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 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
26 /* Portions Copyright 2007 Jeremy Teo */
29 #include <sys/types.h>
30 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/sysmacros.h>
34 #include <sys/resource.h>
35 #include <sys/mntent.h>
36 #include <sys/mkdev.h>
37 #include <sys/u8_textprep.h>
38 #include <sys/dsl_dataset.h>
40 #include <sys/vfs_opreg.h>
41 #include <sys/vnode.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
47 #include <sys/atomic.h>
49 #include "fs/fs_subr.h"
50 #include <sys/zfs_dir.h>
51 #include <sys/zfs_acl.h>
52 #include <sys/zfs_ioctl.h>
53 #include <sys/zfs_rlock.h>
54 #include <sys/zfs_fuid.h>
55 #include <sys/fs/zfs.h>
56 #include <sys/kidmap.h>
60 #include <sys/refcount.h>
63 #include <sys/zfs_znode.h>
68 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
69 * turned on when DEBUG is also defined.
76 #define ZNODE_STAT_ADD(stat) ((stat)++)
78 #define ZNODE_STAT_ADD(stat) /* nothing */
79 #endif /* ZNODE_STATS */
81 #define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3))
82 #define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1))
85 * Functions needed for userland (ie: libzpool) are not put under
86 * #ifdef_KERNEL; the rest of the functions have dependencies
87 * (such as VFS logic) that will not compile easily in userland.
90 static kmem_cache_t
*znode_cache
= NULL
;
94 znode_evict_error(dmu_buf_t
*dbuf
, void *user_ptr
)
97 * We should never drop all dbuf refs without first clearing
98 * the eviction callback.
100 panic("evicting znode %p\n", user_ptr
);
105 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
109 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
111 zp
->z_vnode
= vn_alloc(kmflags
);
112 if (zp
->z_vnode
== NULL
) {
115 ZTOV(zp
)->v_data
= zp
;
117 list_link_init(&zp
->z_link_node
);
119 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
120 rw_init(&zp
->z_map_lock
, NULL
, RW_DEFAULT
, NULL
);
121 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
122 rw_init(&zp
->z_name_lock
, NULL
, RW_DEFAULT
, NULL
);
123 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
125 mutex_init(&zp
->z_range_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
126 avl_create(&zp
->z_range_avl
, zfs_range_compare
,
127 sizeof (rl_t
), offsetof(rl_t
, r_node
));
130 zp
->z_dirlocks
= NULL
;
136 zfs_znode_cache_destructor(void *buf
, void *arg
)
140 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
141 ASSERT(ZTOV(zp
)->v_data
== zp
);
143 ASSERT(!list_link_active(&zp
->z_link_node
));
144 mutex_destroy(&zp
->z_lock
);
145 rw_destroy(&zp
->z_map_lock
);
146 rw_destroy(&zp
->z_parent_lock
);
147 rw_destroy(&zp
->z_name_lock
);
148 mutex_destroy(&zp
->z_acl_lock
);
149 avl_destroy(&zp
->z_range_avl
);
150 mutex_destroy(&zp
->z_range_lock
);
152 ASSERT(zp
->z_dbuf
== NULL
);
153 ASSERT(zp
->z_dirlocks
== NULL
);
158 uint64_t zms_zfsvfs_invalid
;
159 uint64_t zms_zfsvfs_unmounted
;
160 uint64_t zms_zfsvfs_recheck_invalid
;
161 uint64_t zms_obj_held
;
162 uint64_t zms_vnode_locked
;
163 uint64_t zms_not_only_dnlc
;
165 #endif /* ZNODE_STATS */
168 zfs_znode_move_impl(znode_t
*ozp
, znode_t
*nzp
)
173 nzp
->z_zfsvfs
= ozp
->z_zfsvfs
;
177 nzp
->z_vnode
= ozp
->z_vnode
;
178 ozp
->z_vnode
= vp
; /* let destructor free the overwritten vnode */
179 ZTOV(ozp
)->v_data
= ozp
;
180 ZTOV(nzp
)->v_data
= nzp
;
182 nzp
->z_id
= ozp
->z_id
;
183 ASSERT(ozp
->z_dirlocks
== NULL
); /* znode not in use */
184 ASSERT(avl_numnodes(&ozp
->z_range_avl
) == 0);
185 nzp
->z_unlinked
= ozp
->z_unlinked
;
186 nzp
->z_atime_dirty
= ozp
->z_atime_dirty
;
187 nzp
->z_zn_prefetch
= ozp
->z_zn_prefetch
;
188 nzp
->z_blksz
= ozp
->z_blksz
;
189 nzp
->z_seq
= ozp
->z_seq
;
190 nzp
->z_mapcnt
= ozp
->z_mapcnt
;
191 nzp
->z_last_itx
= ozp
->z_last_itx
;
192 nzp
->z_gen
= ozp
->z_gen
;
193 nzp
->z_sync_cnt
= ozp
->z_sync_cnt
;
194 nzp
->z_phys
= ozp
->z_phys
;
195 nzp
->z_dbuf
= ozp
->z_dbuf
;
197 /* Update back pointers. */
198 (void) dmu_buf_update_user(nzp
->z_dbuf
, ozp
, nzp
, &nzp
->z_phys
,
202 * Invalidate the original znode by clearing fields that provide a
203 * pointer back to the znode. Set the low bit of the vfs pointer to
204 * ensure that zfs_znode_move() recognizes the znode as invalid in any
205 * subsequent callback.
208 POINTER_INVALIDATE(&ozp
->z_zfsvfs
);
212 * Wrapper function for ZFS_ENTER that returns 0 if successful and otherwise
213 * returns a non-zero error code.
216 zfs_enter(zfsvfs_t
*zfsvfs
)
224 zfs_znode_move(void *buf
, void *newbuf
, size_t size
, void *arg
)
226 znode_t
*ozp
= buf
, *nzp
= newbuf
;
231 * The znode is on the file system's list of known znodes if the vfs
232 * pointer is valid. We set the low bit of the vfs pointer when freeing
233 * the znode to invalidate it, and the memory patterns written by kmem
234 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
235 * created znode sets the vfs pointer last of all to indicate that the
236 * znode is known and in a valid state to be moved by this function.
238 zfsvfs
= ozp
->z_zfsvfs
;
239 if (!POINTER_IS_VALID(zfsvfs
)) {
240 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_invalid
);
241 return (KMEM_CBRC_DONT_KNOW
);
245 * Ensure that the filesystem is not unmounted during the move.
247 if (zfs_enter(zfsvfs
) != 0) { /* ZFS_ENTER */
248 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_unmounted
);
249 return (KMEM_CBRC_DONT_KNOW
);
252 mutex_enter(&zfsvfs
->z_znodes_lock
);
254 * Recheck the vfs pointer in case the znode was removed just before
255 * acquiring the lock.
257 if (zfsvfs
!= ozp
->z_zfsvfs
) {
258 mutex_exit(&zfsvfs
->z_znodes_lock
);
260 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_recheck_invalid
);
261 return (KMEM_CBRC_DONT_KNOW
);
265 * At this point we know that as long as we hold z_znodes_lock, the
266 * znode cannot be freed and fields within the znode can be safely
267 * accessed. Now, prevent a race with zfs_zget().
269 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs
, ozp
->z_id
) == 0) {
270 mutex_exit(&zfsvfs
->z_znodes_lock
);
272 ZNODE_STAT_ADD(znode_move_stats
.zms_obj_held
);
273 return (KMEM_CBRC_LATER
);
277 if (mutex_tryenter(&vp
->v_lock
) == 0) {
278 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
279 mutex_exit(&zfsvfs
->z_znodes_lock
);
281 ZNODE_STAT_ADD(znode_move_stats
.zms_vnode_locked
);
282 return (KMEM_CBRC_LATER
);
285 /* Only move znodes that are referenced _only_ by the DNLC. */
286 if (vp
->v_count
!= 1 || !vn_in_dnlc(vp
)) {
287 mutex_exit(&vp
->v_lock
);
288 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
289 mutex_exit(&zfsvfs
->z_znodes_lock
);
291 ZNODE_STAT_ADD(znode_move_stats
.zms_not_only_dnlc
);
292 return (KMEM_CBRC_LATER
);
296 * The znode is known and in a valid state to move. We're holding the
297 * locks needed to execute the critical section.
299 zfs_znode_move_impl(ozp
, nzp
);
300 mutex_exit(&vp
->v_lock
);
301 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
303 list_link_replace(&ozp
->z_link_node
, &nzp
->z_link_node
);
304 mutex_exit(&zfsvfs
->z_znodes_lock
);
307 return (KMEM_CBRC_YES
);
316 ASSERT(znode_cache
== NULL
);
317 znode_cache
= kmem_cache_create("zfs_znode_cache",
318 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
319 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, 0);
320 kmem_cache_set_move(znode_cache
, zfs_znode_move
);
327 * Cleanup vfs & vnode ops
329 zfs_remove_op_tables();
335 kmem_cache_destroy(znode_cache
);
339 struct vnodeops
*zfs_dvnodeops
;
340 struct vnodeops
*zfs_fvnodeops
;
341 struct vnodeops
*zfs_symvnodeops
;
342 struct vnodeops
*zfs_xdvnodeops
;
343 struct vnodeops
*zfs_evnodeops
;
346 zfs_remove_op_tables()
352 (void) vfs_freevfsops_by_type(zfsfstype
);
359 vn_freevnodeops(zfs_dvnodeops
);
361 vn_freevnodeops(zfs_fvnodeops
);
363 vn_freevnodeops(zfs_symvnodeops
);
365 vn_freevnodeops(zfs_xdvnodeops
);
367 vn_freevnodeops(zfs_evnodeops
);
369 zfs_dvnodeops
= NULL
;
370 zfs_fvnodeops
= NULL
;
371 zfs_symvnodeops
= NULL
;
372 zfs_xdvnodeops
= NULL
;
373 zfs_evnodeops
= NULL
;
376 extern const fs_operation_def_t zfs_dvnodeops_template
[];
377 extern const fs_operation_def_t zfs_fvnodeops_template
[];
378 extern const fs_operation_def_t zfs_xdvnodeops_template
[];
379 extern const fs_operation_def_t zfs_symvnodeops_template
[];
380 extern const fs_operation_def_t zfs_evnodeops_template
[];
383 zfs_create_op_tables()
388 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
389 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
390 * In this case we just return as the ops vectors are already set up.
395 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_dvnodeops_template
,
400 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_fvnodeops_template
,
405 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_symvnodeops_template
,
410 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_xdvnodeops_template
,
415 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_evnodeops_template
,
422 * zfs_init_fs - Initialize the zfsvfs struct and the file system
423 * incore "master" object. Verify version compatibility.
426 zfs_init_fs(zfsvfs_t
*zfsvfs
, znode_t
**zpp
)
428 extern int zfsfstype
;
430 objset_t
*os
= zfsvfs
->z_os
;
437 error
= zfs_get_zplprop(os
, ZFS_PROP_VERSION
, &zfsvfs
->z_version
);
440 } else if (zfsvfs
->z_version
> ZPL_VERSION
) {
441 (void) printf("Mismatched versions: File system "
442 "is version %llu on-disk format, which is "
443 "incompatible with this software version %lld!",
444 (u_longlong_t
)zfsvfs
->z_version
, ZPL_VERSION
);
448 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_NORMALIZE
, &zval
)) != 0)
450 zfsvfs
->z_norm
= (int)zval
;
451 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_UTF8ONLY
, &zval
)) != 0)
453 zfsvfs
->z_utf8
= (zval
!= 0);
454 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_CASE
, &zval
)) != 0)
456 zfsvfs
->z_case
= (uint_t
)zval
;
458 * Fold case on file systems that are always or sometimes case
461 if (zfsvfs
->z_case
== ZFS_CASE_INSENSITIVE
||
462 zfsvfs
->z_case
== ZFS_CASE_MIXED
)
463 zfsvfs
->z_norm
|= U8_TEXTPREP_TOUPPER
;
466 * The fsid is 64 bits, composed of an 8-bit fs type, which
467 * separates our fsid from any other filesystem types, and a
468 * 56-bit objset unique ID. The objset unique ID is unique to
469 * all objsets open on this system, provided by unique_create().
470 * The 8-bit fs type must be put in the low bits of fsid[1]
471 * because that's where other Solaris filesystems put it.
473 fsid_guid
= dmu_objset_fsid_guid(os
);
474 ASSERT((fsid_guid
& ~((1ULL<<56)-1)) == 0);
475 zfsvfs
->z_vfs
->vfs_fsid
.val
[0] = fsid_guid
;
476 zfsvfs
->z_vfs
->vfs_fsid
.val
[1] = ((fsid_guid
>>32) << 8) |
479 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_ROOT_OBJ
, 8, 1,
483 ASSERT(zfsvfs
->z_root
!= 0);
485 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_UNLINKED_SET
, 8, 1,
486 &zfsvfs
->z_unlinkedobj
);
491 * Initialize zget mutex's
493 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
494 mutex_init(&zfsvfs
->z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
496 error
= zfs_zget(zfsvfs
, zfsvfs
->z_root
, zpp
);
499 * On error, we destroy the mutexes here since it's not
500 * possible for the caller to determine if the mutexes were
501 * initialized properly.
503 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
504 mutex_destroy(&zfsvfs
->z_hold_mtx
[i
]);
507 ASSERT3U((*zpp
)->z_id
, ==, zfsvfs
->z_root
);
508 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_FUID_TABLES
, 8, 1,
509 &zfsvfs
->z_fuid_obj
);
517 * define a couple of values we need available
518 * for both 64 and 32 bit environments.
521 #define NBITSMINOR64 32
524 #define MAXMAJ64 0xffffffffUL
527 #define MAXMIN64 0xffffffffUL
531 * Create special expldev for ZFS private use.
532 * Can't use standard expldev since it doesn't do
533 * what we want. The standard expldev() takes a
534 * dev32_t in LP64 and expands it to a long dev_t.
535 * We need an interface that takes a dev32_t in ILP32
536 * and expands it to a long dev_t.
539 zfs_expldev(dev_t dev
)
542 major_t major
= (major_t
)dev
>> NBITSMINOR32
& MAXMAJ32
;
543 return (((uint64_t)major
<< NBITSMINOR64
) |
544 ((minor_t
)dev
& MAXMIN32
));
551 * Special cmpldev for ZFS private use.
552 * Can't use standard cmpldev since it takes
553 * a long dev_t and compresses it to dev32_t in
554 * LP64. We need to do a compaction of a long dev_t
555 * to a dev32_t in ILP32.
558 zfs_cmpldev(uint64_t dev
)
561 minor_t minor
= (minor_t
)dev
& MAXMIN64
;
562 major_t major
= (major_t
)(dev
>> NBITSMINOR64
) & MAXMAJ64
;
564 if (major
> MAXMAJ32
|| minor
> MAXMIN32
)
567 return (((dev32_t
)major
<< NBITSMINOR32
) | minor
);
574 zfs_znode_dmu_init(zfsvfs_t
*zfsvfs
, znode_t
*zp
, dmu_buf_t
*db
)
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_dbuf
== NULL
);
585 nzp
= dmu_buf_set_user_ie(db
, zp
, &zp
->z_phys
, znode_evict_error
);
589 * concurrent zgets on this object.
592 panic("existing znode %p for dbuf %p", (void *)nzp
, (void *)db
);
595 * Slap on VROOT if we are the root znode
597 if (zp
->z_id
== zfsvfs
->z_root
)
598 ZTOV(zp
)->v_flag
|= VROOT
;
600 mutex_exit(&zp
->z_lock
);
605 zfs_znode_dmu_fini(znode_t
*zp
)
607 dmu_buf_t
*db
= zp
->z_dbuf
;
608 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp
->z_zfsvfs
, zp
->z_id
)) ||
610 RW_WRITE_HELD(&zp
->z_zfsvfs
->z_teardown_inactive_lock
));
611 ASSERT(zp
->z_dbuf
!= NULL
);
613 VERIFY(zp
== dmu_buf_update_user(db
, zp
, NULL
, NULL
, NULL
));
614 dmu_buf_rele(db
, NULL
);
618 * Construct a new znode/vnode and intialize.
620 * This does not do a call to dmu_set_user() that is
621 * up to the caller to do, in case you don't want to
625 zfs_znode_alloc(zfsvfs_t
*zfsvfs
, dmu_buf_t
*db
, int blksz
)
630 zp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
632 ASSERT(zp
->z_dirlocks
== NULL
);
633 ASSERT(zp
->z_dbuf
== NULL
);
634 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
637 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
638 * the zfs_znode_move() callback.
642 zp
->z_atime_dirty
= 0;
645 zp
->z_id
= db
->db_object
;
647 zp
->z_seq
= 0x7A4653;
653 zfs_znode_dmu_init(zfsvfs
, zp
, db
);
655 zp
->z_gen
= zp
->z_phys
->zp_gen
;
657 vp
->v_vfsp
= zfsvfs
->z_parent
->z_vfs
;
658 vp
->v_type
= IFTOVT((mode_t
)zp
->z_phys
->zp_mode
);
660 switch (vp
->v_type
) {
662 if (zp
->z_phys
->zp_flags
& ZFS_XATTR
) {
663 vn_setops(vp
, zfs_xdvnodeops
);
664 vp
->v_flag
|= V_XATTRDIR
;
666 vn_setops(vp
, zfs_dvnodeops
);
668 zp
->z_zn_prefetch
= B_TRUE
; /* z_prefetch default is enabled */
672 vp
->v_rdev
= zfs_cmpldev(zp
->z_phys
->zp_rdev
);
677 vn_setops(vp
, zfs_fvnodeops
);
680 vp
->v_flag
|= VMODSORT
;
681 vn_setops(vp
, zfs_fvnodeops
);
684 vn_setops(vp
, zfs_symvnodeops
);
687 vn_setops(vp
, zfs_evnodeops
);
691 mutex_enter(&zfsvfs
->z_znodes_lock
);
692 list_insert_tail(&zfsvfs
->z_all_znodes
, zp
);
695 * Everything else must be valid before assigning z_zfsvfs makes the
696 * znode eligible for zfs_znode_move().
698 zp
->z_zfsvfs
= zfsvfs
;
699 mutex_exit(&zfsvfs
->z_znodes_lock
);
701 VFS_HOLD(zfsvfs
->z_vfs
);
706 * Create a new DMU object to hold a zfs znode.
708 * IN: dzp - parent directory for new znode
709 * vap - file attributes for new znode
710 * tx - dmu transaction id for zap operations
711 * cr - credentials of caller
713 * IS_ROOT_NODE - new object will be root
714 * IS_XATTR - new object is an attribute
715 * IS_REPLAY - intent log replay
716 * bonuslen - length of bonus buffer
717 * setaclp - File/Dir initial ACL
718 * fuidp - Tracks fuid allocation.
720 * OUT: zpp - allocated znode
724 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
725 uint_t flag
, znode_t
**zpp
, int bonuslen
, zfs_acl_t
*setaclp
,
726 zfs_fuid_info_t
**fuidp
)
730 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
735 ASSERT(vap
&& (vap
->va_mask
& (AT_TYPE
|AT_MODE
)) == (AT_TYPE
|AT_MODE
));
737 if (zfsvfs
->z_assign
>= TXG_INITIAL
) { /* ZIL replay */
738 obj
= vap
->va_nodeid
;
740 now
= vap
->va_ctime
; /* see zfs_replay_create() */
741 gen
= vap
->va_nblocks
; /* ditto */
745 gen
= dmu_tx_get_txg(tx
);
749 * Create a new DMU object.
752 * There's currently no mechanism for pre-reading the blocks that will
753 * be to needed allocate a new object, so we accept the small chance
754 * that there will be an i/o error and we will fail one of the
757 if (vap
->va_type
== VDIR
) {
758 if (flag
& IS_REPLAY
) {
759 err
= zap_create_claim_norm(zfsvfs
->z_os
, obj
,
760 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
761 DMU_OT_ZNODE
, sizeof (znode_phys_t
) + bonuslen
, tx
);
762 ASSERT3U(err
, ==, 0);
764 obj
= zap_create_norm(zfsvfs
->z_os
,
765 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
766 DMU_OT_ZNODE
, sizeof (znode_phys_t
) + bonuslen
, tx
);
769 if (flag
& IS_REPLAY
) {
770 err
= dmu_object_claim(zfsvfs
->z_os
, obj
,
771 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
772 DMU_OT_ZNODE
, sizeof (znode_phys_t
) + bonuslen
, tx
);
773 ASSERT3U(err
, ==, 0);
775 obj
= dmu_object_alloc(zfsvfs
->z_os
,
776 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
777 DMU_OT_ZNODE
, sizeof (znode_phys_t
) + bonuslen
, tx
);
780 VERIFY(0 == dmu_bonus_hold(zfsvfs
->z_os
, obj
, NULL
, &db
));
781 dmu_buf_will_dirty(db
, tx
);
784 * Initialize the znode physical data to zero.
786 ASSERT(db
->db_size
>= sizeof (znode_phys_t
));
787 bzero(db
->db_data
, db
->db_size
);
791 * If this is the root, fix up the half-initialized parent pointer
792 * to reference the just-allocated physical data area.
794 if (flag
& IS_ROOT_NODE
) {
801 * If parent is an xattr, so am I.
803 if (dzp
->z_phys
->zp_flags
& ZFS_XATTR
)
806 if (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
) {
807 pzp
->zp_rdev
= zfs_expldev(vap
->va_rdev
);
810 if (zfsvfs
->z_use_fuids
)
811 pzp
->zp_flags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
813 if (vap
->va_type
== VDIR
) {
814 pzp
->zp_size
= 2; /* contents ("." and "..") */
815 pzp
->zp_links
= (flag
& (IS_ROOT_NODE
| IS_XATTR
)) ? 2 : 1;
818 pzp
->zp_parent
= dzp
->z_id
;
820 pzp
->zp_flags
|= ZFS_XATTR
;
824 ZFS_TIME_ENCODE(&now
, pzp
->zp_crtime
);
825 ZFS_TIME_ENCODE(&now
, pzp
->zp_ctime
);
827 if (vap
->va_mask
& AT_ATIME
) {
828 ZFS_TIME_ENCODE(&vap
->va_atime
, pzp
->zp_atime
);
830 ZFS_TIME_ENCODE(&now
, pzp
->zp_atime
);
833 if (vap
->va_mask
& AT_MTIME
) {
834 ZFS_TIME_ENCODE(&vap
->va_mtime
, pzp
->zp_mtime
);
836 ZFS_TIME_ENCODE(&now
, pzp
->zp_mtime
);
839 pzp
->zp_mode
= MAKEIMODE(vap
->va_type
, vap
->va_mode
);
840 if (!(flag
& IS_ROOT_NODE
)) {
841 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
842 *zpp
= zfs_znode_alloc(zfsvfs
, db
, 0);
843 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
846 * If we are creating the root node, the "parent" we
847 * passed in is the znode for the root.
851 zfs_perm_init(*zpp
, dzp
, flag
, vap
, tx
, cr
, setaclp
, fuidp
);
855 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
)
859 xoap
= xva_getxoptattr(xvap
);
862 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
863 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, zp
->z_phys
->zp_crtime
);
864 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
866 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
867 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
);
868 XVA_SET_RTN(xvap
, XAT_READONLY
);
870 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
871 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
);
872 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
874 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
875 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
);
876 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
878 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
879 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
);
880 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
882 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
883 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
);
884 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
886 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
887 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
);
888 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
890 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
891 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
);
892 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
894 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
895 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
);
896 XVA_SET_RTN(xvap
, XAT_NODUMP
);
898 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
899 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
);
900 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
902 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
903 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
904 xoap
->xoa_av_quarantined
);
905 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
907 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
908 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
);
909 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
911 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
912 (void) memcpy(zp
->z_phys
+ 1, xoap
->xoa_av_scanstamp
,
913 sizeof (xoap
->xoa_av_scanstamp
));
914 zp
->z_phys
->zp_flags
|= ZFS_BONUS_SCANSTAMP
;
915 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
920 zfs_zget(zfsvfs_t
*zfsvfs
, uint64_t obj_num
, znode_t
**zpp
)
922 dmu_object_info_t doi
;
929 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
931 err
= dmu_bonus_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
933 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
937 dmu_object_info_from_db(db
, &doi
);
938 if (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
939 doi
.doi_bonus_size
< sizeof (znode_phys_t
)) {
940 dmu_buf_rele(db
, NULL
);
941 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
945 zp
= dmu_buf_get_user(db
);
947 mutex_enter(&zp
->z_lock
);
950 * Since we do immediate eviction of the z_dbuf, we
951 * should never find a dbuf with a znode that doesn't
952 * know about the dbuf.
954 ASSERT3P(zp
->z_dbuf
, ==, db
);
955 ASSERT3U(zp
->z_id
, ==, obj_num
);
956 if (zp
->z_unlinked
) {
963 dmu_buf_rele(db
, NULL
);
964 mutex_exit(&zp
->z_lock
);
965 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
970 * Not found create new znode/vnode
972 zp
= zfs_znode_alloc(zfsvfs
, db
, doi
.doi_data_block_size
);
973 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
979 zfs_rezget(znode_t
*zp
)
981 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
982 dmu_object_info_t doi
;
984 uint64_t obj_num
= zp
->z_id
;
987 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
989 err
= dmu_bonus_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
991 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
995 dmu_object_info_from_db(db
, &doi
);
996 if (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
997 doi
.doi_bonus_size
< sizeof (znode_phys_t
)) {
998 dmu_buf_rele(db
, NULL
);
999 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1003 if (((znode_phys_t
*)db
->db_data
)->zp_gen
!= zp
->z_gen
) {
1004 dmu_buf_rele(db
, NULL
);
1005 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1009 zfs_znode_dmu_init(zfsvfs
, zp
, db
);
1010 zp
->z_unlinked
= (zp
->z_phys
->zp_links
== 0);
1011 zp
->z_blksz
= doi
.doi_data_block_size
;
1013 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1019 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1021 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1022 objset_t
*os
= zfsvfs
->z_os
;
1023 uint64_t obj
= zp
->z_id
;
1024 uint64_t acl_obj
= zp
->z_phys
->zp_acl
.z_acl_extern_obj
;
1026 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
1028 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1029 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1030 zfs_znode_dmu_fini(zp
);
1031 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1036 zfs_zinactive(znode_t
*zp
)
1038 vnode_t
*vp
= ZTOV(zp
);
1039 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1040 uint64_t z_id
= zp
->z_id
;
1042 ASSERT(zp
->z_dbuf
&& zp
->z_phys
);
1045 * Don't allow a zfs_zget() while were trying to release this znode
1047 ZFS_OBJ_HOLD_ENTER(zfsvfs
, z_id
);
1049 mutex_enter(&zp
->z_lock
);
1050 mutex_enter(&vp
->v_lock
);
1052 if (vp
->v_count
> 0 || vn_has_cached_data(vp
)) {
1054 * If the hold count is greater than zero, somebody has
1055 * obtained a new reference on this znode while we were
1056 * processing it here, so we are done. If we still have
1057 * mapped pages then we are also done, since we don't
1058 * want to inactivate the znode until the pages get pushed.
1060 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1061 * this seems like it would leave the znode hanging with
1062 * no chance to go inactive...
1064 mutex_exit(&vp
->v_lock
);
1065 mutex_exit(&zp
->z_lock
);
1066 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1069 mutex_exit(&vp
->v_lock
);
1072 * If this was the last reference to a file with no links,
1073 * remove the file from the file system.
1075 if (zp
->z_unlinked
) {
1076 mutex_exit(&zp
->z_lock
);
1077 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1081 mutex_exit(&zp
->z_lock
);
1082 zfs_znode_dmu_fini(zp
);
1083 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1088 zfs_znode_free(znode_t
*zp
)
1090 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1092 vn_invalid(ZTOV(zp
));
1094 ASSERT(ZTOV(zp
)->v_count
== 0);
1096 mutex_enter(&zfsvfs
->z_znodes_lock
);
1097 POINTER_INVALIDATE(&zp
->z_zfsvfs
);
1098 list_remove(&zfsvfs
->z_all_znodes
, zp
);
1099 mutex_exit(&zfsvfs
->z_znodes_lock
);
1101 kmem_cache_free(znode_cache
, zp
);
1103 VFS_RELE(zfsvfs
->z_vfs
);
1107 zfs_time_stamper_locked(znode_t
*zp
, uint_t flag
, dmu_tx_t
*tx
)
1111 ASSERT(MUTEX_HELD(&zp
->z_lock
));
1116 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
1117 zp
->z_atime_dirty
= 0;
1120 zp
->z_atime_dirty
= 1;
1123 if (flag
& AT_ATIME
)
1124 ZFS_TIME_ENCODE(&now
, zp
->z_phys
->zp_atime
);
1126 if (flag
& AT_MTIME
) {
1127 ZFS_TIME_ENCODE(&now
, zp
->z_phys
->zp_mtime
);
1128 if (zp
->z_zfsvfs
->z_use_fuids
)
1129 zp
->z_phys
->zp_flags
|= (ZFS_ARCHIVE
| ZFS_AV_MODIFIED
);
1132 if (flag
& AT_CTIME
) {
1133 ZFS_TIME_ENCODE(&now
, zp
->z_phys
->zp_ctime
);
1134 if (zp
->z_zfsvfs
->z_use_fuids
)
1135 zp
->z_phys
->zp_flags
|= ZFS_ARCHIVE
;
1140 * Update the requested znode timestamps with the current time.
1141 * If we are in a transaction, then go ahead and mark the znode
1142 * dirty in the transaction so the timestamps will go to disk.
1143 * Otherwise, we will get pushed next time the znode is updated
1144 * in a transaction, or when this znode eventually goes inactive.
1147 * 1 - Only the ACCESS time is ever updated outside of a transaction.
1148 * 2 - Multiple consecutive updates will be collapsed into a single
1149 * znode update by the transaction grouping semantics of the DMU.
1152 zfs_time_stamper(znode_t
*zp
, uint_t flag
, dmu_tx_t
*tx
)
1154 mutex_enter(&zp
->z_lock
);
1155 zfs_time_stamper_locked(zp
, flag
, tx
);
1156 mutex_exit(&zp
->z_lock
);
1160 * Grow the block size for a file.
1162 * IN: zp - znode of file to free data in.
1163 * size - requested block size
1164 * tx - open transaction.
1166 * NOTE: this function assumes that the znode is write locked.
1169 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1174 if (size
<= zp
->z_blksz
)
1177 * If the file size is already greater than the current blocksize,
1178 * we will not grow. If there is more than one block in a file,
1179 * the blocksize cannot change.
1181 if (zp
->z_blksz
&& zp
->z_phys
->zp_size
> zp
->z_blksz
)
1184 error
= dmu_object_set_blocksize(zp
->z_zfsvfs
->z_os
, zp
->z_id
,
1186 if (error
== ENOTSUP
)
1188 ASSERT3U(error
, ==, 0);
1190 /* What blocksize did we actually get? */
1191 dmu_object_size_from_db(zp
->z_dbuf
, &zp
->z_blksz
, &dummy
);
1195 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1196 * be calling back into the fs for a putpage(). E.g.: when truncating
1197 * a file, the pages being "thrown away* don't need to be written out.
1201 zfs_no_putpage(vnode_t
*vp
, page_t
*pp
, u_offset_t
*offp
, size_t *lenp
,
1202 int flags
, cred_t
*cr
)
1209 * Increase the file length
1211 * IN: zp - znode of file to free data in.
1212 * end - new end-of-file
1214 * RETURN: 0 if success
1215 * error code if failure
1218 zfs_extend(znode_t
*zp
, uint64_t end
)
1220 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1227 * We will change zp_size, lock the whole file.
1229 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1232 * Nothing to do if file already at desired length.
1234 if (end
<= zp
->z_phys
->zp_size
) {
1235 zfs_range_unlock(rl
);
1239 tx
= dmu_tx_create(zfsvfs
->z_os
);
1240 dmu_tx_hold_bonus(tx
, zp
->z_id
);
1241 if (end
> zp
->z_blksz
&&
1242 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zfsvfs
->z_max_blksz
)) {
1244 * We are growing the file past the current block size.
1246 if (zp
->z_blksz
> zp
->z_zfsvfs
->z_max_blksz
) {
1247 ASSERT(!ISP2(zp
->z_blksz
));
1248 newblksz
= MIN(end
, SPA_MAXBLOCKSIZE
);
1250 newblksz
= MIN(end
, zp
->z_zfsvfs
->z_max_blksz
);
1252 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1257 error
= dmu_tx_assign(tx
, zfsvfs
->z_assign
);
1259 if (error
== ERESTART
&& zfsvfs
->z_assign
== TXG_NOWAIT
) {
1265 zfs_range_unlock(rl
);
1268 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
1271 zfs_grow_blocksize(zp
, newblksz
, tx
);
1273 zp
->z_phys
->zp_size
= end
;
1275 zfs_range_unlock(rl
);
1283 * Free space in a file.
1285 * IN: zp - znode of file to free data in.
1286 * off - start of section to free.
1287 * len - length of section to free.
1289 * RETURN: 0 if success
1290 * error code if failure
1293 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1295 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1300 * Lock the range being freed.
1302 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1305 * Nothing to do if file already at desired length.
1307 if (off
>= zp
->z_phys
->zp_size
) {
1308 zfs_range_unlock(rl
);
1312 if (off
+ len
> zp
->z_phys
->zp_size
)
1313 len
= zp
->z_phys
->zp_size
- off
;
1315 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, off
, len
);
1317 zfs_range_unlock(rl
);
1325 * IN: zp - znode of file to free data in.
1326 * end - new end-of-file.
1328 * RETURN: 0 if success
1329 * error code if failure
1332 zfs_trunc(znode_t
*zp
, uint64_t end
)
1334 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1335 vnode_t
*vp
= ZTOV(zp
);
1341 * We will change zp_size, lock the whole file.
1343 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1346 * Nothing to do if file already at desired length.
1348 if (end
>= zp
->z_phys
->zp_size
) {
1349 zfs_range_unlock(rl
);
1353 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, end
, -1);
1355 zfs_range_unlock(rl
);
1359 tx
= dmu_tx_create(zfsvfs
->z_os
);
1360 dmu_tx_hold_bonus(tx
, zp
->z_id
);
1361 error
= dmu_tx_assign(tx
, zfsvfs
->z_assign
);
1363 if (error
== ERESTART
&& zfsvfs
->z_assign
== TXG_NOWAIT
) {
1369 zfs_range_unlock(rl
);
1372 dmu_buf_will_dirty(zp
->z_dbuf
, tx
);
1374 zp
->z_phys
->zp_size
= end
;
1378 zfs_range_unlock(rl
);
1381 * Clear any mapped pages in the truncated region. This has to
1382 * happen outside of the transaction to avoid the possibility of
1383 * a deadlock with someone trying to push a page that we are
1384 * about to invalidate.
1386 rw_enter(&zp
->z_map_lock
, RW_WRITER
);
1387 if (vn_has_cached_data(vp
)) {
1389 uint64_t start
= end
& PAGEMASK
;
1390 int poff
= end
& PAGEOFFSET
;
1392 if (poff
!= 0 && (pp
= page_lookup(vp
, start
, SE_SHARED
))) {
1394 * We need to zero a partial page.
1396 pagezero(pp
, poff
, PAGESIZE
- poff
);
1400 error
= pvn_vplist_dirty(vp
, start
, zfs_no_putpage
,
1401 B_INVAL
| B_TRUNC
, NULL
);
1404 rw_exit(&zp
->z_map_lock
);
1410 * Free space in a file
1412 * IN: zp - znode of file to free data in.
1413 * off - start of range
1414 * len - end of range (0 => EOF)
1415 * flag - current file open mode flags.
1416 * log - TRUE if this action should be logged
1418 * RETURN: 0 if success
1419 * error code if failure
1422 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1424 vnode_t
*vp
= ZTOV(zp
);
1426 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1427 zilog_t
*zilog
= zfsvfs
->z_log
;
1430 if (off
> zp
->z_phys
->zp_size
) {
1431 error
= zfs_extend(zp
, off
+len
);
1432 if (error
== 0 && log
)
1439 * Check for any locks in the region to be freed.
1441 if (MANDLOCK(vp
, (mode_t
)zp
->z_phys
->zp_mode
)) {
1442 uint64_t length
= (len
? len
: zp
->z_phys
->zp_size
- off
);
1443 if (error
= chklock(vp
, FWRITE
, off
, length
, flag
, NULL
))
1448 error
= zfs_trunc(zp
, off
);
1450 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1451 off
+ len
> zp
->z_phys
->zp_size
)
1452 error
= zfs_extend(zp
, off
+len
);
1457 tx
= dmu_tx_create(zfsvfs
->z_os
);
1458 dmu_tx_hold_bonus(tx
, zp
->z_id
);
1459 error
= dmu_tx_assign(tx
, zfsvfs
->z_assign
);
1461 if (error
== ERESTART
&& zfsvfs
->z_assign
== TXG_NOWAIT
) {
1470 zfs_time_stamper(zp
, CONTENT_MODIFIED
, tx
);
1471 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1478 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1481 uint64_t moid
, doid
, version
;
1482 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1486 znode_t
*rootzp
= NULL
;
1492 * First attempt to create master node.
1495 * In an empty objset, there are no blocks to read and thus
1496 * there can be no i/o errors (which we assert below).
1498 moid
= MASTER_NODE_OBJ
;
1499 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1500 DMU_OT_NONE
, 0, tx
);
1504 * Set starting attributes.
1506 if (spa_version(dmu_objset_spa(os
)) >= SPA_VERSION_FUID
)
1507 version
= ZPL_VERSION
;
1509 version
= ZPL_VERSION_FUID
- 1;
1510 error
= zap_update(os
, moid
, ZPL_VERSION_STR
,
1511 8, 1, &version
, tx
);
1513 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1514 /* For the moment we expect all zpl props to be uint64_ts */
1518 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1519 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1520 name
= nvpair_name(elem
);
1521 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1523 error
= zap_update(os
, moid
, ZPL_VERSION_STR
,
1524 8, 1, &version
, tx
);
1526 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1529 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1531 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1534 ASSERT(version
!= 0);
1537 * Create a delete queue.
1539 doid
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1541 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &doid
, tx
);
1545 * Create root znode. Create minimal znode/vnode/zfsvfs
1546 * to allow zfs_mknode to work.
1548 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
1549 vattr
.va_type
= VDIR
;
1550 vattr
.va_mode
= S_IFDIR
|0755;
1551 vattr
.va_uid
= crgetuid(cr
);
1552 vattr
.va_gid
= crgetgid(cr
);
1554 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1555 rootzp
->z_unlinked
= 0;
1556 rootzp
->z_atime_dirty
= 0;
1562 bzero(&zfsvfs
, sizeof (zfsvfs_t
));
1565 zfsvfs
.z_assign
= TXG_NOWAIT
;
1566 zfsvfs
.z_parent
= &zfsvfs
;
1567 zfsvfs
.z_version
= version
;
1568 zfsvfs
.z_use_fuids
= USE_FUIDS(version
, os
);
1569 zfsvfs
.z_norm
= norm
;
1571 * Fold case on file systems that are always or sometimes case
1574 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1575 zfsvfs
.z_norm
|= U8_TEXTPREP_TOUPPER
;
1577 mutex_init(&zfsvfs
.z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1578 list_create(&zfsvfs
.z_all_znodes
, sizeof (znode_t
),
1579 offsetof(znode_t
, z_link_node
));
1581 ASSERT(!POINTER_IS_VALID(rootzp
->z_zfsvfs
));
1582 rootzp
->z_zfsvfs
= &zfsvfs
;
1583 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, 0, NULL
, NULL
);
1584 ASSERT3P(zp
, ==, rootzp
);
1585 ASSERT(!vn_in_dnlc(ZTOV(rootzp
))); /* not valid to move */
1586 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1588 POINTER_INVALIDATE(&rootzp
->z_zfsvfs
);
1590 ZTOV(rootzp
)->v_count
= 0;
1591 dmu_buf_rele(rootzp
->z_dbuf
, NULL
);
1592 rootzp
->z_dbuf
= NULL
;
1593 kmem_cache_free(znode_cache
, rootzp
);
1596 #endif /* _KERNEL */
1598 * Given an object number, return its parent object number and whether
1599 * or not the object is an extended attribute directory.
1602 zfs_obj_to_pobj(objset_t
*osp
, uint64_t obj
, uint64_t *pobjp
, int *is_xattrdir
)
1605 dmu_object_info_t doi
;
1609 if ((error
= dmu_bonus_hold(osp
, obj
, FTAG
, &db
)) != 0)
1612 dmu_object_info_from_db(db
, &doi
);
1613 if (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1614 doi
.doi_bonus_size
< sizeof (znode_phys_t
)) {
1615 dmu_buf_rele(db
, FTAG
);
1620 *pobjp
= zp
->zp_parent
;
1621 *is_xattrdir
= ((zp
->zp_flags
& ZFS_XATTR
) != 0) &&
1622 S_ISDIR(zp
->zp_mode
);
1623 dmu_buf_rele(db
, FTAG
);
1629 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
1631 char *path
= buf
+ len
- 1;
1638 char component
[MAXNAMELEN
+ 2];
1642 if ((error
= zfs_obj_to_pobj(osp
, obj
, &pobj
,
1643 &is_xattrdir
)) != 0)
1654 (void) sprintf(component
+ 1, "<xattrdir>");
1656 error
= zap_value_search(osp
, pobj
, obj
,
1657 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
1662 complen
= strlen(component
);
1664 ASSERT(path
>= buf
);
1665 bcopy(component
, path
, complen
);
1670 (void) memmove(buf
, path
, buf
+ len
- path
);