that the znode is write locked.
1404 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1409 if (size
<= zp
->z_blksz
)
1412 * If the file size is already greater than the current blocksize,
1413 * we will not grow. If there is more than one block in a file,
1414 * the blocksize cannot change.
1416 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1419 error
= dmu_object_set_blocksize(zp
->z_zfsvfs
->z_os
, zp
->z_id
,
1422 if (error
== ENOTSUP
)
1424 ASSERT3U(error
, ==, 0);
1426 /* What blocksize did we actually get? */
1427 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1431 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1432 * be calling back into the fs for a putpage(). E.g.: when truncating
1433 * a file, the pages being "thrown away* don't need to be written out.
1437 zfs_no_putpage(vnode_t
*vp
, page_t
*pp
, u_offset_t
*offp
, size_t *lenp
,
1438 int flags
, cred_t
*cr
)
1445 * Increase the file length
1447 * IN: zp - znode of file to free data in.
1448 * end - new end-of-file
1450 * RETURN: 0 if success
1451 * error code if failure
1454 zfs_extend(znode_t
*zp
, uint64_t end
)
1456 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1463 * We will change zp_size, lock the whole file.
1465 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1468 * Nothing to do if file already at desired length.
1470 if (end
<= zp
->z_size
) {
1471 zfs_range_unlock(rl
);
1475 tx
= dmu_tx_create(zfsvfs
->z_os
);
1476 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1477 zfs_sa_upgrade_txholds(tx
, zp
);
1478 if (end
> zp
->z_blksz
&&
1479 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zfsvfs
->z_max_blksz
)) {
1481 * We are growing the file past the current block size.
1483 if (zp
->z_blksz
> zp
->z_zfsvfs
->z_max_blksz
) {
1484 ASSERT(!ISP2(zp
->z_blksz
));
1485 newblksz
= MIN(end
, SPA_MAXBLOCKSIZE
);
1487 newblksz
= MIN(end
, zp
->z_zfsvfs
->z_max_blksz
);
1489 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1494 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1496 if (error
== ERESTART
) {
1502 zfs_range_unlock(rl
);
1507 zfs_grow_blocksize(zp
, newblksz
, tx
);
1511 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(zp
->z_zfsvfs
),
1512 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1514 zfs_range_unlock(rl
);
1522 * Free space in a file.
1524 * IN: zp - znode of file to free data in.
1525 * off - start of section to free.
1526 * len - length of section to free.
1528 * RETURN: 0 if success
1529 * error code if failure
1532 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1534 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1539 * Lock the range being freed.
1541 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1544 * Nothing to do if file already at desired length.
1546 if (off
>= zp
->z_size
) {
1547 zfs_range_unlock(rl
);
1551 if (off
+ len
> zp
->z_size
)
1552 len
= zp
->z_size
- off
;
1554 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, off
, len
);
1556 zfs_range_unlock(rl
);
1564 * IN: zp - znode of file to free data in.
1565 * end - new end-of-file.
1567 * RETURN: 0 if success
1568 * error code if failure
1571 zfs_trunc(znode_t
*zp
, uint64_t end
)
1573 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1574 vnode_t
*vp
= ZTOV(zp
);
1578 sa_bulk_attr_t bulk
[2];
1582 * We will change zp_size, lock the whole file.
1584 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1587 * Nothing to do if file already at desired length.
1589 if (end
>= zp
->z_size
) {
1590 zfs_range_unlock(rl
);
1594 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, end
, -1);
1596 zfs_range_unlock(rl
);
1600 tx
= dmu_tx_create(zfsvfs
->z_os
);
1601 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1602 zfs_sa_upgrade_txholds(tx
, zp
);
1603 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1605 if (error
== ERESTART
) {
1611 zfs_range_unlock(rl
);
1616 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
),
1617 NULL
, &zp
->z_size
, sizeof (zp
->z_size
));
1620 zp
->z_pflags
&= ~ZFS_SPARSE
;
1621 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1622 NULL
, &zp
->z_pflags
, 8);
1624 VERIFY(sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
) == 0);
1629 * Clear any mapped pages in the truncated region. This has to
1630 * happen outside of the transaction to avoid the possibility of
1631 * a deadlock with someone trying to push a page that we are
1632 * about to invalidate.
1634 if (vn_has_cached_data(vp
)) {
1636 uint64_t start
= end
& PAGEMASK
;
1637 int poff
= end
& PAGEOFFSET
;
1639 if (poff
!= 0 && (pp
= page_lookup(vp
, start
, SE_SHARED
))) {
1641 * We need to zero a partial page.
1643 pagezero(pp
, poff
, PAGESIZE
- poff
);
1647 error
= pvn_vplist_dirty(vp
, start
, zfs_no_putpage
,
1648 B_INVAL
| B_TRUNC
, NULL
);
1652 zfs_range_unlock(rl
);
1658 * Free space in a file
1660 * IN: zp - znode of file to free data in.
1661 * off - start of range
1662 * len - end of range (0 => EOF)
1663 * flag - current file open mode flags.
1664 * log - TRUE if this action should be logged
1666 * RETURN: 0 if success
1667 * error code if failure
1670 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1672 vnode_t
*vp
= ZTOV(zp
);
1674 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1675 zilog_t
*zilog
= zfsvfs
->z_log
;
1677 uint64_t mtime
[2], ctime
[2];
1678 sa_bulk_attr_t bulk
[3];
1682 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zfsvfs
), &mode
,
1683 sizeof (mode
))) != 0)
1686 if (off
> zp
->z_size
) {
1687 error
= zfs_extend(zp
, off
+len
);
1688 if (error
== 0 && log
)
1695 * Check for any locks in the region to be freed.
1698 if (MANDLOCK(vp
, (mode_t
)mode
)) {
1699 uint64_t length
= (len
? len
: zp
->z_size
- off
);
1700 if (error
= chklock(vp
, FWRITE
, off
, length
, flag
, NULL
))
1705 error
= zfs_trunc(zp
, off
);
1707 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1708 off
+ len
> zp
->z_size
)
1709 error
= zfs_extend(zp
, off
+len
);
1714 tx
= dmu_tx_create(zfsvfs
->z_os
);
1715 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1716 zfs_sa_upgrade_txholds(tx
, zp
);
1717 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1719 if (error
== ERESTART
) {
1728 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
, mtime
, 16);
1729 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
, ctime
, 16);
1730 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1731 NULL
, &zp
->z_pflags
, 8);
1732 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
1733 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1736 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1741 #endif /* HAVE_ZPL */
1744 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1746 uint64_t moid
, obj
, sa_obj
, version
;
1747 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1754 znode_t
*rootzp
= NULL
;
1758 zfs_acl_ids_t acl_ids
;
1763 #endif /* HAVE_ZPL */
1766 * First attempt to create master node.
1769 * In an empty objset, there are no blocks to read and thus
1770 * there can be no i/o errors (which we assert below).
1772 moid
= MASTER_NODE_OBJ
;
1773 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1774 DMU_OT_NONE
, 0, tx
);
1778 * Set starting attributes.
1780 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1782 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1783 /* For the moment we expect all zpl props to be uint64_ts */
1787 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1788 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1789 name
= nvpair_name(elem
);
1790 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1794 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1797 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1799 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1802 ASSERT(version
!= 0);
1803 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1806 * Create zap object used for SA attribute registration
1809 if (version
>= ZPL_VERSION_SA
) {
1810 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1811 DMU_OT_NONE
, 0, tx
);
1812 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1818 * Create a delete queue.
1820 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1822 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1827 * Create root znode. Create minimal znode/vnode/zfsvfs
1828 * to allow zfs_mknode to work.
1830 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
1831 vattr
.va_type
= VDIR
;
1832 vattr
.va_mode
= S_IFDIR
|0755;
1833 vattr
.va_uid
= crgetuid(cr
);
1834 vattr
.va_gid
= crgetgid(cr
);
1836 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1837 ASSERT(!POINTER_IS_VALID(rootzp
->z_zfsvfs
));
1838 rootzp
->z_moved
= 0;
1839 rootzp
->z_unlinked
= 0;
1840 rootzp
->z_atime_dirty
= 0;
1841 rootzp
->z_is_sa
= USE_SA(version
, os
);
1847 bzero(&zfsvfs
, sizeof (zfsvfs_t
));
1850 zfsvfs
.z_parent
= &zfsvfs
;
1851 zfsvfs
.z_version
= version
;
1852 zfsvfs
.z_use_fuids
= USE_FUIDS(version
, os
);
1853 zfsvfs
.z_use_sa
= USE_SA(version
, os
);
1854 zfsvfs
.z_norm
= norm
;
1856 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
1857 &zfsvfs
.z_attr_table
);
1862 * Fold case on file systems that are always or sometimes case
1865 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1866 zfsvfs
.z_norm
|= U8_TEXTPREP_TOUPPER
;
1868 /* XXX - This must be destroyed but I'm not quite sure yet so
1869 * I'm just annotating that fact when it's an issue. -Brian */
1870 mutex_init(&zfsvfs
.z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1871 list_create(&zfsvfs
.z_all_znodes
, sizeof (znode_t
),
1872 offsetof(znode_t
, z_link_node
));
1874 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1875 mutex_init(&zfsvfs
.z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1877 rootzp
->z_zfsvfs
= &zfsvfs
;
1878 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1879 cr
, NULL
, &acl_ids
));
1880 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1881 ASSERT3P(zp
, ==, rootzp
);
1882 ASSERT(!vn_in_dnlc(ZTOV(rootzp
))); /* not valid to move */
1883 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1885 zfs_acl_ids_free(&acl_ids
);
1886 POINTER_INVALIDATE(&rootzp
->z_zfsvfs
);
1888 ZTOV(rootzp
)->v_count
= 0;
1889 sa_handle_destroy(rootzp
->z_sa_hdl
);
1890 kmem_cache_free(znode_cache
, rootzp
);
1891 error
= zfs_create_share_dir(&zfsvfs
, tx
);
1893 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1894 mutex_destroy(&zfsvfs
.z_hold_mtx
[i
]);
1897 * Create root znode with code free of VFS dependencies
1899 obj
= zap_create_norm(os
, norm
, DMU_OT_DIRECTORY_CONTENTS
,
1900 DMU_OT_ZNODE
, sizeof (znode_phys_t
), tx
);
1902 VERIFY(0 == dmu_bonus_hold(os
, obj
, FTAG
, &db
));
1903 dmu_buf_will_dirty(db
, tx
);
1906 * Initialize the znode physical data to zero.
1908 ASSERT(db
->db_size
>= sizeof (znode_phys_t
));
1909 bzero(db
->db_data
, db
->db_size
);
1912 if (USE_FUIDS(version
, os
))
1913 pzp
->zp_flags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
1915 pzp
->zp_size
= 2; /* "." and ".." */
1917 pzp
->zp_parent
= obj
;
1918 pzp
->zp_gen
= dmu_tx_get_txg(tx
);
1919 pzp
->zp_mode
= S_IFDIR
| 0755;
1920 pzp
->zp_flags
= ZFS_ACL_TRIVIAL
;
1924 ZFS_TIME_ENCODE(&now
, pzp
->zp_crtime
);
1925 ZFS_TIME_ENCODE(&now
, pzp
->zp_ctime
);
1926 ZFS_TIME_ENCODE(&now
, pzp
->zp_atime
);
1927 ZFS_TIME_ENCODE(&now
, pzp
->zp_mtime
);
1929 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &obj
, tx
);
1932 dmu_buf_rele(db
, FTAG
);
1933 #endif /* HAVE_ZPL */
1936 #endif /* _KERNEL */
1939 zfs_sa_setup(objset_t
*osp
, sa_attr_type_t
**sa_table
)
1941 uint64_t sa_obj
= 0;
1944 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1945 if (error
!= 0 && error
!= ENOENT
)
1948 error
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
, sa_table
);
1953 zfs_grab_sa_handle(objset_t
*osp
, uint64_t obj
, sa_handle_t
**hdlp
,
1956 dmu_object_info_t doi
;
1959 if ((error
= sa_buf_hold(osp
, obj
, FTAG
, db
)) != 0)
1962 dmu_object_info_from_db(*db
, &doi
);
1963 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1964 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
1965 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1966 doi
.doi_bonus_size
< sizeof (znode_phys_t
))) {
1967 sa_buf_rele(*db
, FTAG
);
1971 error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
, hdlp
);
1973 sa_buf_rele(*db
, FTAG
);
1981 zfs_release_sa_handle(sa_handle_t
*hdl
, dmu_buf_t
*db
)
1983 sa_handle_destroy(hdl
);
1984 sa_buf_rele(db
, FTAG
);
1988 * Given an object number, return its parent object number and whether
1989 * or not the object is an extended attribute directory.
1992 zfs_obj_to_pobj(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
, uint64_t *pobjp
,
1998 sa_bulk_attr_t bulk
[3];
2002 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
], NULL
,
2003 &parent
, sizeof (parent
));
2004 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
2005 &pflags
, sizeof (pflags
));
2006 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
2007 &mode
, sizeof (mode
));
2009 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0)
2013 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
2019 * Given an object number, return some zpl level statistics
2022 zfs_obj_to_stats_impl(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
2025 sa_bulk_attr_t bulk
[4];
2028 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
2029 &sb
->zs_mode
, sizeof (sb
->zs_mode
));
2030 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_GEN
], NULL
,
2031 &sb
->zs_gen
, sizeof (sb
->zs_gen
));
2032 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_LINKS
], NULL
,
2033 &sb
->zs_links
, sizeof (sb
->zs_links
));
2034 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_CTIME
], NULL
,
2035 &sb
->zs_ctime
, sizeof (sb
->zs_ctime
));
2037 return (sa_bulk_lookup(hdl
, bulk
, count
));
2041 zfs_obj_to_path_impl(objset_t
*osp
, uint64_t obj
, sa_handle_t
*hdl
,
2042 sa_attr_type_t
*sa_table
, char *buf
, int len
)
2044 sa_handle_t
*sa_hdl
;
2045 sa_handle_t
*prevhdl
= NULL
;
2046 dmu_buf_t
*prevdb
= NULL
;
2047 dmu_buf_t
*sa_db
= NULL
;
2048 char *path
= buf
+ len
- 1;
2056 char component
[MAXNAMELEN
+ 2];
2061 zfs_release_sa_handle(prevhdl
, prevdb
);
2063 if ((error
= zfs_obj_to_pobj(sa_hdl
, sa_table
, &pobj
,
2064 &is_xattrdir
)) != 0)
2075 (void) sprintf(component
+ 1, "<xattrdir>");
2077 error
= zap_value_search(osp
, pobj
, obj
,
2078 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
2083 complen
= strlen(component
);
2085 ASSERT(path
>= buf
);
2086 bcopy(component
, path
, complen
);
2089 if (sa_hdl
!= hdl
) {
2093 error
= zfs_grab_sa_handle(osp
, obj
, &sa_hdl
, &sa_db
);
2101 if (sa_hdl
!= NULL
&& sa_hdl
!= hdl
) {
2102 ASSERT(sa_db
!= NULL
);
2103 zfs_release_sa_handle(sa_hdl
, sa_db
);
2107 (void) memmove(buf
, path
, buf
+ len
- path
);
2113 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
2115 sa_attr_type_t
*sa_table
;
2120 error
= zfs_sa_setup(osp
, &sa_table
);
2124 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
);
2128 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2130 zfs_release_sa_handle(hdl
, db
);
2135 zfs_obj_to_stats(objset_t
*osp
, uint64_t obj
, zfs_stat_t
*sb
,
2138 char *path
= buf
+ len
- 1;
2139 sa_attr_type_t
*sa_table
;
2146 error
= zfs_sa_setup(osp
, &sa_table
);
2150 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
);
2154 error
= zfs_obj_to_stats_impl(hdl
, sa_table
, sb
);
2156 zfs_release_sa_handle(hdl
, db
);
2160 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2162 zfs_release_sa_handle(hdl
, db
);
2166 #if defined(_KERNEL) && defined(HAVE_SPL)
2167 EXPORT_SYMBOL(zfs_create_fs
);
2168 EXPORT_SYMBOL(zfs_obj_to_path
);