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>
68 #include "zfs_comutil.h"
71 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
72 * turned on when DEBUG is also defined.
79 #define ZNODE_STAT_ADD(stat) ((stat)++)
81 #define ZNODE_STAT_ADD(stat) /* nothing */
82 #endif /* ZNODE_STATS */
84 #define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3))
85 #define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1))
88 * Functions needed for userland (ie: libzpool) are not put under
89 * #ifdef_KERNEL; the rest of the functions have dependencies
90 * (such as VFS logic) that will not compile easily in userland.
94 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
95 * be freed before it can be safely accessed.
97 krwlock_t zfsvfs_lock
;
99 static kmem_cache_t
*znode_cache
= NULL
;
103 znode_evict_error(dmu_buf_t
*dbuf
, void *user_ptr
)
106 * We should never drop all dbuf refs without first clearing
107 * the eviction callback.
109 panic("evicting znode %p\n", user_ptr
);
114 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
118 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
120 zp
->z_vnode
= vn_alloc(kmflags
);
121 if (zp
->z_vnode
== NULL
) {
124 ZTOV(zp
)->v_data
= zp
;
126 list_link_init(&zp
->z_link_node
);
128 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
129 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
130 rw_init(&zp
->z_name_lock
, NULL
, RW_DEFAULT
, NULL
);
131 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
133 mutex_init(&zp
->z_range_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
134 avl_create(&zp
->z_range_avl
, zfs_range_compare
,
135 sizeof (rl_t
), offsetof(rl_t
, r_node
));
137 zp
->z_dirlocks
= NULL
;
138 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_last_itx
= ozp
->z_last_itx
;
200 nzp
->z_gen
= ozp
->z_gen
;
201 nzp
->z_sync_cnt
= ozp
->z_sync_cnt
;
202 nzp
->z_is_sa
= ozp
->z_is_sa
;
203 nzp
->z_sa_hdl
= ozp
->z_sa_hdl
;
204 bcopy(ozp
->z_atime
, nzp
->z_atime
, sizeof (uint64_t) * 2);
205 nzp
->z_links
= ozp
->z_links
;
206 nzp
->z_size
= ozp
->z_size
;
207 nzp
->z_pflags
= ozp
->z_pflags
;
208 nzp
->z_uid
= ozp
->z_uid
;
209 nzp
->z_gid
= ozp
->z_gid
;
210 nzp
->z_mode
= ozp
->z_mode
;
213 * Since this is just an idle znode and kmem is already dealing with
214 * memory pressure, release any cached ACL.
216 if (ozp
->z_acl_cached
) {
217 zfs_acl_free(ozp
->z_acl_cached
);
218 ozp
->z_acl_cached
= NULL
;
221 sa_set_userp(nzp
->z_sa_hdl
, nzp
);
224 * Invalidate the original znode by clearing fields that provide a
225 * pointer back to the znode. Set the low bit of the vfs pointer to
226 * ensure that zfs_znode_move() recognizes the znode as invalid in any
227 * subsequent callback.
229 ozp
->z_sa_hdl
= NULL
;
230 POINTER_INVALIDATE(&ozp
->z_zfsvfs
);
235 zfs_znode_move(void *buf
, void *newbuf
, size_t size
, void *arg
)
237 znode_t
*ozp
= buf
, *nzp
= newbuf
;
242 * The znode is on the file system's list of known znodes if the vfs
243 * pointer is valid. We set the low bit of the vfs pointer when freeing
244 * the znode to invalidate it, and the memory patterns written by kmem
245 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
246 * created znode sets the vfs pointer last of all to indicate that the
247 * znode is known and in a valid state to be moved by this function.
249 zfsvfs
= ozp
->z_zfsvfs
;
250 if (!POINTER_IS_VALID(zfsvfs
)) {
251 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_invalid
);
252 return (KMEM_CBRC_DONT_KNOW
);
256 * Close a small window in which it's possible that the filesystem could
257 * be unmounted and freed, and zfsvfs, though valid in the previous
258 * statement, could point to unrelated memory by the time we try to
259 * prevent the filesystem from being unmounted.
261 rw_enter(&zfsvfs_lock
, RW_WRITER
);
262 if (zfsvfs
!= ozp
->z_zfsvfs
) {
263 rw_exit(&zfsvfs_lock
);
264 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_recheck1
);
265 return (KMEM_CBRC_DONT_KNOW
);
269 * If the znode is still valid, then so is the file system. We know that
270 * no valid file system can be freed while we hold zfsvfs_lock, so we
271 * can safely ensure that the filesystem is not and will not be
272 * unmounted. The next statement is equivalent to ZFS_ENTER().
274 rrw_enter(&zfsvfs
->z_teardown_lock
, RW_READER
, FTAG
);
275 if (zfsvfs
->z_unmounted
) {
277 rw_exit(&zfsvfs_lock
);
278 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_unmounted
);
279 return (KMEM_CBRC_DONT_KNOW
);
281 rw_exit(&zfsvfs_lock
);
283 mutex_enter(&zfsvfs
->z_znodes_lock
);
285 * Recheck the vfs pointer in case the znode was removed just before
286 * acquiring the lock.
288 if (zfsvfs
!= ozp
->z_zfsvfs
) {
289 mutex_exit(&zfsvfs
->z_znodes_lock
);
291 ZNODE_STAT_ADD(znode_move_stats
.zms_zfsvfs_recheck2
);
292 return (KMEM_CBRC_DONT_KNOW
);
296 * At this point we know that as long as we hold z_znodes_lock, the
297 * znode cannot be freed and fields within the znode can be safely
298 * accessed. Now, prevent a race with zfs_zget().
300 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs
, ozp
->z_id
) == 0) {
301 mutex_exit(&zfsvfs
->z_znodes_lock
);
303 ZNODE_STAT_ADD(znode_move_stats
.zms_obj_held
);
304 return (KMEM_CBRC_LATER
);
308 if (mutex_tryenter(&vp
->v_lock
) == 0) {
309 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
310 mutex_exit(&zfsvfs
->z_znodes_lock
);
312 ZNODE_STAT_ADD(znode_move_stats
.zms_vnode_locked
);
313 return (KMEM_CBRC_LATER
);
316 /* Only move znodes that are referenced _only_ by the DNLC. */
317 if (vp
->v_count
!= 1 || !vn_in_dnlc(vp
)) {
318 mutex_exit(&vp
->v_lock
);
319 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
320 mutex_exit(&zfsvfs
->z_znodes_lock
);
322 ZNODE_STAT_ADD(znode_move_stats
.zms_not_only_dnlc
);
323 return (KMEM_CBRC_LATER
);
327 * The znode is known and in a valid state to move. We're holding the
328 * locks needed to execute the critical section.
330 zfs_znode_move_impl(ozp
, nzp
);
331 mutex_exit(&vp
->v_lock
);
332 ZFS_OBJ_HOLD_EXIT(zfsvfs
, ozp
->z_id
);
334 list_link_replace(&ozp
->z_link_node
, &nzp
->z_link_node
);
335 mutex_exit(&zfsvfs
->z_znodes_lock
);
338 return (KMEM_CBRC_YES
);
347 rw_init(&zfsvfs_lock
, NULL
, RW_DEFAULT
, NULL
);
348 ASSERT(znode_cache
== NULL
);
349 znode_cache
= kmem_cache_create("zfs_znode_cache",
350 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
351 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, 0);
352 kmem_cache_set_move(znode_cache
, zfs_znode_move
);
359 * Cleanup vfs & vnode ops
361 zfs_remove_op_tables();
367 kmem_cache_destroy(znode_cache
);
369 rw_destroy(&zfsvfs_lock
);
372 struct vnodeops
*zfs_dvnodeops
;
373 struct vnodeops
*zfs_fvnodeops
;
374 struct vnodeops
*zfs_symvnodeops
;
375 struct vnodeops
*zfs_xdvnodeops
;
376 struct vnodeops
*zfs_evnodeops
;
377 struct vnodeops
*zfs_sharevnodeops
;
380 zfs_remove_op_tables()
386 (void) vfs_freevfsops_by_type(zfsfstype
);
393 vn_freevnodeops(zfs_dvnodeops
);
395 vn_freevnodeops(zfs_fvnodeops
);
397 vn_freevnodeops(zfs_symvnodeops
);
399 vn_freevnodeops(zfs_xdvnodeops
);
401 vn_freevnodeops(zfs_evnodeops
);
402 if (zfs_sharevnodeops
)
403 vn_freevnodeops(zfs_sharevnodeops
);
405 zfs_dvnodeops
= NULL
;
406 zfs_fvnodeops
= NULL
;
407 zfs_symvnodeops
= NULL
;
408 zfs_xdvnodeops
= NULL
;
409 zfs_evnodeops
= NULL
;
410 zfs_sharevnodeops
= NULL
;
413 extern const fs_operation_def_t zfs_dvnodeops_template
[];
414 extern const fs_operation_def_t zfs_fvnodeops_template
[];
415 extern const fs_operation_def_t zfs_xdvnodeops_template
[];
416 extern const fs_operation_def_t zfs_symvnodeops_template
[];
417 extern const fs_operation_def_t zfs_evnodeops_template
[];
418 extern const fs_operation_def_t zfs_sharevnodeops_template
[];
421 zfs_create_op_tables()
426 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
427 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
428 * In this case we just return as the ops vectors are already set up.
433 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_dvnodeops_template
,
438 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_fvnodeops_template
,
443 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_symvnodeops_template
,
448 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_xdvnodeops_template
,
453 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_evnodeops_template
,
458 error
= vn_make_ops(MNTTYPE_ZFS
, zfs_sharevnodeops_template
,
465 zfs_create_share_dir(zfsvfs_t
*zfsvfs
, dmu_tx_t
*tx
)
467 zfs_acl_ids_t acl_ids
;
474 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
475 vattr
.va_type
= VDIR
;
476 vattr
.va_mode
= S_IFDIR
|0555;
477 vattr
.va_uid
= crgetuid(kcred
);
478 vattr
.va_gid
= crgetgid(kcred
);
480 sharezp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
481 sharezp
->z_unlinked
= 0;
482 sharezp
->z_atime_dirty
= 0;
483 sharezp
->z_zfsvfs
= zfsvfs
;
484 sharezp
->z_is_sa
= zfsvfs
->z_use_sa
;
490 VERIFY(0 == zfs_acl_ids_create(sharezp
, IS_ROOT_NODE
, &vattr
,
491 kcred
, NULL
, &acl_ids
));
492 zfs_mknode(sharezp
, &vattr
, tx
, kcred
, IS_ROOT_NODE
, &zp
, &acl_ids
);
493 ASSERT3P(zp
, ==, sharezp
);
494 ASSERT(!vn_in_dnlc(ZTOV(sharezp
))); /* not valid to move */
495 POINTER_INVALIDATE(&sharezp
->z_zfsvfs
);
496 error
= zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
497 ZFS_SHARES_DIR
, 8, 1, &sharezp
->z_id
, tx
);
498 zfsvfs
->z_shares_dir
= sharezp
->z_id
;
500 zfs_acl_ids_free(&acl_ids
);
501 ZTOV(sharezp
)->v_count
= 0;
502 sa_handle_destroy(sharezp
->z_sa_hdl
);
503 kmem_cache_free(znode_cache
, sharezp
);
509 * define a couple of values we need available
510 * for both 64 and 32 bit environments.
513 #define NBITSMINOR64 32
516 #define MAXMAJ64 0xffffffffUL
519 #define MAXMIN64 0xffffffffUL
523 * Create special expldev for ZFS private use.
524 * Can't use standard expldev since it doesn't do
525 * what we want. The standard expldev() takes a
526 * dev32_t in LP64 and expands it to a long dev_t.
527 * We need an interface that takes a dev32_t in ILP32
528 * and expands it to a long dev_t.
531 zfs_expldev(dev_t dev
)
534 major_t major
= (major_t
)dev
>> NBITSMINOR32
& MAXMAJ32
;
535 return (((uint64_t)major
<< NBITSMINOR64
) |
536 ((minor_t
)dev
& MAXMIN32
));
543 * Special cmpldev for ZFS private use.
544 * Can't use standard cmpldev since it takes
545 * a long dev_t and compresses it to dev32_t in
546 * LP64. We need to do a compaction of a long dev_t
547 * to a dev32_t in ILP32.
550 zfs_cmpldev(uint64_t dev
)
553 minor_t minor
= (minor_t
)dev
& MAXMIN64
;
554 major_t major
= (major_t
)(dev
>> NBITSMINOR64
) & MAXMAJ64
;
556 if (major
> MAXMAJ32
|| minor
> MAXMIN32
)
559 return (((dev32_t
)major
<< NBITSMINOR32
) | minor
);
566 zfs_znode_sa_init(zfsvfs_t
*zfsvfs
, znode_t
*zp
,
567 dmu_buf_t
*db
, dmu_object_type_t obj_type
, sa_handle_t
*sa_hdl
)
569 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
) || (zfsvfs
== zp
->z_zfsvfs
));
570 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs
, zp
->z_id
)));
572 mutex_enter(&zp
->z_lock
);
574 ASSERT(zp
->z_sa_hdl
== NULL
);
575 ASSERT(zp
->z_acl_cached
== NULL
);
576 if (sa_hdl
== NULL
) {
577 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, zp
,
578 SA_HDL_SHARED
, &zp
->z_sa_hdl
));
580 zp
->z_sa_hdl
= sa_hdl
;
581 sa_set_userp(sa_hdl
, zp
);
584 zp
->z_is_sa
= (obj_type
== DMU_OT_SA
) ? B_TRUE
: B_FALSE
;
587 * Slap on VROOT if we are the root znode
589 if (zp
->z_id
== zfsvfs
->z_root
)
590 ZTOV(zp
)->v_flag
|= VROOT
;
592 mutex_exit(&zp
->z_lock
);
597 zfs_znode_dmu_fini(znode_t
*zp
)
599 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp
->z_zfsvfs
, zp
->z_id
)) ||
601 RW_WRITE_HELD(&zp
->z_zfsvfs
->z_teardown_inactive_lock
));
603 sa_handle_destroy(zp
->z_sa_hdl
);
608 * Construct a new znode/vnode and intialize.
610 * This does not do a call to dmu_set_user() that is
611 * up to the caller to do, in case you don't want to
615 zfs_znode_alloc(zfsvfs_t
*zfsvfs
, dmu_buf_t
*db
, int blksz
,
616 dmu_object_type_t obj_type
, sa_handle_t
*hdl
)
623 sa_bulk_attr_t bulk
[9];
626 zp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
628 ASSERT(zp
->z_dirlocks
== NULL
);
629 ASSERT(!POINTER_IS_VALID(zp
->z_zfsvfs
));
632 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
633 * the zfs_znode_move() callback.
637 zp
->z_atime_dirty
= 0;
640 zp
->z_id
= db
->db_object
;
642 zp
->z_seq
= 0x7A4653;
648 zfs_znode_sa_init(zfsvfs
, zp
, db
, obj_type
, hdl
);
650 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
, &mode
, 8);
651 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
, &zp
->z_gen
, 8);
652 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
654 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
656 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
658 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zfsvfs
), NULL
, &parent
, 8);
659 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
661 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
663 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
666 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
) != 0 || zp
->z_gen
== 0) {
668 sa_handle_destroy(zp
->z_sa_hdl
);
669 kmem_cache_free(znode_cache
, zp
);
673 zp
->z_uid
= zfs_fuid_map_id(zfsvfs
, uid
, CRED(), ZFS_OWNER
);
674 zp
->z_gid
= zfs_fuid_map_id(zfsvfs
, gid
, CRED(), ZFS_GROUP
);
676 vp
->v_vfsp
= zfsvfs
->z_parent
->z_vfs
;
678 vp
->v_type
= IFTOVT((mode_t
)mode
);
680 switch (vp
->v_type
) {
682 if (zp
->z_pflags
& ZFS_XATTR
) {
683 vn_setops(vp
, zfs_xdvnodeops
);
684 vp
->v_flag
|= V_XATTRDIR
;
686 vn_setops(vp
, zfs_dvnodeops
);
688 zp
->z_zn_prefetch
= B_TRUE
; /* z_prefetch default is enabled */
694 VERIFY(sa_lookup(zp
->z_sa_hdl
, SA_ZPL_RDEV(zfsvfs
),
695 &rdev
, sizeof (rdev
)) == 0);
697 vp
->v_rdev
= zfs_cmpldev(rdev
);
703 vn_setops(vp
, zfs_fvnodeops
);
706 vp
->v_flag
|= VMODSORT
;
707 if (parent
== zfsvfs
->z_shares_dir
) {
708 ASSERT(uid
== 0 && gid
== 0);
709 vn_setops(vp
, zfs_sharevnodeops
);
711 vn_setops(vp
, zfs_fvnodeops
);
715 vn_setops(vp
, zfs_symvnodeops
);
718 vn_setops(vp
, zfs_evnodeops
);
722 mutex_enter(&zfsvfs
->z_znodes_lock
);
723 list_insert_tail(&zfsvfs
->z_all_znodes
, zp
);
726 * Everything else must be valid before assigning z_zfsvfs makes the
727 * znode eligible for zfs_znode_move().
729 zp
->z_zfsvfs
= zfsvfs
;
730 mutex_exit(&zfsvfs
->z_znodes_lock
);
732 VFS_HOLD(zfsvfs
->z_vfs
);
736 static uint64_t empty_xattr
;
737 static uint64_t pad
[4];
738 static zfs_acl_phys_t acl_phys
;
740 * Create a new DMU object to hold a zfs znode.
742 * IN: dzp - parent directory for new znode
743 * vap - file attributes for new znode
744 * tx - dmu transaction id for zap operations
745 * cr - credentials of caller
747 * IS_ROOT_NODE - new object will be root
748 * IS_XATTR - new object is an attribute
749 * bonuslen - length of bonus buffer
750 * setaclp - File/Dir initial ACL
751 * fuidp - Tracks fuid allocation.
753 * OUT: zpp - allocated znode
757 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
758 uint_t flag
, znode_t
**zpp
, zfs_acl_ids_t
*acl_ids
)
760 uint64_t crtime
[2], atime
[2], mtime
[2], ctime
[2];
761 uint64_t mode
, size
, links
, parent
, pflags
;
762 uint64_t dzp_pflags
= 0;
764 zfsvfs_t
*zfsvfs
= dzp
->z_zfsvfs
;
771 dmu_object_type_t obj_type
;
772 sa_bulk_attr_t sa_attrs
[ZPL_END
];
774 zfs_acl_locator_cb_t locate
= { 0 };
776 ASSERT(vap
&& (vap
->va_mask
& (AT_TYPE
|AT_MODE
)) == (AT_TYPE
|AT_MODE
));
778 if (zfsvfs
->z_replay
) {
779 obj
= vap
->va_nodeid
;
780 now
= vap
->va_ctime
; /* see zfs_replay_create() */
781 gen
= vap
->va_nblocks
; /* ditto */
785 gen
= dmu_tx_get_txg(tx
);
788 obj_type
= zfsvfs
->z_use_sa
? DMU_OT_SA
: DMU_OT_ZNODE
;
789 bonuslen
= (obj_type
== DMU_OT_SA
) ?
790 DN_MAX_BONUSLEN
: ZFS_OLD_ZNODE_PHYS_SIZE
;
793 * Create a new DMU object.
796 * There's currently no mechanism for pre-reading the blocks that will
797 * be to needed allocate a new object, so we accept the small chance
798 * that there will be an i/o error and we will fail one of the
801 if (vap
->va_type
== VDIR
) {
802 if (zfsvfs
->z_replay
) {
803 err
= zap_create_claim_norm(zfsvfs
->z_os
, obj
,
804 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
805 obj_type
, bonuslen
, tx
);
806 ASSERT3U(err
, ==, 0);
808 obj
= zap_create_norm(zfsvfs
->z_os
,
809 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
810 obj_type
, bonuslen
, tx
);
813 if (zfsvfs
->z_replay
) {
814 err
= dmu_object_claim(zfsvfs
->z_os
, obj
,
815 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
816 obj_type
, bonuslen
, tx
);
817 ASSERT3U(err
, ==, 0);
819 obj
= dmu_object_alloc(zfsvfs
->z_os
,
820 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
821 obj_type
, bonuslen
, tx
);
825 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
826 VERIFY(0 == sa_buf_hold(zfsvfs
->z_os
, obj
, NULL
, &db
));
829 * If this is the root, fix up the half-initialized parent pointer
830 * to reference the just-allocated physical data area.
832 if (flag
& IS_ROOT_NODE
) {
835 dzp_pflags
= dzp
->z_pflags
;
839 * If parent is an xattr, so am I.
841 if (dzp_pflags
& ZFS_XATTR
) {
845 if (zfsvfs
->z_use_fuids
)
846 pflags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
850 if (vap
->va_type
== VDIR
) {
851 size
= 2; /* contents ("." and "..") */
852 links
= (flag
& (IS_ROOT_NODE
| IS_XATTR
)) ? 2 : 1;
857 if (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
) {
858 rdev
= zfs_expldev(vap
->va_rdev
);
862 mode
= acl_ids
->z_mode
;
867 * No execs denied will be deterimed when zfs_mode_compute() is called.
869 pflags
|= acl_ids
->z_aclp
->z_hints
&
870 (ZFS_ACL_TRIVIAL
|ZFS_INHERIT_ACE
|ZFS_ACL_AUTO_INHERIT
|
871 ZFS_ACL_DEFAULTED
|ZFS_ACL_PROTECTED
);
873 ZFS_TIME_ENCODE(&now
, crtime
);
874 ZFS_TIME_ENCODE(&now
, ctime
);
876 if (vap
->va_mask
& AT_ATIME
) {
877 ZFS_TIME_ENCODE(&vap
->va_atime
, atime
);
879 ZFS_TIME_ENCODE(&now
, atime
);
882 if (vap
->va_mask
& AT_MTIME
) {
883 ZFS_TIME_ENCODE(&vap
->va_mtime
, mtime
);
885 ZFS_TIME_ENCODE(&now
, mtime
);
888 /* Now add in all of the "SA" attributes */
889 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, NULL
, SA_HDL_SHARED
,
893 * Setup the array of attributes to be replaced/set on the new file
895 * order for DMU_OT_ZNODE is critical since it needs to be constructed
896 * in the old znode_phys_t format. Don't change this ordering
899 if (obj_type
== DMU_OT_ZNODE
) {
900 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
902 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
904 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
906 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
908 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
910 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
912 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
914 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
917 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
919 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
921 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
923 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
924 &acl_ids
->z_fuid
, 8);
925 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
926 &acl_ids
->z_fgid
, 8);
927 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
929 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
931 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
933 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
935 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
937 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
941 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zfsvfs
), NULL
, &links
, 8);
943 if (obj_type
== DMU_OT_ZNODE
) {
944 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zfsvfs
), NULL
,
947 if (obj_type
== DMU_OT_ZNODE
||
948 (vap
->va_type
== VBLK
|| vap
->va_type
== VCHR
)) {
949 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zfsvfs
),
953 if (obj_type
== DMU_OT_ZNODE
) {
954 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
956 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
957 &acl_ids
->z_fuid
, 8);
958 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
959 &acl_ids
->z_fgid
, 8);
960 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zfsvfs
), NULL
, pad
,
961 sizeof (uint64_t) * 4);
962 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zfsvfs
), NULL
,
963 &acl_phys
, sizeof (zfs_acl_phys_t
));
964 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
965 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zfsvfs
), NULL
,
966 &acl_ids
->z_aclp
->z_acl_count
, 8);
967 locate
.cb_aclp
= acl_ids
->z_aclp
;
968 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zfsvfs
),
969 zfs_acl_data_locator
, &locate
,
970 acl_ids
->z_aclp
->z_acl_bytes
);
971 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
972 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
975 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
977 if (!(flag
& IS_ROOT_NODE
)) {
978 *zpp
= zfs_znode_alloc(zfsvfs
, db
, 0, obj_type
, sa_hdl
);
979 ASSERT(*zpp
!= NULL
);
982 * If we are creating the root node, the "parent" we
983 * passed in is the znode for the root.
987 (*zpp
)->z_sa_hdl
= sa_hdl
;
990 (*zpp
)->z_pflags
= pflags
;
991 (*zpp
)->z_mode
= mode
;
993 if (vap
->va_mask
& AT_XVATTR
)
994 zfs_xvattr_set(*zpp
, (xvattr_t
*)vap
, tx
);
996 if (obj_type
== DMU_OT_ZNODE
||
997 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
998 err
= zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
);
999 ASSERT3P(err
, ==, 0);
1001 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1005 * zfs_xvattr_set only updates the in-core attributes
1006 * it is assumed the caller will be doing an sa_bulk_update
1007 * to push the changes out
1010 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
1014 xoap
= xva_getxoptattr(xvap
);
1017 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
1019 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
1020 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(zp
->z_zfsvfs
),
1021 ×
, sizeof (times
), tx
);
1022 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
1024 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
1025 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
1027 XVA_SET_RTN(xvap
, XAT_READONLY
);
1029 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
1030 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
1032 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
1034 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
1035 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
1037 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
1039 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
1040 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
1042 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
1044 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
1045 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
1047 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
1049 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
1050 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
1052 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
1054 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
1055 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
1057 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
1059 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
1060 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
1062 XVA_SET_RTN(xvap
, XAT_NODUMP
);
1064 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
1065 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
1067 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
1069 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
1070 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
1071 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
1072 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
1074 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
1075 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
1077 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
1079 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
1080 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
1081 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
1083 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
1084 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
1086 XVA_SET_RTN(xvap
, XAT_REPARSE
);
1091 zfs_zget(zfsvfs_t
*zfsvfs
, uint64_t obj_num
, znode_t
**zpp
)
1093 dmu_object_info_t doi
;
1101 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1103 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1105 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1109 dmu_object_info_from_db(db
, &doi
);
1110 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1111 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1112 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1113 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1114 sa_buf_rele(db
, NULL
);
1115 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1119 hdl
= dmu_buf_get_user(db
);
1121 zp
= sa_get_userdata(hdl
);
1125 * Since "SA" does immediate eviction we
1126 * should never find a sa handle that doesn't
1127 * know about the znode.
1130 ASSERT3P(zp
, !=, NULL
);
1132 mutex_enter(&zp
->z_lock
);
1133 ASSERT3U(zp
->z_id
, ==, obj_num
);
1134 if (zp
->z_unlinked
) {
1141 sa_buf_rele(db
, NULL
);
1142 mutex_exit(&zp
->z_lock
);
1143 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1148 * Not found create new znode/vnode
1149 * but only if file exists.
1151 * There is a small window where zfs_vget() could
1152 * find this object while a file create is still in
1153 * progress. This is checked for in zfs_znode_alloc()
1155 * if zfs_znode_alloc() fails it will drop the hold on the
1158 zp
= zfs_znode_alloc(zfsvfs
, db
, doi
.doi_data_block_size
,
1159 doi
.doi_bonus_type
, NULL
);
1165 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1170 zfs_rezget(znode_t
*zp
)
1172 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1173 dmu_object_info_t doi
;
1175 uint64_t obj_num
= zp
->z_id
;
1178 sa_bulk_attr_t bulk
[8];
1183 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj_num
);
1185 mutex_enter(&zp
->z_acl_lock
);
1186 if (zp
->z_acl_cached
) {
1187 zfs_acl_free(zp
->z_acl_cached
);
1188 zp
->z_acl_cached
= NULL
;
1191 mutex_exit(&zp
->z_acl_lock
);
1192 ASSERT(zp
->z_sa_hdl
== NULL
);
1193 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1195 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1199 dmu_object_info_from_db(db
, &doi
);
1200 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1201 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1202 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1203 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1204 sa_buf_rele(db
, NULL
);
1205 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1209 zfs_znode_sa_init(zfsvfs
, zp
, db
, doi
.doi_bonus_type
, NULL
);
1211 /* reload cached values */
1212 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
,
1213 &gen
, sizeof (gen
));
1214 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
1215 &zp
->z_size
, sizeof (zp
->z_size
));
1216 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
1217 &zp
->z_links
, sizeof (zp
->z_links
));
1218 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
1219 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1220 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
1221 &zp
->z_atime
, sizeof (zp
->z_atime
));
1222 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
1223 &uid
, sizeof (uid
));
1224 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
1225 &gid
, sizeof (gid
));
1226 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
,
1227 &mode
, sizeof (mode
));
1231 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
1232 zfs_znode_dmu_fini(zp
);
1233 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1237 if (gen
!= zp
->z_gen
) {
1238 zfs_znode_dmu_fini(zp
);
1239 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1243 zp
->z_uid
= zfs_fuid_map_id(zfsvfs
, uid
, CRED(), ZFS_OWNER
);
1244 zp
->z_gid
= zfs_fuid_map_id(zfsvfs
, gid
, CRED(), ZFS_GROUP
);
1245 zp
->z_unlinked
= (zp
->z_links
== 0);
1246 zp
->z_blksz
= doi
.doi_data_block_size
;
1248 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj_num
);
1254 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1256 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1257 objset_t
*os
= zfsvfs
->z_os
;
1258 uint64_t obj
= zp
->z_id
;
1259 uint64_t acl_obj
= ZFS_EXTERNAL_ACL(zp
);
1261 ZFS_OBJ_HOLD_ENTER(zfsvfs
, obj
);
1263 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1264 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1265 zfs_znode_dmu_fini(zp
);
1266 ZFS_OBJ_HOLD_EXIT(zfsvfs
, obj
);
1271 zfs_zinactive(znode_t
*zp
)
1273 vnode_t
*vp
= ZTOV(zp
);
1274 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1275 uint64_t z_id
= zp
->z_id
;
1277 ASSERT(zp
->z_sa_hdl
);
1280 * Don't allow a zfs_zget() while were trying to release this znode
1282 ZFS_OBJ_HOLD_ENTER(zfsvfs
, z_id
);
1284 mutex_enter(&zp
->z_lock
);
1285 mutex_enter(&vp
->v_lock
);
1287 if (vp
->v_count
> 0 || vn_has_cached_data(vp
)) {
1289 * If the hold count is greater than zero, somebody has
1290 * obtained a new reference on this znode while we were
1291 * processing it here, so we are done. If we still have
1292 * mapped pages then we are also done, since we don't
1293 * want to inactivate the znode until the pages get pushed.
1295 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1296 * this seems like it would leave the znode hanging with
1297 * no chance to go inactive...
1299 mutex_exit(&vp
->v_lock
);
1300 mutex_exit(&zp
->z_lock
);
1301 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1304 mutex_exit(&vp
->v_lock
);
1307 * If this was the last reference to a file with no links,
1308 * remove the file from the file system.
1310 if (zp
->z_unlinked
) {
1311 mutex_exit(&zp
->z_lock
);
1312 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1317 mutex_exit(&zp
->z_lock
);
1318 zfs_znode_dmu_fini(zp
);
1319 ZFS_OBJ_HOLD_EXIT(zfsvfs
, z_id
);
1324 zfs_znode_free(znode_t
*zp
)
1326 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1328 vn_invalid(ZTOV(zp
));
1330 ASSERT(ZTOV(zp
)->v_count
== 0);
1332 mutex_enter(&zfsvfs
->z_znodes_lock
);
1333 POINTER_INVALIDATE(&zp
->z_zfsvfs
);
1334 list_remove(&zfsvfs
->z_all_znodes
, zp
);
1335 mutex_exit(&zfsvfs
->z_znodes_lock
);
1337 if (zp
->z_acl_cached
) {
1338 zfs_acl_free(zp
->z_acl_cached
);
1339 zp
->z_acl_cached
= NULL
;
1342 kmem_cache_free(znode_cache
, zp
);
1344 VFS_RELE(zfsvfs
->z_vfs
);
1348 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1349 uint64_t ctime
[2], boolean_t have_tx
)
1355 if (have_tx
) { /* will sa_bulk_update happen really soon? */
1356 zp
->z_atime_dirty
= 0;
1359 zp
->z_atime_dirty
= 1;
1362 if (flag
& AT_ATIME
) {
1363 ZFS_TIME_ENCODE(&now
, zp
->z_atime
);
1366 if (flag
& AT_MTIME
) {
1367 ZFS_TIME_ENCODE(&now
, mtime
);
1368 if (zp
->z_zfsvfs
->z_use_fuids
) {
1369 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1374 if (flag
& AT_CTIME
) {
1375 ZFS_TIME_ENCODE(&now
, ctime
);
1376 if (zp
->z_zfsvfs
->z_use_fuids
)
1377 zp
->z_pflags
|= ZFS_ARCHIVE
;
1382 * Grow the block size for a file.
1384 * IN: zp - znode of file to free data in.
1385 * size - requested block size
1386 * tx - open transaction.
1388 * NOTE: this function assumes that the znode is write locked.
1391 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1396 if (size
<= zp
->z_blksz
)
1399 * If the file size is already greater than the current blocksize,
1400 * we will not grow. If there is more than one block in a file,
1401 * the blocksize cannot change.
1403 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1406 error
= dmu_object_set_blocksize(zp
->z_zfsvfs
->z_os
, zp
->z_id
,
1409 if (error
== ENOTSUP
)
1411 ASSERT3U(error
, ==, 0);
1413 /* What blocksize did we actually get? */
1414 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1418 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1419 * be calling back into the fs for a putpage(). E.g.: when truncating
1420 * a file, the pages being "thrown away* don't need to be written out.
1424 zfs_no_putpage(vnode_t
*vp
, page_t
*pp
, u_offset_t
*offp
, size_t *lenp
,
1425 int flags
, cred_t
*cr
)
1432 * Increase the file length
1434 * IN: zp - znode of file to free data in.
1435 * end - new end-of-file
1437 * RETURN: 0 if success
1438 * error code if failure
1441 zfs_extend(znode_t
*zp
, uint64_t end
)
1443 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1450 * We will change zp_size, lock the whole file.
1452 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1455 * Nothing to do if file already at desired length.
1457 if (end
<= zp
->z_size
) {
1458 zfs_range_unlock(rl
);
1462 tx
= dmu_tx_create(zfsvfs
->z_os
);
1463 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1464 zfs_sa_upgrade_txholds(tx
, zp
);
1465 if (end
> zp
->z_blksz
&&
1466 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zfsvfs
->z_max_blksz
)) {
1468 * We are growing the file past the current block size.
1470 if (zp
->z_blksz
> zp
->z_zfsvfs
->z_max_blksz
) {
1471 ASSERT(!ISP2(zp
->z_blksz
));
1472 newblksz
= MIN(end
, SPA_MAXBLOCKSIZE
);
1474 newblksz
= MIN(end
, zp
->z_zfsvfs
->z_max_blksz
);
1476 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1481 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1483 if (error
== ERESTART
) {
1489 zfs_range_unlock(rl
);
1494 zfs_grow_blocksize(zp
, newblksz
, tx
);
1498 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(zp
->z_zfsvfs
),
1499 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1501 zfs_range_unlock(rl
);
1509 * Free space in a file.
1511 * IN: zp - znode of file to free data in.
1512 * off - start of section to free.
1513 * len - length of section to free.
1515 * RETURN: 0 if success
1516 * error code if failure
1519 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1521 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1526 * Lock the range being freed.
1528 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1531 * Nothing to do if file already at desired length.
1533 if (off
>= zp
->z_size
) {
1534 zfs_range_unlock(rl
);
1538 if (off
+ len
> zp
->z_size
)
1539 len
= zp
->z_size
- off
;
1541 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, off
, len
);
1543 zfs_range_unlock(rl
);
1551 * IN: zp - znode of file to free data in.
1552 * end - new end-of-file.
1554 * RETURN: 0 if success
1555 * error code if failure
1558 zfs_trunc(znode_t
*zp
, uint64_t end
)
1560 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1561 vnode_t
*vp
= ZTOV(zp
);
1567 * We will change zp_size, lock the whole file.
1569 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1572 * Nothing to do if file already at desired length.
1574 if (end
>= zp
->z_size
) {
1575 zfs_range_unlock(rl
);
1579 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, end
, -1);
1581 zfs_range_unlock(rl
);
1585 tx
= dmu_tx_create(zfsvfs
->z_os
);
1586 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1587 zfs_sa_upgrade_txholds(tx
, zp
);
1588 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1590 if (error
== ERESTART
) {
1596 zfs_range_unlock(rl
);
1602 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(zp
->z_zfsvfs
),
1603 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1608 * Clear any mapped pages in the truncated region. This has to
1609 * happen outside of the transaction to avoid the possibility of
1610 * a deadlock with someone trying to push a page that we are
1611 * about to invalidate.
1613 if (vn_has_cached_data(vp
)) {
1615 uint64_t start
= end
& PAGEMASK
;
1616 int poff
= end
& PAGEOFFSET
;
1618 if (poff
!= 0 && (pp
= page_lookup(vp
, start
, SE_SHARED
))) {
1620 * We need to zero a partial page.
1622 pagezero(pp
, poff
, PAGESIZE
- poff
);
1626 error
= pvn_vplist_dirty(vp
, start
, zfs_no_putpage
,
1627 B_INVAL
| B_TRUNC
, NULL
);
1631 zfs_range_unlock(rl
);
1637 * Free space in a file
1639 * IN: zp - znode of file to free data in.
1640 * off - start of range
1641 * len - end of range (0 => EOF)
1642 * flag - current file open mode flags.
1643 * log - TRUE if this action should be logged
1645 * RETURN: 0 if success
1646 * error code if failure
1649 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1651 vnode_t
*vp
= ZTOV(zp
);
1653 zfsvfs_t
*zfsvfs
= zp
->z_zfsvfs
;
1654 zilog_t
*zilog
= zfsvfs
->z_log
;
1656 uint64_t mtime
[2], ctime
[2];
1657 sa_bulk_attr_t bulk
[3];
1661 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zfsvfs
), &mode
,
1662 sizeof (mode
))) != 0)
1665 if (off
> zp
->z_size
) {
1666 error
= zfs_extend(zp
, off
+len
);
1667 if (error
== 0 && log
)
1674 * Check for any locks in the region to be freed.
1677 if (MANDLOCK(vp
, (mode_t
)mode
)) {
1678 uint64_t length
= (len
? len
: zp
->z_size
- off
);
1679 if (error
= chklock(vp
, FWRITE
, off
, length
, flag
, NULL
))
1684 error
= zfs_trunc(zp
, off
);
1686 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1687 off
+ len
> zp
->z_size
)
1688 error
= zfs_extend(zp
, off
+len
);
1693 tx
= dmu_tx_create(zfsvfs
->z_os
);
1694 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1695 zfs_sa_upgrade_txholds(tx
, zp
);
1696 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1698 if (error
== ERESTART
) {
1707 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
, mtime
, 16);
1708 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
, ctime
, 16);
1709 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1710 NULL
, &zp
->z_pflags
, 8);
1711 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
1712 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1715 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1722 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1725 uint64_t moid
, obj
, sa_obj
, version
;
1726 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1731 znode_t
*rootzp
= NULL
;
1735 zfs_acl_ids_t acl_ids
;
1738 * First attempt to create master node.
1741 * In an empty objset, there are no blocks to read and thus
1742 * there can be no i/o errors (which we assert below).
1744 moid
= MASTER_NODE_OBJ
;
1745 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1746 DMU_OT_NONE
, 0, tx
);
1750 * Set starting attributes.
1752 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1754 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1755 /* For the moment we expect all zpl props to be uint64_ts */
1759 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1760 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1761 name
= nvpair_name(elem
);
1762 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1766 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1769 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1771 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1774 ASSERT(version
!= 0);
1775 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1778 * Create zap object used for SA attribute registration
1781 if (version
>= ZPL_VERSION_SA
) {
1782 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1783 DMU_OT_NONE
, 0, tx
);
1784 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1790 * Create a delete queue.
1792 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1794 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1798 * Create root znode. Create minimal znode/vnode/zfsvfs
1799 * to allow zfs_mknode to work.
1801 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
1802 vattr
.va_type
= VDIR
;
1803 vattr
.va_mode
= S_IFDIR
|0755;
1804 vattr
.va_uid
= crgetuid(cr
);
1805 vattr
.va_gid
= crgetgid(cr
);
1807 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1808 rootzp
->z_unlinked
= 0;
1809 rootzp
->z_atime_dirty
= 0;
1810 rootzp
->z_is_sa
= USE_SA(version
, os
);
1816 bzero(&zfsvfs
, sizeof (zfsvfs_t
));
1819 zfsvfs
.z_parent
= &zfsvfs
;
1820 zfsvfs
.z_version
= version
;
1821 zfsvfs
.z_use_fuids
= USE_FUIDS(version
, os
);
1822 zfsvfs
.z_use_sa
= USE_SA(version
, os
);
1823 zfsvfs
.z_norm
= norm
;
1825 zfsvfs
.z_attr_table
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
);
1828 * Fold case on file systems that are always or sometimes case
1831 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1832 zfsvfs
.z_norm
|= U8_TEXTPREP_TOUPPER
;
1834 mutex_init(&zfsvfs
.z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1835 list_create(&zfsvfs
.z_all_znodes
, sizeof (znode_t
),
1836 offsetof(znode_t
, z_link_node
));
1838 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1839 mutex_init(&zfsvfs
.z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1841 ASSERT(!POINTER_IS_VALID(rootzp
->z_zfsvfs
));
1842 rootzp
->z_zfsvfs
= &zfsvfs
;
1843 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1844 cr
, NULL
, &acl_ids
));
1845 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1846 ASSERT3P(zp
, ==, rootzp
);
1847 ASSERT(!vn_in_dnlc(ZTOV(rootzp
))); /* not valid to move */
1848 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1850 zfs_acl_ids_free(&acl_ids
);
1851 POINTER_INVALIDATE(&rootzp
->z_zfsvfs
);
1853 ZTOV(rootzp
)->v_count
= 0;
1854 sa_handle_destroy(rootzp
->z_sa_hdl
);
1855 kmem_cache_free(znode_cache
, rootzp
);
1858 * Create shares directory
1861 error
= zfs_create_share_dir(&zfsvfs
, tx
);
1865 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1866 mutex_destroy(&zfsvfs
.z_hold_mtx
[i
]);
1869 #endif /* _KERNEL */
1872 * Given an object number, return its parent object number and whether
1873 * or not the object is an extended attribute directory.
1876 zfs_obj_to_pobj(objset_t
*osp
, uint64_t obj
, uint64_t *pobjp
, int *is_xattrdir
,
1877 sa_attr_type_t
*sa_table
)
1880 dmu_object_info_t doi
;
1885 sa_bulk_attr_t bulk
[3];
1889 if ((error
= sa_buf_hold(osp
, obj
, FTAG
, &db
)) != 0)
1892 dmu_object_info_from_db(db
, &doi
);
1893 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1894 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
1895 doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1896 doi
.doi_bonus_size
< sizeof (znode_phys_t
)) {
1897 sa_buf_rele(db
, FTAG
);
1901 if ((error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
,
1903 sa_buf_rele(db
, FTAG
);
1907 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
],
1909 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
1911 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1914 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0) {
1915 sa_buf_rele(db
, FTAG
);
1916 sa_handle_destroy(hdl
);
1920 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
1921 sa_handle_destroy(hdl
);
1922 sa_buf_rele(db
, FTAG
);
1928 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
1930 char *path
= buf
+ len
- 1;
1931 sa_attr_type_t
*sa_table
;
1933 uint64_t sa_obj
= 0;
1937 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1939 if (error
!= 0 && error
!= ENOENT
)
1942 sa_table
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
);
1946 char component
[MAXNAMELEN
+ 2];
1950 if ((error
= zfs_obj_to_pobj(osp
, obj
, &pobj
,
1951 &is_xattrdir
, sa_table
)) != 0)
1962 (void) sprintf(component
+ 1, "<xattrdir>");
1964 error
= zap_value_search(osp
, pobj
, obj
,
1965 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
1970 complen
= strlen(component
);
1972 ASSERT(path
>= buf
);
1973 bcopy(component
, path
, complen
);
1978 (void) memmove(buf
, path
, buf
+ len
- path
);