4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
24 * Copyright (c) 2014 Integros [integros.com]
27 /* Portions Copyright 2007 Jeremy Teo */
30 #include <sys/types.h>
31 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
36 #include <sys/mntent.h>
37 #include <sys/mkdev.h>
38 #include <sys/u8_textprep.h>
39 #include <sys/dsl_dataset.h>
41 #include <sys/vfs_opreg.h>
42 #include <sys/vnode.h>
45 #include <sys/errno.h>
46 #include <sys/unistd.h>
48 #include <sys/atomic.h>
50 #include "fs/fs_subr.h"
51 #include <sys/zfs_dir.h>
52 #include <sys/zfs_acl.h>
53 #include <sys/zfs_ioctl.h>
54 #include <sys/zfs_rlock.h>
55 #include <sys/zfs_fuid.h>
56 #include <sys/dnode.h>
57 #include <sys/fs/zfs.h>
58 #include <sys/kidmap.h>
62 #include <sys/dmu_objset.h>
63 #include <sys/refcount.h>
66 #include <sys/zfs_znode.h>
68 #include <sys/zfs_sa.h>
69 #include <sys/zfs_stat.h>
72 #include "zfs_comutil.h"
75 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
76 * turned on when DEBUG is also defined.
83 #define ZNODE_STAT_ADD(stat) ((stat)++)
85 #define ZNODE_STAT_ADD(stat) /* nothing */
86 #endif /* ZNODE_STATS */
89 * Functions needed for userland (ie: libzpool) are not put under
90 * #ifdef_KERNEL; the rest of the functions have dependencies
91 * (such as VFS logic) that will not compile easily in userland.
95 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
96 * be freed before it can be safely accessed.
98 krwlock_t zfsvfs_lock
;
100 static kmem_cache_t
*znode_cache
= NULL
;
104 znode_evict_error(dmu_buf_t
*dbuf
, void *user_ptr
)
107 * We should never drop all dbuf refs without first clearing
108 * the eviction callback.
110 panic("evicting znode %p\n", user_ptr
);
115 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
119 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
121 zp
->z_vnode
= vn_alloc(kmflags
);
122 if (zp
->z_vnode
== NULL
) {
125 ZTOV(zp
)->v_data
= zp
;
127 list_link_init(&zp
->z_link_node
);
129 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
130 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
131 rw_init(&zp
->z_name_lock
, NULL
, RW_DEFAULT
, NULL
);
132 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
134 mutex_init(&zp
->z_range_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
135 avl_create(&zp
->z_range_avl
, zfs_range_compare
,
136 sizeof (rl_t
), offsetof(rl_t
, r_node
));
138 zp
->z_dirlocks
= NULL
;
139 zp
->z_acl_cached
= NULL
;
146 zfs_znode_cache_destructor(void *buf
, void *arg
)
150 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
151 ASSERT(ZTOV(zp
)->v_data
== zp
);
153 ASSERT(!list_link_active(&zp
->z_link_node
));
154 mutex_destroy(&zp
->z_lock
);
155 rw_destroy(&zp
->z_parent_lock
);
156 rw_destroy(&zp
->z_name_lock
);
157 mutex_destroy(&zp
->z_acl_lock
);
158 avl_destroy(&zp
->z_range_avl
);
159 mutex_destroy(&zp
->z_range_lock
);
161 ASSERT(zp
->z_dirlocks
== NULL
);
162 ASSERT(zp
->z_acl_cached
== NULL
);
167 uint64_t zms_zfsvfs_invalid
;
168 uint64_t zms_zfsvfs_recheck1
;
169 uint64_t zms_zfsvfs_unmounted
;
170 uint64_t zms_zfsvfs_recheck2
;
171 uint64_t zms_obj_held
;
172 uint64_t zms_vnode_locked
;
173 uint64_t zms_not_only_dnlc
;
175 #endif /* ZNODE_STATS */
178 zfs_znode_move_impl(znode_t
*ozp
, znode_t
*nzp
)
183 nzp
->z_zfsvfs
= ozp
->z_zfsvfs
;
187 nzp
->z_vnode
= ozp
->z_vnode
;
188 ozp
->z_vnode
= vp
; /* let destructor free the overwritten vnode */
189 ZTOV(ozp
)->v_data
= ozp
;
190 ZTOV(nzp
)->v_data
= nzp
;
192 nzp
->z_id
= ozp
->z_id
;
193 ASSERT(ozp
->z_dirlocks
== NULL
); /* znode not in use */
194 ASSERT(avl_numnodes(&ozp
->z_range_avl
) == 0);
195 nzp
->z_unlinked
= ozp
->z_unlinked
;
196 nzp
->z_atime_dirty
= ozp
->z_atime_dirty
;
197 nzp
->z_zn_prefetch
= ozp
->z_zn_prefetch
;
198 nzp
->z_blksz
= ozp
->z_blksz
;
199 nzp
->z_seq
= ozp
->z_seq
;
200 nzp
->z_mapcnt
= ozp
->z_mapcnt
;
201 nzp
->z_gen
= ozp
->z_gen
;
202 nzp
->z_sync_cnt
= ozp
->z_sync_cnt
;
203 nzp
->z_is_sa
= ozp
->z_is_sa
;
204 nzp
->z_sa_hdl
= ozp
->z_sa_hdl
;
205 bcopy(ozp
->z_atime
, nzp
->z_atime
, sizeof (uint64_t) * 2);
206 nzp
->z_links
= ozp
->z_links
;
207 nzp
->z_size
= ozp
->z_size
;
208 nzp
->z_pflags
= ozp
->z_pflags
;
209 nzp
->z_uid
= ozp
->z_uid
;
210 nzp
->z_gid
= ozp
->z_gid
;
211 nzp
->z_mode
= ozp
->z_mode
;
214 * Since this is just an idle znode and kmem is already dealing with
215 * memory pressure, release any cached ACL.
217 if (ozp
->z_acl_cached
) {
218 zfs_acl_free(ozp
->z_acl_cached
);
219 ozp
->z_acl_cached
= NULL
;
222 sa_set_userp(nzp
->z_sa_hdl
, nzp
);
225 * Invalidate the original znode by clearing fields that provide a
226 * pointer back to the znode. Set the low bit of the vfs pointer to
227 * ensure that zfs_znode_move() recognizes the znode as invalid in any
228 * subsequent callback.
230 ozp
->z_sa_hdl
= NULL
;
231 POINTER_INVALIDATE(&ozp
->z_zfsvfs
);
237 ozp
->z_moved
= (uint8_t)-1;
242 zfs_znode_move(void *buf
, void *newbuf
, size_t size
, void *arg
)
244 znode_t
*ozp
= buf
, *nzp
= newbuf
;
249 * The znode is on the file system's list of known znodes if the vfs
250 * pointer is valid. We set the low bit of the vfs pointer when freeing
251 * the znode to invalidate it, and the memory patterns written by kmem
252 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
253 * created znode sets the vfs pointer last of all to indicate that the
254 * znode is known and in a valid state to be moved by this function.
256 zfsvfs
= ozp
->z_zfsvfs
;
257 if (!POINTER_IS_VALID(zfsvfs
)) {
258 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_invalid
);
259 return (KMEM_CBRC_DONT_KNOW
);
263 * Close a small window in which it's possible that the filesystem could
264 * be unmounted and freed, and zfsvfs, though valid in the previous
265 * statement, could point to unrelated memory by the time we try to
266 * prevent the filesystem from being unmounted.
268 rw_enter(&zfsvfs_lock
, RW_WRITER
);
269 if (zfsvfs
!= ozp
->z_zfsvfs
) {
270 rw_exit(&zfsvfs_lock
);
271 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_recheck1
);
272 return (KMEM_CBRC_DONT_KNOW
);
276 * If the znode is still valid, then so is the file system. We know that
277 * no valid file system can be freed while we hold zfsvfs_lock, so we
278 * can safely ensure that the filesystem is not and will not be
279 * unmounted. The next statement is equivalent to ZFS_ENTER().
281 rrm_enter(&zfsvfs
->z_teardown_lock
, RW_READER
, FTAG
);
282 if (zfsvfs
->z_unmounted
) {
284 rw_exit(&zfsvfs_lock
);
285 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_unmounted
);
286 return (KMEM_CBRC_DONT_KNOW
);
288 rw_exit(&zfsvfs_lock
);
290 mutex_enter(&zfsvfs
->z_znodes_lock
);
292 * Recheck the vfs pointer in case the znode was removed just before
293 * acquiring the lock.
295 if (zfsvfs
!= ozp
->z_zfsvfs
) {
296 mutex_exit(&zfsvfs
->z_znodes_lock
);
298 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_recheck2
);
299 return (KMEM_CBRC_DONT_KNOW
);
303 * At this point we know that as long as we hold z_znodes_lock, the
304 * znode cannot be freed and fields within the znode can be safely
305 * accessed. Now, prevent a race with zfs_zget().
307 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs
, ozp
->z_id
) == 0) {
308 mutex_exit(&zfsvfs
->z_znodes_lock
);
310 ZNODE_STAT_ADD(znode_move_stats
.zms_obj_held
);
311 return (KMEM_CBRC_LATER
);
315 if (mutex_tryenter(&vp
->v_lock
) == 0) {
316 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
317 mutex_exit(&zfsvfs
->z_znodes_lock
);
319 ZNODE_STAT_ADD(znode_move_stats
.zms_vnode_locked
);
320 return (KMEM_CBRC_LATER
);
323 /* Only move znodes that are referenced _only_ by the DNLC. */
324 if (vp
->v_count
!= 1 || !vn_in_dnlc(vp
)) {
325 mutex_exit(&vp
->v_lock
);
326 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
327 mutex_exit(&zfsvfs
->z_znodes_lock
);
329 ZNODE_STAT_ADD(znode_move_stats
.zms_not_only_dnlc
);
330 return (KMEM_CBRC_LATER
);
334 * The znode is known and in a valid state to move. We're holding the
335 * locks needed to execute the critical section.
337 zfs_znode_move_impl(ozp
, nzp
);
338 mutex_exit(&vp
->v_lock
);
339 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
341 list_link_replace(&ozp
->z_link_node
, &nzp
->z_link_node
);
342 mutex_exit(&zfsvfs
->z_znodes_lock
);
345 return (KMEM_CBRC_YES
);
354 rw_init(&zfsvfs_lock
, NULL
, RW_DEFAULT
, NULL
);
355 ASSERT(znode_cache
== NULL
);
356 znode_cache
= kmem_cache_create("zfs_znode_cache",
357 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
358 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, 0);
359 kmem_cache_set_move(znode_cache
, zfs_znode_move
);
366 * Cleanup vfs & vnode ops
368 zfs_remove_op_tables();
374 kmem_cache_destroy(znode_cache
);
376 rw_destroy(&zfsvfs_lock
);
379 struct vnodeops
*zfs_dvnodeops
;
380 struct vnodeops
*zfs_fvnodeops
;
381 struct vnodeops
*zfs_symvnodeops
;
382 struct vnodeops
*zfs_xdvnodeops
;
383 struct vnodeops
*zfs_evnodeops
;
384 struct vnodeops
*zfs_sharevnodeops
;
387 zfs_remove_op_tables()
393 (void) vfs_freevfsops_by_type(zfsfstype
);
400 vn_freevnodeops(zfs_dvnodeops
);
402 vn_freevnodeops(zfs_fvnodeops
);
404 vn_freevnodeops(zfs_symvnodeops
);
406 vn_freevnodeops(zfs_xdvnodeops
);
408 vn_freevnodeops(zfs_evnodeops
);
409 if (zfs_sharevnodeops
)
410 vn_freevnodeops(zfs_sharevnodeops
);
412 zfs_dvnodeops
= NULL
;
413 zfs_fvnodeops
= NULL
;
414 zfs_symvnodeops
= NULL
;
415 zfs_xdvnodeops
= NULL
;
416 zfs_evnodeops
= NULL
;
417 zfs_sharevnodeops
= NULL
;
420 extern const fs_operation_def_t zfs_dvnodeops_template
[];
421 extern const fs_operation_def_t zfs_fvnodeops_template
[];
422 extern const fs_operation_def_t zfs_xdvnodeops_template
[];
423 extern const fs_operation_def_t zfs_symvnodeops_template
[];
424 extern const fs_operation_def_t zfs_evnodeops_template
[];
425 extern const fs_operation_def_t zfs_sharevnodeops_template
[];
428 zfs_create_op_tables()
433 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
434 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
435 * In this case we just return as the ops vectors are already set up.
440 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_dvnodeops_template
,
445 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_fvnodeops_template
,
450 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_symvnodeops_template
,
455 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_xdvnodeops_template
,
460 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_evnodeops_template
,
465 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_sharevnodeops_template
,
472 zfs_create_share_dir(zfsvfs_t
*zfsvfs
, dmu_tx_t
*tx
)
474 zfs_acl_ids_t acl_ids
;
481 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
482 vattr
.va_type
= VDIR
;
483 vattr
.va_mode
= S_IFDIR
|0555;
484 vattr
.va_uid
= crgetuid(kcred
);
485 vattr
.va_gid
= crgetgid(kcred
);
487 sharezp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
488 ASSERT(!POINTER_IS_VALID(sharezp
->z_zfsvfs
));
489 sharezp
->z_moved
= 0;
490 sharezp
->z_unlinked
= 0;
491 sharezp
->z_atime_dirty
= 0;
492 sharezp
->z_zfsvfs
= zfsvfs
;
493 sharezp
->z_is_sa
= zfsvfs
->z_use_sa
;
499 VERIFY(0 == zfs_acl_ids_create(sharezp
, IS_ROOT_NODE
, &vattr
,
500 kcred
, NULL
, &acl_ids
));
501 zfs_mknode(sharezp
, &vattr
, tx
, kcred
, IS_ROOT_NODE
, &zp
, &acl_ids
);
502 ASSERT3P(zp
, ==, sharezp
);
503 ASSERT(!vn_in_dnlc(ZTOV(sharezp
))); /* not valid to move */
504 POINTER_INVALIDATE(&sharezp
->z_zfsvfs
);
505 error
= zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
506 ZFS_SHARES_DIR
, 8, 1, &sharezp
->z_id
, tx
);
507 zfsvfs
->z_shares_dir
= sharezp
->z_id
;
509 zfs_acl_ids_free(&acl_ids
);
510 ZTOV(sharezp
)->v_count
= 0;
511 sa_handle_destroy(sharezp
->z_sa_hdl
);
512 kmem_cache_free(znode_cache
, sharezp
);
518 * define a couple of values we need available
519 * for both 64 and 32 bit environments.
522 #define NBITSMINOR64 32
525 #define MAXMAJ64 0xffffffffUL
528 #define MAXMIN64 0xffffffffUL
532 * Create special expldev for ZFS private use.
533 * Can't use standard expldev since it doesn't do
534 * what we want. The standard expldev() takes a
535 * dev32_t in LP64 and expands it to a long dev_t.
536 * We need an interface that takes a dev32_t in ILP32
537 * and expands it to a long dev_t.
540 zfs_expldev(dev_t dev
)
543 major_t major
= (major_t
)dev
>> NBITSMINOR32
& MAXMAJ32
;
544 return (((uint64_t)major
<< NBITSMINOR64
) |
545 ((minor_t
)dev
& MAXMIN32
));
552 * Special cmpldev for ZFS private use.
553 * Can't use standard cmpldev since it takes
554 * a long dev_t and compresses it to dev32_t in
555 * LP64. We need to do a compaction of a long dev_t
556 * to a dev32_t in ILP32.
559 zfs_cmpldev(uint64_t dev
)
562 minor_t minor
= (minor_t
)dev
& MAXMIN64
;
563 major_t major
= (major_t
)(dev
>> NBITSMINOR64
) & MAXMAJ64
;
565 if (major
> MAXMAJ32
|| minor
> MAXMIN32
)
568 return (((dev32_t
)major
<< NBITSMINOR32
) | minor
);
575 zfs_znode_sa_init(zfsvfs_t
*zfsvfs
, znode_t
*zp
,
576 dmu_buf_t
*db
, dmu_object_type_t obj_type
, sa_handle_t
*sa_hdl
)
578 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
) || (zfsvfs
== zp
->z_zfsvfs
));
579 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs
, zp
->z_id
)));
581 mutex_enter(&zp
->z_lock
);
583 ASSERT(zp
->z_sa_hdl
== NULL
);
584 ASSERT(zp
->z_acl_cached
== NULL
);
585 if (sa_hdl
== NULL
) {
586 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, zp
,
587 SA_HDL_SHARED
, &zp
->z_sa_hdl
));
589 zp
->z_sa_hdl
= sa_hdl
;
590 sa_set_userp(sa_hdl
, zp
);
593 zp
->z_is_sa
= (obj_type
== DMU_OT_SA
) ? B_TRUE
: B_FALSE
;
596 * Slap on VROOT if we are the root znode
598 if (zp
->z_id
== zfsvfs
->z_root
)
599 ZTOV(zp
)->v_flag
|= VROOT
;
601 mutex_exit(&zp
->z_lock
);
606 zfs_znode_dmu_fini(znode_t
*zp
)
608 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp
->z_zfsvfs
, zp
->z_id
)) ||
610 RW_WRITE_HELD(&zp
->z_zfsvfs
->z_teardown_inactive_lock
));
612 sa_handle_destroy(zp
->z_sa_hdl
);
617 * Construct a new znode/vnode and intialize.
619 * This does not do a call to dmu_set_user() that is
620 * up to the caller to do, in case you don't want to
624 zfs_znode_alloc(zfsvfs_t
*zfsvfs
, dmu_buf_t
*db
, int blksz
,
625 dmu_object_type_t obj_type
, sa_handle_t
*hdl
)
631 sa_bulk_attr_t bulk
[9];
634 zp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
636 ASSERT(zp
->z_dirlocks
== NULL
);
637 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
641 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
642 * the zfs_znode_move() callback.
646 zp
->z_atime_dirty
= 0;
648 zp
->z_id
= db
->db_object
;
650 zp
->z_seq
= 0x7A4653;
656 zfs_znode_sa_init(zfsvfs
, zp
, db
, obj_type
, hdl
);
658 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
, &mode
, 8);
659 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
, &zp
->z_gen
, 8);
660 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
662 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
664 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
666 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zfsvfs
), NULL
, &parent
, 8);
667 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
669 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
671 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
674 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
) != 0 || zp
->z_gen
== 0) {
676 sa_handle_destroy(zp
->z_sa_hdl
);
677 kmem_cache_free(znode_cache
, zp
);
682 vp
->v_vfsp
= zfsvfs
->z_parent
->z_vfs
;
684 vp
->v_type
= IFTOVT((mode_t
)mode
);
686 switch (vp
->v_type
) {
688 if (zp
->z_pflags
& ZFS_XATTR
) {
689 vn_setops(vp
, zfs_xdvnodeops
);
690 vp
->v_flag
|= V_XATTRDIR
;
692 vn_setops(vp
, zfs_dvnodeops
);
694 zp
->z_zn_prefetch
= B_TRUE
; /* z_prefetch default is enabled */
700 VERIFY(sa_lookup(zp
->z_sa_hdl
, SA_ZPL_RDEV(zfsvfs
),
701 &rdev
, sizeof (rdev
)) == 0);
703 vp
->v_rdev
= zfs_cmpldev(rdev
);
709 vn_setops(vp
, zfs_fvnodeops
);
712 vp
->v_flag
|= VMODSORT
;
713 if (parent
== zfsvfs
->z_shares_dir
) {
714 ASSERT(zp
->z_uid
== 0 && zp
->z_gid
== 0);
715 vn_setops(vp
, zfs_sharevnodeops
);
717 vn_setops(vp
, zfs_fvnodeops
);
721 vn_setops(vp
, zfs_symvnodeops
);
724 vn_setops(vp
, zfs_evnodeops
);
728 mutex_enter(&zfsvfs
->z_znodes_lock
);
729 list_insert_tail(&zfsvfs
->z_all_znodes
, zp
);
732 * Everything else must be valid before assigning z_zfsvfs makes the
733 * znode eligible for zfs_znode_move().
735 zp
->z_zfsvfs
= zfsvfs
;
736 mutex_exit(&zfsvfs
->z_znodes_lock
);
738 VFS_HOLD(zfsvfs
->z_vfs
);
742 static uint64_t empty_xattr
;
743 static uint64_t pad
[4];
744 static zfs_acl_phys_t acl_phys
;
746 * Create a new DMU object to hold a zfs znode.
748 * IN: dzp - parent directory for new znode
749 * vap - file attributes for new znode
750 * tx - dmu transaction id for zap operations
751 * cr - credentials of caller
753 * IS_ROOT_NODE - new object will be root
754 * IS_XATTR - new object is an attribute
755 * bonuslen - length of bonus buffer
756 * setaclp - File/Dir initial ACL
757 * fuidp - Tracks fuid allocation.
759 * OUT: zpp - allocated znode
763 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
764 uint_t flag
, znode_t
**zpp
, zfs_acl_ids_t
*acl_ids
)
766 uint64_t crtime
[2], atime
[2], mtime
[2], ctime
[2];
767 uint64_t mode
, size
, links
, parent
, pflags
;
768 uint64_t dzp_pflags
= 0;
770 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
776 dmu_object_type_t obj_type
;
777 sa_bulk_attr_t sa_attrs
[ZPL_END
];
779 zfs_acl_locator_cb_t locate
= { 0 };
781 ASSERT(vap
&& (vap
->va_mask
& (AT_TYPE
|AT_MODE
)) == (AT_TYPE
|AT_MODE
));
783 if (zfsvfs
->z_replay
) {
784 obj
= vap
->va_nodeid
;
785 now
= vap
->va_ctime
; /* see zfs_replay_create() */
786 gen
= vap
->va_nblocks
; /* ditto */
790 gen
= dmu_tx_get_txg(tx
);
793 obj_type
= zfsvfs
->z_use_sa
? DMU_OT_SA
: DMU_OT_ZNODE
;
794 bonuslen
= (obj_type
== DMU_OT_SA
) ?
795 DN_MAX_BONUSLEN
: ZFS_OLD_ZNODE_PHYS_SIZE
;
798 * Create a new DMU object.
801 * There's currently no mechanism for pre-reading the blocks that will
802 * be needed to allocate a new object, so we accept the small chance
803 * that there will be an i/o error and we will fail one of the
806 if (vap
->va_type
== VDIR
) {
807 if (zfsvfs
->z_replay
) {
808 VERIFY0(zap_create_claim_norm(zfsvfs
->z_os
, obj
,
809 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
810 obj_type
, bonuslen
, tx
));
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 VERIFY0(dmu_object_claim(zfsvfs
->z_os
, obj
,
819 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
820 obj_type
, bonuslen
, tx
));
822 obj
= dmu_object_alloc(zfsvfs
->z_os
,
823 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
824 obj_type
, bonuslen
, tx
);
828 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
829 VERIFY(0 == sa_buf_hold(zfsvfs
->z_os
, obj
, NULL
, &db
));
832 * If this is the root, fix up the half-initialized parent pointer
833 * to reference the just-allocated physical data area.
835 if (flag
& IS_ROOT_NODE
) {
838 dzp_pflags
= dzp
->z_pflags
;
842 * If parent is an xattr, so am I.
844 if (dzp_pflags
& ZFS_XATTR
) {
848 if (zfsvfs
->z_use_fuids
)
849 pflags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
853 if (vap
->va_type
== VDIR
) {
854 size
= 2; /* contents ("." and "..") */
855 links
= (flag
& (IS_ROOT_NODE
| IS_XATTR
)) ? 2 : 1;
860 if (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
) {
861 rdev
= zfs_expldev(vap
->va_rdev
);
865 mode
= acl_ids
->z_mode
;
870 * No execs denied will be deterimed when zfs_mode_compute() is called.
872 pflags
|= acl_ids
->z_aclp
->z_hints
&
873 (ZFS_ACL_TRIVIAL
|ZFS_INHERIT_ACE
|ZFS_ACL_AUTO_INHERIT
|
874 ZFS_ACL_DEFAULTED
|ZFS_ACL_PROTECTED
);
876 ZFS_TIME_ENCODE(&now
, crtime
);
877 ZFS_TIME_ENCODE(&now
, ctime
);
879 if (vap
->va_mask
& AT_ATIME
) {
880 ZFS_TIME_ENCODE(&vap
->va_atime
, atime
);
882 ZFS_TIME_ENCODE(&now
, atime
);
885 if (vap
->va_mask
& AT_MTIME
) {
886 ZFS_TIME_ENCODE(&vap
->va_mtime
, mtime
);
888 ZFS_TIME_ENCODE(&now
, mtime
);
891 /* Now add in all of the "SA" attributes */
892 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, NULL
, SA_HDL_SHARED
,
896 * Setup the array of attributes to be replaced/set on the new file
898 * order for DMU_OT_ZNODE is critical since it needs to be constructed
899 * in the old znode_phys_t format. Don't change this ordering
902 if (obj_type
== DMU_OT_ZNODE
) {
903 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
905 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
907 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
909 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
911 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
913 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
915 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
917 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
920 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
922 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
924 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
926 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
927 &acl_ids
->z_fuid
, 8);
928 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
929 &acl_ids
->z_fgid
, 8);
930 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
932 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
934 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
936 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
938 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
940 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
944 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zfsvfs
), NULL
, &links
, 8);
946 if (obj_type
== DMU_OT_ZNODE
) {
947 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zfsvfs
), NULL
,
950 if (obj_type
== DMU_OT_ZNODE
||
951 (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
)) {
952 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zfsvfs
),
956 if (obj_type
== DMU_OT_ZNODE
) {
957 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
959 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
960 &acl_ids
->z_fuid
, 8);
961 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
962 &acl_ids
->z_fgid
, 8);
963 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zfsvfs
), NULL
, pad
,
964 sizeof (uint64_t) * 4);
965 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zfsvfs
), NULL
,
966 &acl_phys
, sizeof (zfs_acl_phys_t
));
967 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
968 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zfsvfs
), NULL
,
969 &acl_ids
->z_aclp
->z_acl_count
, 8);
970 locate
.cb_aclp
= acl_ids
->z_aclp
;
971 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zfsvfs
),
972 zfs_acl_data_locator
, &locate
,
973 acl_ids
->z_aclp
->z_acl_bytes
);
974 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
975 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
978 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
980 if (!(flag
& IS_ROOT_NODE
)) {
981 *zpp
= zfs_znode_alloc(zfsvfs
, db
, 0, obj_type
, sa_hdl
);
982 ASSERT(*zpp
!= NULL
);
985 * If we are creating the root node, the "parent" we
986 * passed in is the znode for the root.
990 (*zpp
)->z_sa_hdl
= sa_hdl
;
993 (*zpp
)->z_pflags
= pflags
;
994 (*zpp
)->z_mode
= mode
;
996 if (vap
->va_mask
& AT_XVATTR
)
997 zfs_xvattr_set(*zpp
, (xvattr_t
*)vap
, tx
);
999 if (obj_type
== DMU_OT_ZNODE
||
1000 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
1001 VERIFY0(zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
));
1003 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1007 * Update in-core attributes. It is assumed the caller will be doing an
1008 * sa_bulk_update to push the changes out.
1011 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
1015 xoap
= xva_getxoptattr(xvap
);
1018 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
1020 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
1021 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(zp
->z_zfsvfs
),
1022 ×
, sizeof (times
), tx
);
1023 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
1025 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
1026 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
1028 XVA_SET_RTN(xvap
, XAT_READONLY
);
1030 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
1031 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
1033 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
1035 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
1036 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
1038 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
1040 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
1041 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
1043 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
1045 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
1046 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
1048 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
1050 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
1051 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
1053 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
1055 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
1056 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
1058 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
1060 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
1061 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
1063 XVA_SET_RTN(xvap
, XAT_NODUMP
);
1065 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
1066 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
1068 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
1070 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
1071 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
1072 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
1073 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
1075 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
1076 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
1078 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
1080 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
1081 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
1082 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
1084 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
1085 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
1087 XVA_SET_RTN(xvap
, XAT_REPARSE
);
1089 if (XVA_ISSET_REQ(xvap
, XAT_OFFLINE
)) {
1090 ZFS_ATTR_SET(zp
, ZFS_OFFLINE
, xoap
->xoa_offline
,
1092 XVA_SET_RTN(xvap
, XAT_OFFLINE
);
1094 if (XVA_ISSET_REQ(xvap
, XAT_SPARSE
)) {
1095 ZFS_ATTR_SET(zp
, ZFS_SPARSE
, xoap
->xoa_sparse
,
1097 XVA_SET_RTN(xvap
, XAT_SPARSE
);
1102 zfs_zget(zfsvfs_t
*zfsvfs
, uint64_t obj_num
, znode_t
**zpp
)
1104 dmu_object_info_t doi
;
1112 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1114 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1116 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1120 dmu_object_info_from_db(db
, &doi
);
1121 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1122 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1123 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1124 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1125 sa_buf_rele(db
, NULL
);
1126 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1127 return (SET_ERROR(EINVAL
));
1130 hdl
= dmu_buf_get_user(db
);
1132 zp
= sa_get_userdata(hdl
);
1136 * Since "SA" does immediate eviction we
1137 * should never find a sa handle that doesn't
1138 * know about the znode.
1141 ASSERT3P(zp
, !=, NULL
);
1143 mutex_enter(&zp
->z_lock
);
1144 ASSERT3U(zp
->z_id
, ==, obj_num
);
1145 if (zp
->z_unlinked
) {
1146 err
= SET_ERROR(ENOENT
);
1152 mutex_exit(&zp
->z_lock
);
1153 sa_buf_rele(db
, NULL
);
1154 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1159 * Not found create new znode/vnode
1160 * but only if file exists.
1162 * There is a small window where zfs_vget() could
1163 * find this object while a file create is still in
1164 * progress. This is checked for in zfs_znode_alloc()
1166 * if zfs_znode_alloc() fails it will drop the hold on the
1169 zp
= zfs_znode_alloc(zfsvfs
, db
, doi
.doi_data_block_size
,
1170 doi
.doi_bonus_type
, NULL
);
1172 err
= SET_ERROR(ENOENT
);
1176 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1181 zfs_rezget(znode_t
*zp
)
1183 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1184 dmu_object_info_t doi
;
1186 uint64_t obj_num
= zp
->z_id
;
1188 sa_bulk_attr_t bulk
[8];
1193 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1195 mutex_enter(&zp
->z_acl_lock
);
1196 if (zp
->z_acl_cached
) {
1197 zfs_acl_free(zp
->z_acl_cached
);
1198 zp
->z_acl_cached
= NULL
;
1201 mutex_exit(&zp
->z_acl_lock
);
1202 ASSERT(zp
->z_sa_hdl
== NULL
);
1203 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1205 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1209 dmu_object_info_from_db(db
, &doi
);
1210 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1211 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1212 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1213 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1214 sa_buf_rele(db
, NULL
);
1215 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1216 return (SET_ERROR(EINVAL
));
1219 zfs_znode_sa_init(zfsvfs
, zp
, db
, doi
.doi_bonus_type
, NULL
);
1221 /* reload cached values */
1222 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
,
1223 &gen
, sizeof (gen
));
1224 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
1225 &zp
->z_size
, sizeof (zp
->z_size
));
1226 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
1227 &zp
->z_links
, sizeof (zp
->z_links
));
1228 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
1229 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1230 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
1231 &zp
->z_atime
, sizeof (zp
->z_atime
));
1232 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
1233 &zp
->z_uid
, sizeof (zp
->z_uid
));
1234 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
1235 &zp
->z_gid
, sizeof (zp
->z_gid
));
1236 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
,
1237 &mode
, sizeof (mode
));
1239 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
1240 zfs_znode_dmu_fini(zp
);
1241 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1242 return (SET_ERROR(EIO
));
1247 if (gen
!= zp
->z_gen
) {
1248 zfs_znode_dmu_fini(zp
);
1249 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1250 return (SET_ERROR(EIO
));
1253 zp
->z_blksz
= doi
.doi_data_block_size
;
1256 * If the file has zero links, then it has been unlinked on the send
1257 * side and it must be in the received unlinked set.
1258 * We call zfs_znode_dmu_fini() now to prevent any accesses to the
1259 * stale data and to prevent automatical removal of the file in
1260 * zfs_zinactive(). The file will be removed either when it is removed
1261 * on the send side and the next incremental stream is received or
1262 * when the unlinked set gets processed.
1264 zp
->z_unlinked
= (zp
->z_links
== 0);
1266 zfs_znode_dmu_fini(zp
);
1268 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1274 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1276 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1277 objset_t
*os
= zfsvfs
->z_os
;
1278 uint64_t obj
= zp
->z_id
;
1279 uint64_t acl_obj
= zfs_external_acl(zp
);
1281 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
1283 VERIFY(!zp
->z_is_sa
);
1284 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1286 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1287 zfs_znode_dmu_fini(zp
);
1288 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1293 zfs_zinactive(znode_t
*zp
)
1295 vnode_t
*vp
= ZTOV(zp
);
1296 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1297 uint64_t z_id
= zp
->z_id
;
1299 ASSERT(zp
->z_sa_hdl
);
1302 * Don't allow a zfs_zget() while were trying to release this znode
1304 ZFS_OBJ_HOLD_ENTER(zfsvfs
, z_id
);
1306 mutex_enter(&zp
->z_lock
);
1307 mutex_enter(&vp
->v_lock
);
1309 if (vp
->v_count
> 0 || vn_has_cached_data(vp
)) {
1311 * If the hold count is greater than zero, somebody has
1312 * obtained a new reference on this znode while we were
1313 * processing it here, so we are done. If we still have
1314 * mapped pages then we are also done, since we don't
1315 * want to inactivate the znode until the pages get pushed.
1317 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1318 * this seems like it would leave the znode hanging with
1319 * no chance to go inactive...
1321 mutex_exit(&vp
->v_lock
);
1322 mutex_exit(&zp
->z_lock
);
1323 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1326 mutex_exit(&vp
->v_lock
);
1329 * If this was the last reference to a file with no links, remove
1330 * the file from the file system unless the file system is mounted
1331 * read-only. That can happen, for example, if the file system was
1332 * originally read-write, the file was opened, then unlinked and
1333 * the file system was made read-only before the file was finally
1334 * closed. The file will remain in the unlinked set.
1336 if (zp
->z_unlinked
) {
1337 ASSERT(!zfsvfs
->z_issnap
);
1338 if ((zfsvfs
->z_vfs
->vfs_flag
& VFS_RDONLY
) == 0) {
1339 mutex_exit(&zp
->z_lock
);
1340 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1346 mutex_exit(&zp
->z_lock
);
1347 zfs_znode_dmu_fini(zp
);
1348 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1353 zfs_znode_free(znode_t
*zp
)
1355 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1357 vn_invalid(ZTOV(zp
));
1359 ASSERT(ZTOV(zp
)->v_count
== 0);
1361 mutex_enter(&zfsvfs
->z_znodes_lock
);
1362 POINTER_INVALIDATE(&zp
->z_zfsvfs
);
1363 list_remove(&zfsvfs
->z_all_znodes
, zp
);
1364 mutex_exit(&zfsvfs
->z_znodes_lock
);
1366 if (zp
->z_acl_cached
) {
1367 zfs_acl_free(zp
->z_acl_cached
);
1368 zp
->z_acl_cached
= NULL
;
1371 kmem_cache_free(znode_cache
, zp
);
1373 VFS_RELE(zfsvfs
->z_vfs
);
1377 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1378 uint64_t ctime
[2], boolean_t have_tx
)
1384 if (have_tx
) { /* will sa_bulk_update happen really soon? */
1385 zp
->z_atime_dirty
= 0;
1388 zp
->z_atime_dirty
= 1;
1391 if (flag
& AT_ATIME
) {
1392 ZFS_TIME_ENCODE(&now
, zp
->z_atime
);
1395 if (flag
& AT_MTIME
) {
1396 ZFS_TIME_ENCODE(&now
, mtime
);
1397 if (zp
->z_zfsvfs
->z_use_fuids
) {
1398 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1403 if (flag
& AT_CTIME
) {
1404 ZFS_TIME_ENCODE(&now
, ctime
);
1405 if (zp
->z_zfsvfs
->z_use_fuids
)
1406 zp
->z_pflags
|= ZFS_ARCHIVE
;
1411 * Grow the block size for a file.
1413 * IN: zp - znode of file to free data in.
1414 * size - requested block size
1415 * tx - open transaction.
1417 * NOTE: this function assumes that the znode is write locked.
1420 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1425 if (size
<= zp
->z_blksz
)
1428 * If the file size is already greater than the current blocksize,
1429 * we will not grow. If there is more than one block in a file,
1430 * the blocksize cannot change.
1432 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1435 error
= dmu_object_set_blocksize(zp
->z_zfsvfs
->z_os
, zp
->z_id
,
1438 if (error
== ENOTSUP
)
1442 /* What blocksize did we actually get? */
1443 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1447 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1448 * be calling back into the fs for a putpage(). E.g.: when truncating
1449 * a file, the pages being "thrown away* don't need to be written out.
1453 zfs_no_putpage(vnode_t
*vp
, page_t
*pp
, u_offset_t
*offp
, size_t *lenp
,
1454 int flags
, cred_t
*cr
)
1461 * Increase the file length
1463 * IN: zp - znode of file to free data in.
1464 * end - new end-of-file
1466 * RETURN: 0 on success, error code on failure
1469 zfs_extend(znode_t
*zp
, uint64_t end
)
1471 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1478 * We will change zp_size, lock the whole file.
1480 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1483 * Nothing to do if file already at desired length.
1485 if (end
<= zp
->z_size
) {
1486 zfs_range_unlock(rl
);
1489 tx
= dmu_tx_create(zfsvfs
->z_os
);
1490 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1491 zfs_sa_upgrade_txholds(tx
, zp
);
1492 if (end
> zp
->z_blksz
&&
1493 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zfsvfs
->z_max_blksz
)) {
1495 * We are growing the file past the current block size.
1497 if (zp
->z_blksz
> zp
->z_zfsvfs
->z_max_blksz
) {
1499 * File's blocksize is already larger than the
1500 * "recordsize" property. Only let it grow to
1501 * the next power of 2.
1503 ASSERT(!ISP2(zp
->z_blksz
));
1504 newblksz
= MIN(end
, 1 << highbit64(zp
->z_blksz
));
1506 newblksz
= MIN(end
, zp
->z_zfsvfs
->z_max_blksz
);
1508 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1513 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1516 zfs_range_unlock(rl
);
1521 zfs_grow_blocksize(zp
, newblksz
, tx
);
1525 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(zp
->z_zfsvfs
),
1526 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1528 zfs_range_unlock(rl
);
1536 * Free space in a file.
1538 * IN: zp - znode of file to free data in.
1539 * off - start of section to free.
1540 * len - length of section to free.
1542 * RETURN: 0 on success, error code on failure
1545 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1547 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1552 * Lock the range being freed.
1554 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1557 * Nothing to do if file already at desired length.
1559 if (off
>= zp
->z_size
) {
1560 zfs_range_unlock(rl
);
1564 if (off
+ len
> zp
->z_size
)
1565 len
= zp
->z_size
- off
;
1567 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, off
, len
);
1569 zfs_range_unlock(rl
);
1577 * IN: zp - znode of file to free data in.
1578 * end - new end-of-file.
1580 * RETURN: 0 on success, error code on failure
1583 zfs_trunc(znode_t
*zp
, uint64_t end
)
1585 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1586 vnode_t
*vp
= ZTOV(zp
);
1590 sa_bulk_attr_t bulk
[2];
1594 * We will change zp_size, lock the whole file.
1596 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1599 * Nothing to do if file already at desired length.
1601 if (end
>= zp
->z_size
) {
1602 zfs_range_unlock(rl
);
1606 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, end
, -1);
1608 zfs_range_unlock(rl
);
1611 tx
= dmu_tx_create(zfsvfs
->z_os
);
1612 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1613 zfs_sa_upgrade_txholds(tx
, zp
);
1614 dmu_tx_mark_netfree(tx
);
1615 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1618 zfs_range_unlock(rl
);
1623 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
),
1624 NULL
, &zp
->z_size
, sizeof (zp
->z_size
));
1627 zp
->z_pflags
&= ~ZFS_SPARSE
;
1628 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1629 NULL
, &zp
->z_pflags
, 8);
1631 VERIFY(sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
) == 0);
1636 * Clear any mapped pages in the truncated region. This has to
1637 * happen outside of the transaction to avoid the possibility of
1638 * a deadlock with someone trying to push a page that we are
1639 * about to invalidate.
1641 if (vn_has_cached_data(vp
)) {
1643 uint64_t start
= end
& PAGEMASK
;
1644 int poff
= end
& PAGEOFFSET
;
1646 if (poff
!= 0 && (pp
= page_lookup(vp
, start
, SE_SHARED
))) {
1648 * We need to zero a partial page.
1650 pagezero(pp
, poff
, PAGESIZE
- poff
);
1654 error
= pvn_vplist_dirty(vp
, start
, zfs_no_putpage
,
1655 B_INVAL
| B_TRUNC
, NULL
);
1659 zfs_range_unlock(rl
);
1665 * Free space in a file
1667 * IN: zp - znode of file to free data in.
1668 * off - start of range
1669 * len - end of range (0 => EOF)
1670 * flag - current file open mode flags.
1671 * log - TRUE if this action should be logged
1673 * RETURN: 0 on success, error code on failure
1676 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1678 vnode_t
*vp
= ZTOV(zp
);
1680 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1681 zilog_t
*zilog
= zfsvfs
->z_log
;
1683 uint64_t mtime
[2], ctime
[2];
1684 sa_bulk_attr_t bulk
[3];
1688 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zfsvfs
), &mode
,
1689 sizeof (mode
))) != 0)
1692 if (off
> zp
->z_size
) {
1693 error
= zfs_extend(zp
, off
+len
);
1694 if (error
== 0 && log
)
1701 * Check for any locks in the region to be freed.
1704 if (MANDLOCK(vp
, (mode_t
)mode
)) {
1705 uint64_t length
= (len
? len
: zp
->z_size
- off
);
1706 if (error
= chklock(vp
, FWRITE
, off
, length
, flag
, NULL
))
1711 error
= zfs_trunc(zp
, off
);
1713 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1714 off
+ len
> zp
->z_size
)
1715 error
= zfs_extend(zp
, off
+len
);
1720 tx
= dmu_tx_create(zfsvfs
->z_os
);
1721 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1722 zfs_sa_upgrade_txholds(tx
, zp
);
1723 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1729 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
, mtime
, 16);
1730 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
, ctime
, 16);
1731 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1732 NULL
, &zp
->z_pflags
, 8);
1733 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
1734 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1737 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
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
;
1752 znode_t
*rootzp
= NULL
;
1757 zfs_acl_ids_t acl_ids
;
1760 * First attempt to create master node.
1763 * In an empty objset, there are no blocks to read and thus
1764 * there can be no i/o errors (which we assert below).
1766 moid
= MASTER_NODE_OBJ
;
1767 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1768 DMU_OT_NONE
, 0, tx
);
1772 * Set starting attributes.
1774 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1776 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1777 /* For the moment we expect all zpl props to be uint64_ts */
1781 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1782 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1783 name
= nvpair_name(elem
);
1784 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1788 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1791 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1793 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1796 ASSERT(version
!= 0);
1797 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1800 * Create zap object used for SA attribute registration
1803 if (version
>= ZPL_VERSION_SA
) {
1804 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1805 DMU_OT_NONE
, 0, tx
);
1806 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1812 * Create a delete queue.
1814 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1816 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1820 * Create root znode. Create minimal znode/vnode/zfsvfs
1821 * to allow zfs_mknode to work.
1823 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
1824 vattr
.va_type
= VDIR
;
1825 vattr
.va_mode
= S_IFDIR
|0755;
1826 vattr
.va_uid
= crgetuid(cr
);
1827 vattr
.va_gid
= crgetgid(cr
);
1829 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1830 ASSERT(!POINTER_IS_VALID(rootzp
->z_zfsvfs
));
1831 rootzp
->z_moved
= 0;
1832 rootzp
->z_unlinked
= 0;
1833 rootzp
->z_atime_dirty
= 0;
1834 rootzp
->z_is_sa
= USE_SA(version
, os
);
1840 zfsvfs
= kmem_zalloc(sizeof (zfsvfs_t
), KM_SLEEP
);
1842 zfsvfs
->z_parent
= zfsvfs
;
1843 zfsvfs
->z_version
= version
;
1844 zfsvfs
->z_use_fuids
= USE_FUIDS(version
, os
);
1845 zfsvfs
->z_use_sa
= USE_SA(version
, os
);
1846 zfsvfs
->z_norm
= norm
;
1848 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
1849 &zfsvfs
->z_attr_table
);
1854 * Fold case on file systems that are always or sometimes case
1857 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1858 zfsvfs
->z_norm
|= U8_TEXTPREP_TOUPPER
;
1860 mutex_init(&zfsvfs
->z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1861 list_create(&zfsvfs
->z_all_znodes
, sizeof (znode_t
),
1862 offsetof(znode_t
, z_link_node
));
1864 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1865 mutex_init(&zfsvfs
->z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1867 rootzp
->z_zfsvfs
= zfsvfs
;
1868 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1869 cr
, NULL
, &acl_ids
));
1870 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1871 ASSERT3P(zp
, ==, rootzp
);
1872 ASSERT(!vn_in_dnlc(ZTOV(rootzp
))); /* not valid to move */
1873 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1875 zfs_acl_ids_free(&acl_ids
);
1876 POINTER_INVALIDATE(&rootzp
->z_zfsvfs
);
1878 ZTOV(rootzp
)->v_count
= 0;
1879 sa_handle_destroy(rootzp
->z_sa_hdl
);
1880 kmem_cache_free(znode_cache
, rootzp
);
1883 * Create shares directory
1886 error
= zfs_create_share_dir(zfsvfs
, tx
);
1890 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1891 mutex_destroy(&zfsvfs
->z_hold_mtx
[i
]);
1892 kmem_free(zfsvfs
, sizeof (zfsvfs_t
));
1895 #endif /* _KERNEL */
1898 zfs_sa_setup(objset_t
*osp
, sa_attr_type_t
**sa_table
)
1900 uint64_t sa_obj
= 0;
1903 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1904 if (error
!= 0 && error
!= ENOENT
)
1907 error
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
, sa_table
);
1912 zfs_grab_sa_handle(objset_t
*osp
, uint64_t obj
, sa_handle_t
**hdlp
,
1913 dmu_buf_t
**db
, void *tag
)
1915 dmu_object_info_t doi
;
1918 if ((error
= sa_buf_hold(osp
, obj
, tag
, db
)) != 0)
1921 dmu_object_info_from_db(*db
, &doi
);
1922 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1923 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
1924 doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1925 doi
.doi_bonus_size
< sizeof (znode_phys_t
)) {
1926 sa_buf_rele(*db
, tag
);
1927 return (SET_ERROR(ENOTSUP
));
1930 error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
, hdlp
);
1932 sa_buf_rele(*db
, tag
);
1940 zfs_release_sa_handle(sa_handle_t
*hdl
, dmu_buf_t
*db
, void *tag
)
1942 sa_handle_destroy(hdl
);
1943 sa_buf_rele(db
, tag
);
1947 * Given an object number, return its parent object number and whether
1948 * or not the object is an extended attribute directory.
1951 zfs_obj_to_pobj(objset_t
*osp
, sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
1952 uint64_t *pobjp
, int *is_xattrdir
)
1957 uint64_t parent_mode
;
1958 sa_bulk_attr_t bulk
[3];
1959 sa_handle_t
*sa_hdl
;
1964 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
], NULL
,
1965 &parent
, sizeof (parent
));
1966 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
1967 &pflags
, sizeof (pflags
));
1968 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1969 &mode
, sizeof (mode
));
1971 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0)
1975 * When a link is removed its parent pointer is not changed and will
1976 * be invalid. There are two cases where a link is removed but the
1977 * file stays around, when it goes to the delete queue and when there
1978 * are additional links.
1980 error
= zfs_grab_sa_handle(osp
, parent
, &sa_hdl
, &sa_db
, FTAG
);
1984 error
= sa_lookup(sa_hdl
, ZPL_MODE
, &parent_mode
, sizeof (parent_mode
));
1985 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
1989 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
1992 * Extended attributes can be applied to files, directories, etc.
1993 * Otherwise the parent must be a directory.
1995 if (!*is_xattrdir
&& !S_ISDIR(parent_mode
))
1996 return (SET_ERROR(EINVAL
));
2004 * Given an object number, return some zpl level statistics
2007 zfs_obj_to_stats_impl(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
2010 sa_bulk_attr_t bulk
[4];
2013 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
2014 &sb
->zs_mode
, sizeof (sb
->zs_mode
));
2015 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_GEN
], NULL
,
2016 &sb
->zs_gen
, sizeof (sb
->zs_gen
));
2017 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_LINKS
], NULL
,
2018 &sb
->zs_links
, sizeof (sb
->zs_links
));
2019 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_CTIME
], NULL
,
2020 &sb
->zs_ctime
, sizeof (sb
->zs_ctime
));
2022 return (sa_bulk_lookup(hdl
, bulk
, count
));
2026 zfs_obj_to_path_impl(objset_t
*osp
, uint64_t obj
, sa_handle_t
*hdl
,
2027 sa_attr_type_t
*sa_table
, char *buf
, int len
)
2029 sa_handle_t
*sa_hdl
;
2030 sa_handle_t
*prevhdl
= NULL
;
2031 dmu_buf_t
*prevdb
= NULL
;
2032 dmu_buf_t
*sa_db
= NULL
;
2033 char *path
= buf
+ len
- 1;
2041 char component
[MAXNAMELEN
+ 2];
2046 zfs_release_sa_handle(prevhdl
, prevdb
, FTAG
);
2048 if ((error
= zfs_obj_to_pobj(osp
, sa_hdl
, sa_table
, &pobj
,
2049 &is_xattrdir
)) != 0)
2060 (void) sprintf(component
+ 1, "<xattrdir>");
2062 error
= zap_value_search(osp
, pobj
, obj
,
2063 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
2068 complen
= strlen(component
);
2070 ASSERT(path
>= buf
);
2071 bcopy(component
, path
, complen
);
2074 if (sa_hdl
!= hdl
) {
2078 error
= zfs_grab_sa_handle(osp
, obj
, &sa_hdl
, &sa_db
, FTAG
);
2086 if (sa_hdl
!= NULL
&& sa_hdl
!= hdl
) {
2087 ASSERT(sa_db
!= NULL
);
2088 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
2092 (void) memmove(buf
, path
, buf
+ len
- path
);
2098 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
2100 sa_attr_type_t
*sa_table
;
2105 error
= zfs_sa_setup(osp
, &sa_table
);
2109 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
2113 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2115 zfs_release_sa_handle(hdl
, db
, FTAG
);
2120 zfs_obj_to_stats(objset_t
*osp
, uint64_t obj
, zfs_stat_t
*sb
,
2123 char *path
= buf
+ len
- 1;
2124 sa_attr_type_t
*sa_table
;
2131 error
= zfs_sa_setup(osp
, &sa_table
);
2135 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
2139 error
= zfs_obj_to_stats_impl(hdl
, sa_table
, sb
);
2141 zfs_release_sa_handle(hdl
, db
, FTAG
);
2145 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2147 zfs_release_sa_handle(hdl
, db
, FTAG
);