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, 2014 by Delphix. All rights reserved.
26 /* Portions Copyright 2007 Jeremy Teo */
29 #include <sys/types.h>
30 #include <sys/param.h>
32 #include <sys/sysmacros.h>
33 #include <sys/mntent.h>
34 #include <sys/u8_textprep.h>
35 #include <sys/dsl_dataset.h>
37 #include <sys/vnode.h>
40 #include <sys/errno.h>
42 #include <sys/atomic.h>
43 #include <sys/zfs_dir.h>
44 #include <sys/zfs_acl.h>
45 #include <sys/zfs_ioctl.h>
46 #include <sys/zfs_rlock.h>
47 #include <sys/zfs_fuid.h>
48 #include <sys/zfs_vnops.h>
49 #include <sys/zfs_ctldir.h>
50 #include <sys/dnode.h>
51 #include <sys/fs/zfs.h>
56 #include <sys/dmu_objset.h>
57 #include <sys/dmu_tx.h>
58 #include <sys/refcount.h>
61 #include <sys/zfs_znode.h>
63 #include <sys/zfs_sa.h>
64 #include <sys/zfs_stat.h>
67 #include "zfs_comutil.h"
70 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
71 * turned on when DEBUG is also defined.
78 #define ZNODE_STAT_ADD(stat) ((stat)++)
80 #define ZNODE_STAT_ADD(stat) /* nothing */
81 #endif /* ZNODE_STATS */
84 * Functions needed for userland (ie: libzpool) are not put under
85 * #ifdef_KERNEL; the rest of the functions have dependencies
86 * (such as VFS logic) that will not compile easily in userland.
90 static kmem_cache_t
*znode_cache
= NULL
;
91 static kmem_cache_t
*znode_hold_cache
= NULL
;
92 unsigned int zfs_object_mutex_size
= ZFS_OBJ_MTX_SZ
;
96 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
100 inode_init_once(ZTOI(zp
));
101 list_link_init(&zp
->z_link_node
);
103 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
104 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
105 rw_init(&zp
->z_name_lock
, NULL
, RW_NOLOCKDEP
, NULL
);
106 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
107 rw_init(&zp
->z_xattr_lock
, NULL
, RW_DEFAULT
, NULL
);
109 zfs_rlock_init(&zp
->z_range_lock
);
111 zp
->z_dirlocks
= NULL
;
112 zp
->z_acl_cached
= NULL
;
113 zp
->z_xattr_cached
= NULL
;
114 zp
->z_xattr_parent
= 0;
121 zfs_znode_cache_destructor(void *buf
, void *arg
)
125 ASSERT(!list_link_active(&zp
->z_link_node
));
126 mutex_destroy(&zp
->z_lock
);
127 rw_destroy(&zp
->z_parent_lock
);
128 rw_destroy(&zp
->z_name_lock
);
129 mutex_destroy(&zp
->z_acl_lock
);
130 rw_destroy(&zp
->z_xattr_lock
);
131 zfs_rlock_destroy(&zp
->z_range_lock
);
133 ASSERT(zp
->z_dirlocks
== NULL
);
134 ASSERT(zp
->z_acl_cached
== NULL
);
135 ASSERT(zp
->z_xattr_cached
== NULL
);
139 zfs_znode_hold_cache_constructor(void *buf
, void *arg
, int kmflags
)
141 znode_hold_t
*zh
= buf
;
143 mutex_init(&zh
->zh_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
144 refcount_create(&zh
->zh_refcount
);
145 zh
->zh_obj
= ZFS_NO_OBJECT
;
151 zfs_znode_hold_cache_destructor(void *buf
, void *arg
)
153 znode_hold_t
*zh
= buf
;
155 mutex_destroy(&zh
->zh_lock
);
156 refcount_destroy(&zh
->zh_refcount
);
163 * Initialize zcache. The KMC_SLAB hint is used in order that it be
164 * backed by kmalloc() when on the Linux slab in order that any
165 * wait_on_bit() operations on the related inode operate properly.
167 ASSERT(znode_cache
== NULL
);
168 znode_cache
= kmem_cache_create("zfs_znode_cache",
169 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
170 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, KMC_SLAB
);
172 ASSERT(znode_hold_cache
== NULL
);
173 znode_hold_cache
= kmem_cache_create("zfs_znode_hold_cache",
174 sizeof (znode_hold_t
), 0, zfs_znode_hold_cache_constructor
,
175 zfs_znode_hold_cache_destructor
, NULL
, NULL
, NULL
, 0);
185 kmem_cache_destroy(znode_cache
);
188 if (znode_hold_cache
)
189 kmem_cache_destroy(znode_hold_cache
);
190 znode_hold_cache
= NULL
;
194 * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to
195 * serialize access to a znode and its SA buffer while the object is being
196 * created or destroyed. This kind of locking would normally reside in the
197 * znode itself but in this case that's impossible because the znode and SA
198 * buffer may not yet exist. Therefore the locking is handled externally
199 * with an array of mutexs and AVLs trees which contain per-object locks.
201 * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted
202 * in to the correct AVL tree and finally the per-object lock is held. In
203 * zfs_znode_hold_exit() the process is reversed. The per-object lock is
204 * released, removed from the AVL tree and destroyed if there are no waiters.
206 * This scheme has two important properties:
208 * 1) No memory allocations are performed while holding one of the z_hold_locks.
209 * This ensures evict(), which can be called from direct memory reclaim, will
210 * never block waiting on a z_hold_locks which just happens to have hashed
213 * 2) All locks used to serialize access to an object are per-object and never
214 * shared. This minimizes lock contention without creating a large number
215 * of dedicated locks.
217 * On the downside it does require znode_lock_t structures to be frequently
218 * allocated and freed. However, because these are backed by a kmem cache
219 * and very short lived this cost is minimal.
222 zfs_znode_hold_compare(const void *a
, const void *b
)
224 const znode_hold_t
*zh_a
= (const znode_hold_t
*)a
;
225 const znode_hold_t
*zh_b
= (const znode_hold_t
*)b
;
227 return (AVL_CMP(zh_a
->zh_obj
, zh_b
->zh_obj
));
231 zfs_znode_held(zfsvfs_t
*zfsvfs
, uint64_t obj
)
233 znode_hold_t
*zh
, search
;
234 int i
= ZFS_OBJ_HASH(zfsvfs
, obj
);
239 mutex_enter(&zfsvfs
->z_hold_locks
[i
]);
240 zh
= avl_find(&zfsvfs
->z_hold_trees
[i
], &search
, NULL
);
241 held
= (zh
&& MUTEX_HELD(&zh
->zh_lock
)) ? B_TRUE
: B_FALSE
;
242 mutex_exit(&zfsvfs
->z_hold_locks
[i
]);
247 static znode_hold_t
*
248 zfs_znode_hold_enter(zfsvfs_t
*zfsvfs
, uint64_t obj
)
250 znode_hold_t
*zh
, *zh_new
, search
;
251 int i
= ZFS_OBJ_HASH(zfsvfs
, obj
);
252 boolean_t found
= B_FALSE
;
254 zh_new
= kmem_cache_alloc(znode_hold_cache
, KM_SLEEP
);
255 zh_new
->zh_obj
= obj
;
258 mutex_enter(&zfsvfs
->z_hold_locks
[i
]);
259 zh
= avl_find(&zfsvfs
->z_hold_trees
[i
], &search
, NULL
);
260 if (likely(zh
== NULL
)) {
262 avl_add(&zfsvfs
->z_hold_trees
[i
], zh
);
264 ASSERT3U(zh
->zh_obj
, ==, obj
);
267 refcount_add(&zh
->zh_refcount
, NULL
);
268 mutex_exit(&zfsvfs
->z_hold_locks
[i
]);
271 kmem_cache_free(znode_hold_cache
, zh_new
);
273 ASSERT(MUTEX_NOT_HELD(&zh
->zh_lock
));
274 ASSERT3S(refcount_count(&zh
->zh_refcount
), >, 0);
275 mutex_enter(&zh
->zh_lock
);
281 zfs_znode_hold_exit(zfsvfs_t
*zfsvfs
, znode_hold_t
*zh
)
283 int i
= ZFS_OBJ_HASH(zfsvfs
, zh
->zh_obj
);
284 boolean_t remove
= B_FALSE
;
286 ASSERT(zfs_znode_held(zfsvfs
, zh
->zh_obj
));
287 ASSERT3S(refcount_count(&zh
->zh_refcount
), >, 0);
288 mutex_exit(&zh
->zh_lock
);
290 mutex_enter(&zfsvfs
->z_hold_locks
[i
]);
291 if (refcount_remove(&zh
->zh_refcount
, NULL
) == 0) {
292 avl_remove(&zfsvfs
->z_hold_trees
[i
], zh
);
295 mutex_exit(&zfsvfs
->z_hold_locks
[i
]);
297 if (remove
== B_TRUE
)
298 kmem_cache_free(znode_hold_cache
, zh
);
302 zfs_create_share_dir(zfsvfs_t
*zfsvfs
, dmu_tx_t
*tx
)
304 #ifdef HAVE_SMB_SHARE
305 zfs_acl_ids_t acl_ids
;
312 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
313 vattr
.va_mode
= S_IFDIR
| 0555;
314 vattr
.va_uid
= crgetuid(kcred
);
315 vattr
.va_gid
= crgetgid(kcred
);
317 sharezp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
318 sharezp
->z_moved
= 0;
319 sharezp
->z_unlinked
= 0;
320 sharezp
->z_atime_dirty
= 0;
321 sharezp
->z_zfsvfs
= zfsvfs
;
322 sharezp
->z_is_sa
= zfsvfs
->z_use_sa
;
323 sharezp
->z_pflags
= 0;
329 VERIFY(0 == zfs_acl_ids_create(sharezp
, IS_ROOT_NODE
, &vattr
,
330 kcred
, NULL
, &acl_ids
));
331 zfs_mknode(sharezp
, &vattr
, tx
, kcred
, IS_ROOT_NODE
, &zp
, &acl_ids
);
332 ASSERT3P(zp
, ==, sharezp
);
333 ASSERT(!vn_in_dnlc(ZTOV(sharezp
))); /* not valid to move */
334 POINTER_INVALIDATE(&sharezp
->z_zfsvfs
);
335 error
= zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
336 ZFS_SHARES_DIR
, 8, 1, &sharezp
->z_id
, tx
);
337 zfsvfs
->z_shares_dir
= sharezp
->z_id
;
339 zfs_acl_ids_free(&acl_ids
);
340 // ZTOV(sharezp)->v_count = 0;
341 sa_handle_destroy(sharezp
->z_sa_hdl
);
342 kmem_cache_free(znode_cache
, sharezp
);
347 #endif /* HAVE_SMB_SHARE */
351 zfs_znode_sa_init(zfsvfs_t
*zfsvfs
, znode_t
*zp
,
352 dmu_buf_t
*db
, dmu_object_type_t obj_type
, sa_handle_t
*sa_hdl
)
354 ASSERT(zfs_znode_held(zfsvfs
, zp
->z_id
));
356 mutex_enter(&zp
->z_lock
);
358 ASSERT(zp
->z_sa_hdl
== NULL
);
359 ASSERT(zp
->z_acl_cached
== NULL
);
360 if (sa_hdl
== NULL
) {
361 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, zp
,
362 SA_HDL_SHARED
, &zp
->z_sa_hdl
));
364 zp
->z_sa_hdl
= sa_hdl
;
365 sa_set_userp(sa_hdl
, zp
);
368 zp
->z_is_sa
= (obj_type
== DMU_OT_SA
) ? B_TRUE
: B_FALSE
;
370 mutex_exit(&zp
->z_lock
);
374 zfs_znode_dmu_fini(znode_t
*zp
)
376 ASSERT(zfs_znode_held(ZTOZSB(zp
), zp
->z_id
) || zp
->z_unlinked
||
377 RW_WRITE_HELD(&ZTOZSB(zp
)->z_teardown_inactive_lock
));
379 sa_handle_destroy(zp
->z_sa_hdl
);
384 * Called by new_inode() to allocate a new inode.
387 zfs_inode_alloc(struct super_block
*sb
, struct inode
**ip
)
391 zp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
398 * Called in multiple places when an inode should be destroyed.
401 zfs_inode_destroy(struct inode
*ip
)
403 znode_t
*zp
= ITOZ(ip
);
404 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
406 mutex_enter(&zfsvfs
->z_znodes_lock
);
407 if (list_link_active(&zp
->z_link_node
)) {
408 list_remove(&zfsvfs
->z_all_znodes
, zp
);
409 zfsvfs
->z_nr_znodes
--;
411 mutex_exit(&zfsvfs
->z_znodes_lock
);
413 if (zp
->z_acl_cached
) {
414 zfs_acl_free(zp
->z_acl_cached
);
415 zp
->z_acl_cached
= NULL
;
418 if (zp
->z_xattr_cached
) {
419 nvlist_free(zp
->z_xattr_cached
);
420 zp
->z_xattr_cached
= NULL
;
423 kmem_cache_free(znode_cache
, zp
);
427 zfs_inode_set_ops(zfsvfs_t
*zfsvfs
, struct inode
*ip
)
431 switch (ip
->i_mode
& S_IFMT
) {
433 ip
->i_op
= &zpl_inode_operations
;
434 ip
->i_fop
= &zpl_file_operations
;
435 ip
->i_mapping
->a_ops
= &zpl_address_space_operations
;
439 ip
->i_op
= &zpl_dir_inode_operations
;
440 ip
->i_fop
= &zpl_dir_file_operations
;
441 ITOZ(ip
)->z_zn_prefetch
= B_TRUE
;
445 ip
->i_op
= &zpl_symlink_inode_operations
;
449 * rdev is only stored in a SA only for device files.
453 (void) sa_lookup(ITOZ(ip
)->z_sa_hdl
, SA_ZPL_RDEV(zfsvfs
), &rdev
,
458 init_special_inode(ip
, ip
->i_mode
, rdev
);
459 ip
->i_op
= &zpl_special_inode_operations
;
463 zfs_panic_recover("inode %llu has invalid mode: 0x%x\n",
464 (u_longlong_t
)ip
->i_ino
, ip
->i_mode
);
466 /* Assume the inode is a file and attempt to continue */
467 ip
->i_mode
= S_IFREG
| 0644;
468 ip
->i_op
= &zpl_inode_operations
;
469 ip
->i_fop
= &zpl_file_operations
;
470 ip
->i_mapping
->a_ops
= &zpl_address_space_operations
;
476 zfs_set_inode_flags(znode_t
*zp
, struct inode
*ip
)
479 * Linux and Solaris have different sets of file attributes, so we
480 * restrict this conversion to the intersection of the two.
482 #ifdef HAVE_INODE_SET_FLAGS
483 unsigned int flags
= 0;
484 if (zp
->z_pflags
& ZFS_IMMUTABLE
)
485 flags
|= S_IMMUTABLE
;
486 if (zp
->z_pflags
& ZFS_APPENDONLY
)
489 inode_set_flags(ip
, flags
, S_IMMUTABLE
|S_APPEND
);
491 if (zp
->z_pflags
& ZFS_IMMUTABLE
)
492 ip
->i_flags
|= S_IMMUTABLE
;
494 ip
->i_flags
&= ~S_IMMUTABLE
;
496 if (zp
->z_pflags
& ZFS_APPENDONLY
)
497 ip
->i_flags
|= S_APPEND
;
499 ip
->i_flags
&= ~S_APPEND
;
504 * Update the embedded inode given the znode. We should work toward
505 * eliminating this function as soon as possible by removing values
506 * which are duplicated between the znode and inode. If the generic
507 * inode has the correct field it should be used, and the ZFS code
508 * updated to access the inode. This can be done incrementally.
511 zfs_inode_update(znode_t
*zp
)
516 u_longlong_t i_blocks
;
522 /* Skip .zfs control nodes which do not exist on disk. */
523 if (zfsctl_is_node(ip
))
526 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &blksize
, &i_blocks
);
528 spin_lock(&ip
->i_lock
);
529 ip
->i_blocks
= i_blocks
;
530 i_size_write(ip
, zp
->z_size
);
531 spin_unlock(&ip
->i_lock
);
536 * Construct a znode+inode and initialize.
538 * This does not do a call to dmu_set_user() that is
539 * up to the caller to do, in case you don't want to
543 zfs_znode_alloc(zfsvfs_t
*zfsvfs
, dmu_buf_t
*db
, int blksz
,
544 dmu_object_type_t obj_type
, uint64_t obj
, sa_handle_t
*hdl
)
552 uint64_t z_uid
, z_gid
;
553 uint64_t atime
[2], mtime
[2], ctime
[2];
554 uint64_t projid
= ZFS_DEFAULT_PROJID
;
555 sa_bulk_attr_t bulk
[11];
558 ASSERT(zfsvfs
!= NULL
);
560 ip
= new_inode(zfsvfs
->z_sb
);
565 ASSERT(zp
->z_dirlocks
== NULL
);
566 ASSERT3P(zp
->z_acl_cached
, ==, NULL
);
567 ASSERT3P(zp
->z_xattr_cached
, ==, NULL
);
571 zp
->z_atime_dirty
= 0;
573 zp
->z_id
= db
->db_object
;
575 zp
->z_seq
= 0x7A4653;
577 zp
->z_is_mapped
= B_FALSE
;
578 zp
->z_is_ctldir
= B_FALSE
;
579 zp
->z_is_stale
= B_FALSE
;
580 zp
->z_range_lock
.zr_size
= &zp
->z_size
;
581 zp
->z_range_lock
.zr_blksz
= &zp
->z_blksz
;
582 zp
->z_range_lock
.zr_max_blksz
= &ZTOZSB(zp
)->z_max_blksz
;
584 zfs_znode_sa_init(zfsvfs
, zp
, db
, obj_type
, hdl
);
586 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
, &mode
, 8);
587 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
, &tmp_gen
, 8);
588 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
590 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
, &links
, 8);
591 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
593 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zfsvfs
), NULL
,
595 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
, &z_uid
, 8);
596 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
, &z_gid
, 8);
597 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
, &atime
, 16);
598 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
, &mtime
, 16);
599 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
, &ctime
, 16);
601 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
) != 0 || tmp_gen
== 0 ||
602 (dmu_objset_projectquota_enabled(zfsvfs
->z_os
) &&
603 (zp
->z_pflags
& ZFS_PROJID
) &&
604 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_PROJID(zfsvfs
), &projid
, 8) != 0)) {
606 sa_handle_destroy(zp
->z_sa_hdl
);
611 zp
->z_projid
= projid
;
612 zp
->z_mode
= ip
->i_mode
= mode
;
613 ip
->i_generation
= (uint32_t)tmp_gen
;
614 ip
->i_blkbits
= SPA_MINBLOCKSHIFT
;
615 set_nlink(ip
, (uint32_t)links
);
616 zfs_uid_write(ip
, z_uid
);
617 zfs_gid_write(ip
, z_gid
);
618 zfs_set_inode_flags(zp
, ip
);
620 /* Cache the xattr parent id */
621 if (zp
->z_pflags
& ZFS_XATTR
)
622 zp
->z_xattr_parent
= parent
;
624 ZFS_TIME_DECODE(&ip
->i_atime
, atime
);
625 ZFS_TIME_DECODE(&ip
->i_mtime
, mtime
);
626 ZFS_TIME_DECODE(&ip
->i_ctime
, ctime
);
629 zfs_inode_update(zp
);
630 zfs_inode_set_ops(zfsvfs
, ip
);
633 * The only way insert_inode_locked() can fail is if the ip->i_ino
634 * number is already hashed for this super block. This can never
635 * happen because the inode numbers map 1:1 with the object numbers.
637 * The one exception is rolling back a mounted file system, but in
638 * this case all the active inode are unhashed during the rollback.
640 VERIFY3S(insert_inode_locked(ip
), ==, 0);
642 mutex_enter(&zfsvfs
->z_znodes_lock
);
643 list_insert_tail(&zfsvfs
->z_all_znodes
, zp
);
644 zfsvfs
->z_nr_znodes
++;
646 mutex_exit(&zfsvfs
->z_znodes_lock
);
648 unlock_new_inode(ip
);
657 * Safely mark an inode dirty. Inodes which are part of a read-only
658 * file system or snapshot may not be dirtied.
661 zfs_mark_inode_dirty(struct inode
*ip
)
663 zfsvfs_t
*zfsvfs
= ITOZSB(ip
);
665 if (zfs_is_readonly(zfsvfs
) || dmu_objset_is_snapshot(zfsvfs
->z_os
))
668 mark_inode_dirty(ip
);
671 static uint64_t empty_xattr
;
672 static uint64_t pad
[4];
673 static zfs_acl_phys_t acl_phys
;
675 * Create a new DMU object to hold a zfs znode.
677 * IN: dzp - parent directory for new znode
678 * vap - file attributes for new znode
679 * tx - dmu transaction id for zap operations
680 * cr - credentials of caller
682 * IS_ROOT_NODE - new object will be root
683 * IS_XATTR - new object is an attribute
684 * bonuslen - length of bonus buffer
685 * setaclp - File/Dir initial ACL
686 * fuidp - Tracks fuid allocation.
688 * OUT: zpp - allocated znode
692 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
693 uint_t flag
, znode_t
**zpp
, zfs_acl_ids_t
*acl_ids
)
695 uint64_t crtime
[2], atime
[2], mtime
[2], ctime
[2];
696 uint64_t mode
, size
, links
, parent
, pflags
;
697 uint64_t projid
= ZFS_DEFAULT_PROJID
;
699 zfsvfs_t
*zfsvfs
= ZTOZSB(dzp
);
701 inode_timespec_t now
;
706 dmu_object_type_t obj_type
;
707 sa_bulk_attr_t
*sa_attrs
;
709 zfs_acl_locator_cb_t locate
= { 0 };
712 if (zfsvfs
->z_replay
) {
713 obj
= vap
->va_nodeid
;
714 now
= vap
->va_ctime
; /* see zfs_replay_create() */
715 gen
= vap
->va_nblocks
; /* ditto */
716 dnodesize
= vap
->va_fsid
; /* ditto */
720 gen
= dmu_tx_get_txg(tx
);
721 dnodesize
= dmu_objset_dnodesize(zfsvfs
->z_os
);
725 dnodesize
= DNODE_MIN_SIZE
;
727 obj_type
= zfsvfs
->z_use_sa
? DMU_OT_SA
: DMU_OT_ZNODE
;
729 bonuslen
= (obj_type
== DMU_OT_SA
) ?
730 DN_BONUS_SIZE(dnodesize
) : ZFS_OLD_ZNODE_PHYS_SIZE
;
733 * Create a new DMU object.
736 * There's currently no mechanism for pre-reading the blocks that will
737 * be needed to allocate a new object, so we accept the small chance
738 * that there will be an i/o error and we will fail one of the
741 if (S_ISDIR(vap
->va_mode
)) {
742 if (zfsvfs
->z_replay
) {
743 VERIFY0(zap_create_claim_norm_dnsize(zfsvfs
->z_os
, obj
,
744 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
745 obj_type
, bonuslen
, dnodesize
, tx
));
747 obj
= zap_create_norm_dnsize(zfsvfs
->z_os
,
748 zfsvfs
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
749 obj_type
, bonuslen
, dnodesize
, tx
);
752 if (zfsvfs
->z_replay
) {
753 VERIFY0(dmu_object_claim_dnsize(zfsvfs
->z_os
, obj
,
754 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
755 obj_type
, bonuslen
, dnodesize
, tx
));
757 obj
= dmu_object_alloc_dnsize(zfsvfs
->z_os
,
758 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
759 obj_type
, bonuslen
, dnodesize
, tx
);
763 zh
= zfs_znode_hold_enter(zfsvfs
, obj
);
764 VERIFY0(sa_buf_hold(zfsvfs
->z_os
, obj
, NULL
, &db
));
767 * If this is the root, fix up the half-initialized parent pointer
768 * to reference the just-allocated physical data area.
770 if (flag
& IS_ROOT_NODE
) {
775 * If parent is an xattr, so am I.
777 if (dzp
->z_pflags
& ZFS_XATTR
) {
781 if (zfsvfs
->z_use_fuids
)
782 pflags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
786 if (S_ISDIR(vap
->va_mode
)) {
787 size
= 2; /* contents ("." and "..") */
791 links
= (flag
& IS_TMPFILE
) ? 0 : 1;
794 if (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))
798 mode
= acl_ids
->z_mode
;
802 if (S_ISREG(vap
->va_mode
) || S_ISDIR(vap
->va_mode
)) {
804 * With ZFS_PROJID flag, we can easily know whether there is
805 * project ID stored on disk or not. See zfs_space_delta_cb().
807 if (obj_type
!= DMU_OT_ZNODE
&&
808 dmu_objset_projectquota_enabled(zfsvfs
->z_os
))
809 pflags
|= ZFS_PROJID
;
812 * Inherit project ID from parent if required.
814 projid
= zfs_inherit_projid(dzp
);
815 if (dzp
->z_pflags
& ZFS_PROJINHERIT
)
816 pflags
|= ZFS_PROJINHERIT
;
820 * No execs denied will be deterimed when zfs_mode_compute() is called.
822 pflags
|= acl_ids
->z_aclp
->z_hints
&
823 (ZFS_ACL_TRIVIAL
|ZFS_INHERIT_ACE
|ZFS_ACL_AUTO_INHERIT
|
824 ZFS_ACL_DEFAULTED
|ZFS_ACL_PROTECTED
);
826 ZFS_TIME_ENCODE(&now
, crtime
);
827 ZFS_TIME_ENCODE(&now
, ctime
);
829 if (vap
->va_mask
& ATTR_ATIME
) {
830 ZFS_TIME_ENCODE(&vap
->va_atime
, atime
);
832 ZFS_TIME_ENCODE(&now
, atime
);
835 if (vap
->va_mask
& ATTR_MTIME
) {
836 ZFS_TIME_ENCODE(&vap
->va_mtime
, mtime
);
838 ZFS_TIME_ENCODE(&now
, mtime
);
841 /* Now add in all of the "SA" attributes */
842 VERIFY(0 == sa_handle_get_from_db(zfsvfs
->z_os
, db
, NULL
, SA_HDL_SHARED
,
846 * Setup the array of attributes to be replaced/set on the new file
848 * order for DMU_OT_ZNODE is critical since it needs to be constructed
849 * in the old znode_phys_t format. Don't change this ordering
851 sa_attrs
= kmem_alloc(sizeof (sa_bulk_attr_t
) * ZPL_END
, KM_SLEEP
);
853 if (obj_type
== DMU_OT_ZNODE
) {
854 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
856 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
858 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
860 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
862 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
864 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
866 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
868 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
871 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zfsvfs
),
873 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zfsvfs
),
875 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zfsvfs
),
877 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
),
878 NULL
, &acl_ids
->z_fuid
, 8);
879 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
),
880 NULL
, &acl_ids
->z_fgid
, 8);
881 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zfsvfs
),
883 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
885 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zfsvfs
),
887 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zfsvfs
),
889 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zfsvfs
),
891 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zfsvfs
),
895 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zfsvfs
), NULL
, &links
, 8);
897 if (obj_type
== DMU_OT_ZNODE
) {
898 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zfsvfs
), NULL
,
900 } else if (dmu_objset_projectquota_enabled(zfsvfs
->z_os
) &&
901 pflags
& ZFS_PROJID
) {
902 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PROJID(zfsvfs
),
905 if (obj_type
== DMU_OT_ZNODE
||
906 (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))) {
907 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zfsvfs
),
910 if (obj_type
== DMU_OT_ZNODE
) {
911 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zfsvfs
),
913 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zfsvfs
), NULL
,
914 &acl_ids
->z_fuid
, 8);
915 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zfsvfs
), NULL
,
916 &acl_ids
->z_fgid
, 8);
917 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zfsvfs
), NULL
, pad
,
918 sizeof (uint64_t) * 4);
919 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zfsvfs
), NULL
,
920 &acl_phys
, sizeof (zfs_acl_phys_t
));
921 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
922 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zfsvfs
), NULL
,
923 &acl_ids
->z_aclp
->z_acl_count
, 8);
924 locate
.cb_aclp
= acl_ids
->z_aclp
;
925 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zfsvfs
),
926 zfs_acl_data_locator
, &locate
,
927 acl_ids
->z_aclp
->z_acl_bytes
);
928 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
929 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
932 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
934 if (!(flag
& IS_ROOT_NODE
)) {
936 * The call to zfs_znode_alloc() may fail if memory is low
937 * via the call path: alloc_inode() -> inode_init_always() ->
938 * security_inode_alloc() -> inode_alloc_security(). Since
939 * the existing code is written such that zfs_mknode() can
940 * not fail retry until sufficient memory has been reclaimed.
943 *zpp
= zfs_znode_alloc(zfsvfs
, db
, 0, obj_type
, obj
,
945 } while (*zpp
== NULL
);
947 VERIFY(*zpp
!= NULL
);
951 * If we are creating the root node, the "parent" we
952 * passed in is the znode for the root.
956 (*zpp
)->z_sa_hdl
= sa_hdl
;
959 (*zpp
)->z_pflags
= pflags
;
960 (*zpp
)->z_mode
= ZTOI(*zpp
)->i_mode
= mode
;
961 (*zpp
)->z_dnodesize
= dnodesize
;
962 (*zpp
)->z_projid
= projid
;
964 if (obj_type
== DMU_OT_ZNODE
||
965 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
966 VERIFY0(zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
));
968 kmem_free(sa_attrs
, sizeof (sa_bulk_attr_t
) * ZPL_END
);
969 zfs_znode_hold_exit(zfsvfs
, zh
);
973 * Update in-core attributes. It is assumed the caller will be doing an
974 * sa_bulk_update to push the changes out.
977 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
980 boolean_t update_inode
= B_FALSE
;
982 xoap
= xva_getxoptattr(xvap
);
985 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
987 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
988 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(ZTOZSB(zp
)),
989 ×
, sizeof (times
), tx
);
990 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
992 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
993 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
995 XVA_SET_RTN(xvap
, XAT_READONLY
);
997 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
998 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
1000 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
1002 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
1003 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
1005 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
1007 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
1008 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
1010 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
1012 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
1013 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
1015 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
1017 update_inode
= B_TRUE
;
1019 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
1020 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
1022 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
1024 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
1025 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
1027 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
1029 update_inode
= B_TRUE
;
1031 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
1032 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
1034 XVA_SET_RTN(xvap
, XAT_NODUMP
);
1036 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
1037 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
1039 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
1041 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
1042 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
1043 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
1044 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
1046 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
1047 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
1049 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
1051 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
1052 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
1053 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
1055 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
1056 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
1058 XVA_SET_RTN(xvap
, XAT_REPARSE
);
1060 if (XVA_ISSET_REQ(xvap
, XAT_OFFLINE
)) {
1061 ZFS_ATTR_SET(zp
, ZFS_OFFLINE
, xoap
->xoa_offline
,
1063 XVA_SET_RTN(xvap
, XAT_OFFLINE
);
1065 if (XVA_ISSET_REQ(xvap
, XAT_SPARSE
)) {
1066 ZFS_ATTR_SET(zp
, ZFS_SPARSE
, xoap
->xoa_sparse
,
1068 XVA_SET_RTN(xvap
, XAT_SPARSE
);
1070 if (XVA_ISSET_REQ(xvap
, XAT_PROJINHERIT
)) {
1071 ZFS_ATTR_SET(zp
, ZFS_PROJINHERIT
, xoap
->xoa_projinherit
,
1073 XVA_SET_RTN(xvap
, XAT_PROJINHERIT
);
1077 zfs_set_inode_flags(zp
, ZTOI(zp
));
1081 zfs_zget(zfsvfs_t
*zfsvfs
, uint64_t obj_num
, znode_t
**zpp
)
1083 dmu_object_info_t doi
;
1093 zh
= zfs_znode_hold_enter(zfsvfs
, obj_num
);
1095 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1097 zfs_znode_hold_exit(zfsvfs
, zh
);
1101 dmu_object_info_from_db(db
, &doi
);
1102 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1103 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1104 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1105 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1106 sa_buf_rele(db
, NULL
);
1107 zfs_znode_hold_exit(zfsvfs
, zh
);
1108 return (SET_ERROR(EINVAL
));
1111 hdl
= dmu_buf_get_user(db
);
1113 zp
= sa_get_userdata(hdl
);
1117 * Since "SA" does immediate eviction we
1118 * should never find a sa handle that doesn't
1119 * know about the znode.
1122 ASSERT3P(zp
, !=, NULL
);
1124 mutex_enter(&zp
->z_lock
);
1125 ASSERT3U(zp
->z_id
, ==, obj_num
);
1127 * If igrab() returns NULL the VFS has independently
1128 * determined the inode should be evicted and has
1129 * called iput_final() to start the eviction process.
1130 * The SA handle is still valid but because the VFS
1131 * requires that the eviction succeed we must drop
1132 * our locks and references to allow the eviction to
1133 * complete. The zfs_zget() may then be retried.
1135 * This unlikely case could be optimized by registering
1136 * a sops->drop_inode() callback. The callback would
1137 * need to detect the active SA hold thereby informing
1138 * the VFS that this inode should not be evicted.
1140 if (igrab(ZTOI(zp
)) == NULL
) {
1141 mutex_exit(&zp
->z_lock
);
1142 sa_buf_rele(db
, NULL
);
1143 zfs_znode_hold_exit(zfsvfs
, zh
);
1144 /* inode might need this to finish evict */
1150 mutex_exit(&zp
->z_lock
);
1151 sa_buf_rele(db
, NULL
);
1152 zfs_znode_hold_exit(zfsvfs
, zh
);
1157 * Not found create new znode/vnode but only if file exists.
1159 * There is a small window where zfs_vget() could
1160 * find this object while a file create is still in
1161 * progress. This is checked for in zfs_znode_alloc()
1163 * if zfs_znode_alloc() fails it will drop the hold on the
1166 zp
= zfs_znode_alloc(zfsvfs
, db
, doi
.doi_data_block_size
,
1167 doi
.doi_bonus_type
, obj_num
, NULL
);
1169 err
= SET_ERROR(ENOENT
);
1173 zfs_znode_hold_exit(zfsvfs
, zh
);
1178 zfs_rezget(znode_t
*zp
)
1180 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1181 dmu_object_info_t doi
;
1183 uint64_t obj_num
= zp
->z_id
;
1186 sa_bulk_attr_t bulk
[10];
1190 uint64_t z_uid
, z_gid
;
1191 uint64_t atime
[2], mtime
[2], ctime
[2];
1192 uint64_t projid
= ZFS_DEFAULT_PROJID
;
1196 * skip ctldir, otherwise they will always get invalidated. This will
1197 * cause funny behaviour for the mounted snapdirs. Especially for
1198 * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent
1199 * anyone automount it again as long as someone is still using the
1202 if (zp
->z_is_ctldir
)
1205 zh
= zfs_znode_hold_enter(zfsvfs
, obj_num
);
1207 mutex_enter(&zp
->z_acl_lock
);
1208 if (zp
->z_acl_cached
) {
1209 zfs_acl_free(zp
->z_acl_cached
);
1210 zp
->z_acl_cached
= NULL
;
1212 mutex_exit(&zp
->z_acl_lock
);
1214 rw_enter(&zp
->z_xattr_lock
, RW_WRITER
);
1215 if (zp
->z_xattr_cached
) {
1216 nvlist_free(zp
->z_xattr_cached
);
1217 zp
->z_xattr_cached
= NULL
;
1219 rw_exit(&zp
->z_xattr_lock
);
1221 ASSERT(zp
->z_sa_hdl
== NULL
);
1222 err
= sa_buf_hold(zfsvfs
->z_os
, obj_num
, NULL
, &db
);
1224 zfs_znode_hold_exit(zfsvfs
, zh
);
1228 dmu_object_info_from_db(db
, &doi
);
1229 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
1230 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
1231 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1232 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
1233 sa_buf_rele(db
, NULL
);
1234 zfs_znode_hold_exit(zfsvfs
, zh
);
1235 return (SET_ERROR(EINVAL
));
1238 zfs_znode_sa_init(zfsvfs
, zp
, db
, doi
.doi_bonus_type
, NULL
);
1240 /* reload cached values */
1241 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zfsvfs
), NULL
,
1242 &gen
, sizeof (gen
));
1243 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
), NULL
,
1244 &zp
->z_size
, sizeof (zp
->z_size
));
1245 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zfsvfs
), NULL
,
1246 &links
, sizeof (links
));
1247 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
), NULL
,
1248 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1249 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zfsvfs
), NULL
,
1250 &z_uid
, sizeof (z_uid
));
1251 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zfsvfs
), NULL
,
1252 &z_gid
, sizeof (z_gid
));
1253 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zfsvfs
), NULL
,
1254 &mode
, sizeof (mode
));
1255 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zfsvfs
), NULL
,
1257 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
,
1259 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
,
1262 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
1263 zfs_znode_dmu_fini(zp
);
1264 zfs_znode_hold_exit(zfsvfs
, zh
);
1265 return (SET_ERROR(EIO
));
1268 if (dmu_objset_projectquota_enabled(zfsvfs
->z_os
)) {
1269 err
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_PROJID(zfsvfs
),
1271 if (err
!= 0 && err
!= ENOENT
) {
1272 zfs_znode_dmu_fini(zp
);
1273 zfs_znode_hold_exit(zfsvfs
, zh
);
1274 return (SET_ERROR(err
));
1278 zp
->z_projid
= projid
;
1279 zp
->z_mode
= ZTOI(zp
)->i_mode
= mode
;
1280 zfs_uid_write(ZTOI(zp
), z_uid
);
1281 zfs_gid_write(ZTOI(zp
), z_gid
);
1283 ZFS_TIME_DECODE(&ZTOI(zp
)->i_atime
, atime
);
1284 ZFS_TIME_DECODE(&ZTOI(zp
)->i_mtime
, mtime
);
1285 ZFS_TIME_DECODE(&ZTOI(zp
)->i_ctime
, ctime
);
1287 if (gen
!= ZTOI(zp
)->i_generation
) {
1288 zfs_znode_dmu_fini(zp
);
1289 zfs_znode_hold_exit(zfsvfs
, zh
);
1290 return (SET_ERROR(EIO
));
1293 set_nlink(ZTOI(zp
), (uint32_t)links
);
1294 zfs_set_inode_flags(zp
, ZTOI(zp
));
1296 zp
->z_blksz
= doi
.doi_data_block_size
;
1297 zp
->z_atime_dirty
= 0;
1298 zfs_inode_update(zp
);
1301 * If the file has zero links, then it has been unlinked on the send
1302 * side and it must be in the received unlinked set.
1303 * We call zfs_znode_dmu_fini() now to prevent any accesses to the
1304 * stale data and to prevent automatical removal of the file in
1305 * zfs_zinactive(). The file will be removed either when it is removed
1306 * on the send side and the next incremental stream is received or
1307 * when the unlinked set gets processed.
1309 zp
->z_unlinked
= (ZTOI(zp
)->i_nlink
== 0);
1311 zfs_znode_dmu_fini(zp
);
1313 zfs_znode_hold_exit(zfsvfs
, zh
);
1319 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1321 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1322 objset_t
*os
= zfsvfs
->z_os
;
1323 uint64_t obj
= zp
->z_id
;
1324 uint64_t acl_obj
= zfs_external_acl(zp
);
1327 zh
= zfs_znode_hold_enter(zfsvfs
, obj
);
1329 VERIFY(!zp
->z_is_sa
);
1330 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1332 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1333 zfs_znode_dmu_fini(zp
);
1334 zfs_znode_hold_exit(zfsvfs
, zh
);
1338 zfs_zinactive(znode_t
*zp
)
1340 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1341 uint64_t z_id
= zp
->z_id
;
1344 ASSERT(zp
->z_sa_hdl
);
1347 * Don't allow a zfs_zget() while were trying to release this znode.
1349 zh
= zfs_znode_hold_enter(zfsvfs
, z_id
);
1351 mutex_enter(&zp
->z_lock
);
1354 * If this was the last reference to a file with no links, remove
1355 * the file from the file system unless the file system is mounted
1356 * read-only. That can happen, for example, if the file system was
1357 * originally read-write, the file was opened, then unlinked and
1358 * the file system was made read-only before the file was finally
1359 * closed. The file will remain in the unlinked set.
1361 if (zp
->z_unlinked
) {
1362 ASSERT(!zfsvfs
->z_issnap
);
1363 if (!zfs_is_readonly(zfsvfs
)) {
1364 mutex_exit(&zp
->z_lock
);
1365 zfs_znode_hold_exit(zfsvfs
, zh
);
1371 mutex_exit(&zp
->z_lock
);
1372 zfs_znode_dmu_fini(zp
);
1374 zfs_znode_hold_exit(zfsvfs
, zh
);
1378 zfs_compare_timespec(struct timespec
*t1
, struct timespec
*t2
)
1380 if (t1
->tv_sec
< t2
->tv_sec
)
1383 if (t1
->tv_sec
> t2
->tv_sec
)
1386 return (t1
->tv_nsec
- t2
->tv_nsec
);
1390 * Prepare to update znode time stamps.
1392 * IN: zp - znode requiring timestamp update
1393 * flag - ATTR_MTIME, ATTR_CTIME flags
1399 * Note: We don't update atime here, because we rely on Linux VFS to do
1403 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1406 inode_timespec_t now
;
1412 if (flag
& ATTR_MTIME
) {
1413 ZFS_TIME_ENCODE(&now
, mtime
);
1414 ZFS_TIME_DECODE(&(ZTOI(zp
)->i_mtime
), mtime
);
1415 if (ZTOZSB(zp
)->z_use_fuids
) {
1416 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1421 if (flag
& ATTR_CTIME
) {
1422 ZFS_TIME_ENCODE(&now
, ctime
);
1423 ZFS_TIME_DECODE(&(ZTOI(zp
)->i_ctime
), ctime
);
1424 if (ZTOZSB(zp
)->z_use_fuids
)
1425 zp
->z_pflags
|= ZFS_ARCHIVE
;
1430 * Grow the block size for a file.
1432 * IN: zp - znode of file to free data in.
1433 * size - requested block size
1434 * tx - open transaction.
1436 * NOTE: this function assumes that the znode is write locked.
1439 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1444 if (size
<= zp
->z_blksz
)
1447 * If the file size is already greater than the current blocksize,
1448 * we will not grow. If there is more than one block in a file,
1449 * the blocksize cannot change.
1451 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1454 error
= dmu_object_set_blocksize(ZTOZSB(zp
)->z_os
, zp
->z_id
,
1457 if (error
== ENOTSUP
)
1461 /* What blocksize did we actually get? */
1462 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1466 * Increase the file length
1468 * IN: zp - znode of file to free data in.
1469 * end - new end-of-file
1471 * RETURN: 0 on success, error code on failure
1474 zfs_extend(znode_t
*zp
, uint64_t end
)
1476 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1483 * We will change zp_size, lock the whole file.
1485 rl
= zfs_range_lock(&zp
->z_range_lock
, 0, UINT64_MAX
, RL_WRITER
);
1488 * Nothing to do if file already at desired length.
1490 if (end
<= zp
->z_size
) {
1491 zfs_range_unlock(rl
);
1494 tx
= dmu_tx_create(zfsvfs
->z_os
);
1495 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1496 zfs_sa_upgrade_txholds(tx
, zp
);
1497 if (end
> zp
->z_blksz
&&
1498 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zfsvfs
->z_max_blksz
)) {
1500 * We are growing the file past the current block size.
1502 if (zp
->z_blksz
> ZTOZSB(zp
)->z_max_blksz
) {
1504 * File's blocksize is already larger than the
1505 * "recordsize" property. Only let it grow to
1506 * the next power of 2.
1508 ASSERT(!ISP2(zp
->z_blksz
));
1509 newblksz
= MIN(end
, 1 << highbit64(zp
->z_blksz
));
1511 newblksz
= MIN(end
, ZTOZSB(zp
)->z_max_blksz
);
1513 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1518 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1521 zfs_range_unlock(rl
);
1526 zfs_grow_blocksize(zp
, newblksz
, tx
);
1530 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(ZTOZSB(zp
)),
1531 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1533 zfs_range_unlock(rl
);
1541 * zfs_zero_partial_page - Modeled after update_pages() but
1542 * with different arguments and semantics for use by zfs_freesp().
1544 * Zeroes a piece of a single page cache entry for zp at offset
1545 * start and length len.
1547 * Caller must acquire a range lock on the file for the region
1548 * being zeroed in order that the ARC and page cache stay in sync.
1551 zfs_zero_partial_page(znode_t
*zp
, uint64_t start
, uint64_t len
)
1553 struct address_space
*mp
= ZTOI(zp
)->i_mapping
;
1558 ASSERT((start
& PAGE_MASK
) == ((start
+ len
- 1) & PAGE_MASK
));
1560 off
= start
& (PAGE_SIZE
- 1);
1563 pp
= find_lock_page(mp
, start
>> PAGE_SHIFT
);
1565 if (mapping_writably_mapped(mp
))
1566 flush_dcache_page(pp
);
1569 bzero(pb
+ off
, len
);
1572 if (mapping_writably_mapped(mp
))
1573 flush_dcache_page(pp
);
1575 mark_page_accessed(pp
);
1576 SetPageUptodate(pp
);
1584 * Free space in a file.
1586 * IN: zp - znode of file to free data in.
1587 * off - start of section to free.
1588 * len - length of section to free.
1590 * RETURN: 0 on success, error code on failure
1593 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1595 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1600 * Lock the range being freed.
1602 rl
= zfs_range_lock(&zp
->z_range_lock
, off
, len
, RL_WRITER
);
1605 * Nothing to do if file already at desired length.
1607 if (off
>= zp
->z_size
) {
1608 zfs_range_unlock(rl
);
1612 if (off
+ len
> zp
->z_size
)
1613 len
= zp
->z_size
- off
;
1615 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, off
, len
);
1618 * Zero partial page cache entries. This must be done under a
1619 * range lock in order to keep the ARC and page cache in sync.
1621 if (zp
->z_is_mapped
) {
1622 loff_t first_page
, last_page
, page_len
;
1623 loff_t first_page_offset
, last_page_offset
;
1625 /* first possible full page in hole */
1626 first_page
= (off
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1627 /* last page of hole */
1628 last_page
= (off
+ len
) >> PAGE_SHIFT
;
1630 /* offset of first_page */
1631 first_page_offset
= first_page
<< PAGE_SHIFT
;
1632 /* offset of last_page */
1633 last_page_offset
= last_page
<< PAGE_SHIFT
;
1635 /* truncate whole pages */
1636 if (last_page_offset
> first_page_offset
) {
1637 truncate_inode_pages_range(ZTOI(zp
)->i_mapping
,
1638 first_page_offset
, last_page_offset
- 1);
1641 /* truncate sub-page ranges */
1642 if (first_page
> last_page
) {
1643 /* entire punched area within a single page */
1644 zfs_zero_partial_page(zp
, off
, len
);
1646 /* beginning of punched area at the end of a page */
1647 page_len
= first_page_offset
- off
;
1649 zfs_zero_partial_page(zp
, off
, page_len
);
1651 /* end of punched area at the beginning of a page */
1652 page_len
= off
+ len
- last_page_offset
;
1654 zfs_zero_partial_page(zp
, last_page_offset
,
1658 zfs_range_unlock(rl
);
1666 * IN: zp - znode of file to free data in.
1667 * end - new end-of-file.
1669 * RETURN: 0 on success, error code on failure
1672 zfs_trunc(znode_t
*zp
, uint64_t end
)
1674 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1678 sa_bulk_attr_t bulk
[2];
1682 * We will change zp_size, lock the whole file.
1684 rl
= zfs_range_lock(&zp
->z_range_lock
, 0, UINT64_MAX
, RL_WRITER
);
1687 * Nothing to do if file already at desired length.
1689 if (end
>= zp
->z_size
) {
1690 zfs_range_unlock(rl
);
1694 error
= dmu_free_long_range(zfsvfs
->z_os
, zp
->z_id
, end
,
1697 zfs_range_unlock(rl
);
1700 tx
= dmu_tx_create(zfsvfs
->z_os
);
1701 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1702 zfs_sa_upgrade_txholds(tx
, zp
);
1703 dmu_tx_mark_netfree(tx
);
1704 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1707 zfs_range_unlock(rl
);
1712 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zfsvfs
),
1713 NULL
, &zp
->z_size
, sizeof (zp
->z_size
));
1716 zp
->z_pflags
&= ~ZFS_SPARSE
;
1717 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1718 NULL
, &zp
->z_pflags
, 8);
1720 VERIFY(sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
) == 0);
1724 zfs_range_unlock(rl
);
1730 * Free space in a file
1732 * IN: zp - znode of file to free data in.
1733 * off - start of range
1734 * len - end of range (0 => EOF)
1735 * flag - current file open mode flags.
1736 * log - TRUE if this action should be logged
1738 * RETURN: 0 on success, error code on failure
1741 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1744 zfsvfs_t
*zfsvfs
= ZTOZSB(zp
);
1745 zilog_t
*zilog
= zfsvfs
->z_log
;
1747 uint64_t mtime
[2], ctime
[2];
1748 sa_bulk_attr_t bulk
[3];
1752 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zfsvfs
), &mode
,
1753 sizeof (mode
))) != 0)
1756 if (off
> zp
->z_size
) {
1757 error
= zfs_extend(zp
, off
+len
);
1758 if (error
== 0 && log
)
1764 error
= zfs_trunc(zp
, off
);
1766 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1767 off
+ len
> zp
->z_size
)
1768 error
= zfs_extend(zp
, off
+len
);
1773 tx
= dmu_tx_create(zfsvfs
->z_os
);
1774 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1775 zfs_sa_upgrade_txholds(tx
, zp
);
1776 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1782 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zfsvfs
), NULL
, mtime
, 16);
1783 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zfsvfs
), NULL
, ctime
, 16);
1784 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zfsvfs
),
1785 NULL
, &zp
->z_pflags
, 8);
1786 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
);
1787 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1790 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1794 zfs_inode_update(zp
);
1799 * Truncate the page cache - for file truncate operations, use
1800 * the purpose-built API for truncations. For punching operations,
1801 * the truncation is handled under a range lock in zfs_free_range.
1804 truncate_setsize(ZTOI(zp
), off
);
1809 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1811 struct super_block
*sb
;
1813 uint64_t moid
, obj
, sa_obj
, version
;
1814 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1820 znode_t
*rootzp
= NULL
;
1823 zfs_acl_ids_t acl_ids
;
1826 * First attempt to create master node.
1829 * In an empty objset, there are no blocks to read and thus
1830 * there can be no i/o errors (which we assert below).
1832 moid
= MASTER_NODE_OBJ
;
1833 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1834 DMU_OT_NONE
, 0, tx
);
1838 * Set starting attributes.
1840 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1842 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1843 /* For the moment we expect all zpl props to be uint64_ts */
1847 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1848 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1849 name
= nvpair_name(elem
);
1850 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1854 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1857 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1859 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1862 ASSERT(version
!= 0);
1863 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1866 * Create zap object used for SA attribute registration
1869 if (version
>= ZPL_VERSION_SA
) {
1870 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1871 DMU_OT_NONE
, 0, tx
);
1872 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1878 * Create a delete queue.
1880 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1882 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1886 * Create root znode. Create minimal znode/inode/zfsvfs/sb
1887 * to allow zfs_mknode to work.
1889 vattr
.va_mask
= ATTR_MODE
|ATTR_UID
|ATTR_GID
;
1890 vattr
.va_mode
= S_IFDIR
|0755;
1891 vattr
.va_uid
= crgetuid(cr
);
1892 vattr
.va_gid
= crgetgid(cr
);
1894 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1895 rootzp
->z_moved
= 0;
1896 rootzp
->z_unlinked
= 0;
1897 rootzp
->z_atime_dirty
= 0;
1898 rootzp
->z_is_sa
= USE_SA(version
, os
);
1899 rootzp
->z_pflags
= 0;
1901 zfsvfs
= kmem_zalloc(sizeof (zfsvfs_t
), KM_SLEEP
);
1903 zfsvfs
->z_parent
= zfsvfs
;
1904 zfsvfs
->z_version
= version
;
1905 zfsvfs
->z_use_fuids
= USE_FUIDS(version
, os
);
1906 zfsvfs
->z_use_sa
= USE_SA(version
, os
);
1907 zfsvfs
->z_norm
= norm
;
1909 sb
= kmem_zalloc(sizeof (struct super_block
), KM_SLEEP
);
1910 sb
->s_fs_info
= zfsvfs
;
1912 ZTOI(rootzp
)->i_sb
= sb
;
1914 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
1915 &zfsvfs
->z_attr_table
);
1920 * Fold case on file systems that are always or sometimes case
1923 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1924 zfsvfs
->z_norm
|= U8_TEXTPREP_TOUPPER
;
1926 mutex_init(&zfsvfs
->z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1927 list_create(&zfsvfs
->z_all_znodes
, sizeof (znode_t
),
1928 offsetof(znode_t
, z_link_node
));
1930 size
= MIN(1 << (highbit64(zfs_object_mutex_size
)-1), ZFS_OBJ_MTX_MAX
);
1931 zfsvfs
->z_hold_size
= size
;
1932 zfsvfs
->z_hold_trees
= vmem_zalloc(sizeof (avl_tree_t
) * size
,
1934 zfsvfs
->z_hold_locks
= vmem_zalloc(sizeof (kmutex_t
) * size
, KM_SLEEP
);
1935 for (i
= 0; i
!= size
; i
++) {
1936 avl_create(&zfsvfs
->z_hold_trees
[i
], zfs_znode_hold_compare
,
1937 sizeof (znode_hold_t
), offsetof(znode_hold_t
, zh_node
));
1938 mutex_init(&zfsvfs
->z_hold_locks
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1941 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1942 cr
, NULL
, &acl_ids
));
1943 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1944 ASSERT3P(zp
, ==, rootzp
);
1945 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1947 zfs_acl_ids_free(&acl_ids
);
1949 atomic_set(&ZTOI(rootzp
)->i_count
, 0);
1950 sa_handle_destroy(rootzp
->z_sa_hdl
);
1951 kmem_cache_free(znode_cache
, rootzp
);
1954 * Create shares directory
1956 error
= zfs_create_share_dir(zfsvfs
, tx
);
1959 for (i
= 0; i
!= size
; i
++) {
1960 avl_destroy(&zfsvfs
->z_hold_trees
[i
]);
1961 mutex_destroy(&zfsvfs
->z_hold_locks
[i
]);
1964 mutex_destroy(&zfsvfs
->z_znodes_lock
);
1966 vmem_free(zfsvfs
->z_hold_trees
, sizeof (avl_tree_t
) * size
);
1967 vmem_free(zfsvfs
->z_hold_locks
, sizeof (kmutex_t
) * size
);
1968 kmem_free(sb
, sizeof (struct super_block
));
1969 kmem_free(zfsvfs
, sizeof (zfsvfs_t
));
1971 #endif /* _KERNEL */
1974 zfs_sa_setup(objset_t
*osp
, sa_attr_type_t
**sa_table
)
1976 uint64_t sa_obj
= 0;
1979 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1980 if (error
!= 0 && error
!= ENOENT
)
1983 error
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
, sa_table
);
1988 zfs_grab_sa_handle(objset_t
*osp
, uint64_t obj
, sa_handle_t
**hdlp
,
1989 dmu_buf_t
**db
, void *tag
)
1991 dmu_object_info_t doi
;
1994 if ((error
= sa_buf_hold(osp
, obj
, tag
, db
)) != 0)
1997 dmu_object_info_from_db(*db
, &doi
);
1998 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1999 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
2000 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
2001 doi
.doi_bonus_size
< sizeof (znode_phys_t
))) {
2002 sa_buf_rele(*db
, tag
);
2003 return (SET_ERROR(ENOTSUP
));
2006 error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
, hdlp
);
2008 sa_buf_rele(*db
, tag
);
2016 zfs_release_sa_handle(sa_handle_t
*hdl
, dmu_buf_t
*db
, void *tag
)
2018 sa_handle_destroy(hdl
);
2019 sa_buf_rele(db
, tag
);
2023 * Given an object number, return its parent object number and whether
2024 * or not the object is an extended attribute directory.
2027 zfs_obj_to_pobj(objset_t
*osp
, sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
2028 uint64_t *pobjp
, int *is_xattrdir
)
2033 uint64_t parent_mode
;
2034 sa_bulk_attr_t bulk
[3];
2035 sa_handle_t
*sa_hdl
;
2040 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
], NULL
,
2041 &parent
, sizeof (parent
));
2042 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
2043 &pflags
, sizeof (pflags
));
2044 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
2045 &mode
, sizeof (mode
));
2047 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0)
2051 * When a link is removed its parent pointer is not changed and will
2052 * be invalid. There are two cases where a link is removed but the
2053 * file stays around, when it goes to the delete queue and when there
2054 * are additional links.
2056 error
= zfs_grab_sa_handle(osp
, parent
, &sa_hdl
, &sa_db
, FTAG
);
2060 error
= sa_lookup(sa_hdl
, ZPL_MODE
, &parent_mode
, sizeof (parent_mode
));
2061 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
2065 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
2068 * Extended attributes can be applied to files, directories, etc.
2069 * Otherwise the parent must be a directory.
2071 if (!*is_xattrdir
&& !S_ISDIR(parent_mode
))
2072 return (SET_ERROR(EINVAL
));
2080 * Given an object number, return some zpl level statistics
2083 zfs_obj_to_stats_impl(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
2086 sa_bulk_attr_t bulk
[4];
2089 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
2090 &sb
->zs_mode
, sizeof (sb
->zs_mode
));
2091 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_GEN
], NULL
,
2092 &sb
->zs_gen
, sizeof (sb
->zs_gen
));
2093 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_LINKS
], NULL
,
2094 &sb
->zs_links
, sizeof (sb
->zs_links
));
2095 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_CTIME
], NULL
,
2096 &sb
->zs_ctime
, sizeof (sb
->zs_ctime
));
2098 return (sa_bulk_lookup(hdl
, bulk
, count
));
2102 zfs_obj_to_path_impl(objset_t
*osp
, uint64_t obj
, sa_handle_t
*hdl
,
2103 sa_attr_type_t
*sa_table
, char *buf
, int len
)
2105 sa_handle_t
*sa_hdl
;
2106 sa_handle_t
*prevhdl
= NULL
;
2107 dmu_buf_t
*prevdb
= NULL
;
2108 dmu_buf_t
*sa_db
= NULL
;
2109 char *path
= buf
+ len
- 1;
2115 uint64_t deleteq_obj
;
2116 VERIFY0(zap_lookup(osp
, MASTER_NODE_OBJ
,
2117 ZFS_UNLINKED_SET
, sizeof (uint64_t), 1, &deleteq_obj
));
2118 error
= zap_lookup_int(osp
, deleteq_obj
, obj
);
2121 } else if (error
!= ENOENT
) {
2128 char component
[MAXNAMELEN
+ 2];
2130 int is_xattrdir
= 0;
2133 zfs_release_sa_handle(prevhdl
, prevdb
, FTAG
);
2135 if ((error
= zfs_obj_to_pobj(osp
, sa_hdl
, sa_table
, &pobj
,
2136 &is_xattrdir
)) != 0)
2147 (void) sprintf(component
+ 1, "<xattrdir>");
2149 error
= zap_value_search(osp
, pobj
, obj
,
2150 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
2155 complen
= strlen(component
);
2157 ASSERT(path
>= buf
);
2158 bcopy(component
, path
, complen
);
2161 if (sa_hdl
!= hdl
) {
2165 error
= zfs_grab_sa_handle(osp
, obj
, &sa_hdl
, &sa_db
, FTAG
);
2173 if (sa_hdl
!= NULL
&& sa_hdl
!= hdl
) {
2174 ASSERT(sa_db
!= NULL
);
2175 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
2179 (void) memmove(buf
, path
, buf
+ len
- path
);
2185 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
2187 sa_attr_type_t
*sa_table
;
2192 error
= zfs_sa_setup(osp
, &sa_table
);
2196 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
2200 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2202 zfs_release_sa_handle(hdl
, db
, FTAG
);
2207 zfs_obj_to_stats(objset_t
*osp
, uint64_t obj
, zfs_stat_t
*sb
,
2210 char *path
= buf
+ len
- 1;
2211 sa_attr_type_t
*sa_table
;
2218 error
= zfs_sa_setup(osp
, &sa_table
);
2222 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
2226 error
= zfs_obj_to_stats_impl(hdl
, sa_table
, sb
);
2228 zfs_release_sa_handle(hdl
, db
, FTAG
);
2232 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
2234 zfs_release_sa_handle(hdl
, db
, FTAG
);
2238 #if defined(_KERNEL)
2239 EXPORT_SYMBOL(zfs_create_fs
);
2240 EXPORT_SYMBOL(zfs_obj_to_path
);
2243 module_param(zfs_object_mutex_size
, uint
, 0644);
2244 MODULE_PARM_DESC(zfs_object_mutex_size
, "Size of znode hold array");