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/zfs_vnops.h>
55 #include <sys/zfs_ctldir.h>
56 #include <sys/dnode.h>
57 #include <sys/fs/zfs.h>
58 #include <sys/kidmap.h>
63 #include <sys/refcount.h>
66 #include <sys/zfs_znode.h>
68 #include <sys/zfs_sa.h>
69 #include <sys/zfs_stat.h>
72 #include "zfs_comutil.h"
75 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
76 * turned on when DEBUG is also defined.
83 #define ZNODE_STAT_ADD(stat) ((stat)++)
85 #define ZNODE_STAT_ADD(stat) /* nothing */
86 #endif /* ZNODE_STATS */
89 * Functions needed for userland (ie: libzpool) are not put under
90 * #ifdef_KERNEL; the rest of the functions have dependencies
91 * (such as VFS logic) that will not compile easily in userland.
95 static kmem_cache_t
*znode_cache
= NULL
;
99 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
103 inode_init_once(ZTOI(zp
));
104 list_link_init(&zp
->z_link_node
);
106 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
107 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
108 rw_init(&zp
->z_name_lock
, NULL
, RW_DEFAULT
, NULL
);
109 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
110 rw_init(&zp
->z_xattr_lock
, NULL
, RW_DEFAULT
, NULL
);
112 mutex_init(&zp
->z_range_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
113 avl_create(&zp
->z_range_avl
, zfs_range_compare
,
114 sizeof (rl_t
), offsetof(rl_t
, r_node
));
116 zp
->z_dirlocks
= NULL
;
117 zp
->z_acl_cached
= NULL
;
118 zp
->z_xattr_cached
= NULL
;
125 zfs_znode_cache_destructor(void *buf
, void *arg
)
129 ASSERT(!list_link_active(&zp
->z_link_node
));
130 mutex_destroy(&zp
->z_lock
);
131 rw_destroy(&zp
->z_parent_lock
);
132 rw_destroy(&zp
->z_name_lock
);
133 mutex_destroy(&zp
->z_acl_lock
);
134 rw_destroy(&zp
->z_xattr_lock
);
135 avl_destroy(&zp
->z_range_avl
);
136 mutex_destroy(&zp
->z_range_lock
);
138 ASSERT(zp
->z_dirlocks
== NULL
);
139 ASSERT(zp
->z_acl_cached
== NULL
);
140 ASSERT(zp
->z_xattr_cached
== NULL
);
149 ASSERT(znode_cache
== NULL
);
150 znode_cache
= kmem_cache_create("zfs_znode_cache",
151 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
152 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, KMC_KMEM
);
162 kmem_cache_destroy(znode_cache
);
167 zfs_create_share_dir(zfs_sb_t
*zsb
, dmu_tx_t
*tx
)
169 #ifdef HAVE_SMB_SHARE
170 zfs_acl_ids_t acl_ids
;
177 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
178 vattr
.va_mode
= S_IFDIR
| 0555;
179 vattr
.va_uid
= crgetuid(kcred
);
180 vattr
.va_gid
= crgetgid(kcred
);
182 sharezp
= kmem_cache_alloc(znode_cache
, KM_PUSHPAGE
);
183 sharezp
->z_moved
= 0;
184 sharezp
->z_unlinked
= 0;
185 sharezp
->z_atime_dirty
= 0;
186 sharezp
->z_zfsvfs
= zfsvfs
;
187 sharezp
->z_is_sa
= zfsvfs
->z_use_sa
;
193 VERIFY(0 == zfs_acl_ids_create(sharezp
, IS_ROOT_NODE
, &vattr
,
194 kcred
, NULL
, &acl_ids
));
195 zfs_mknode(sharezp
, &vattr
, tx
, kcred
, IS_ROOT_NODE
, &zp
, &acl_ids
);
196 ASSERT3P(zp
, ==, sharezp
);
197 ASSERT(!vn_in_dnlc(ZTOV(sharezp
))); /* not valid to move */
198 POINTER_INVALIDATE(&sharezp
->z_zfsvfs
);
199 error
= zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
200 ZFS_SHARES_DIR
, 8, 1, &sharezp
->z_id
, tx
);
201 zfsvfs
->z_shares_dir
= sharezp
->z_id
;
203 zfs_acl_ids_free(&acl_ids
);
204 // ZTOV(sharezp)->v_count = 0;
205 sa_handle_destroy(sharezp
->z_sa_hdl
);
206 kmem_cache_free(znode_cache
, sharezp
);
211 #endif /* HAVE_SMB_SHARE */
215 zfs_znode_sa_init(zfs_sb_t
*zsb
, znode_t
*zp
,
216 dmu_buf_t
*db
, dmu_object_type_t obj_type
, sa_handle_t
*sa_hdl
)
218 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb
, zp
->z_id
)));
220 mutex_enter(&zp
->z_lock
);
222 ASSERT(zp
->z_sa_hdl
== NULL
);
223 ASSERT(zp
->z_acl_cached
== NULL
);
224 if (sa_hdl
== NULL
) {
225 VERIFY(0 == sa_handle_get_from_db(zsb
->z_os
, db
, zp
,
226 SA_HDL_SHARED
, &zp
->z_sa_hdl
));
228 zp
->z_sa_hdl
= sa_hdl
;
229 sa_set_userp(sa_hdl
, zp
);
232 zp
->z_is_sa
= (obj_type
== DMU_OT_SA
) ? B_TRUE
: B_FALSE
;
234 mutex_exit(&zp
->z_lock
);
238 zfs_znode_dmu_fini(znode_t
*zp
)
240 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp
), zp
->z_id
)) ||
242 RW_WRITE_HELD(&ZTOZSB(zp
)->z_teardown_inactive_lock
));
244 sa_handle_destroy(zp
->z_sa_hdl
);
249 * Called by new_inode() to allocate a new inode.
252 zfs_inode_alloc(struct super_block
*sb
, struct inode
**ip
)
256 zp
= kmem_cache_alloc(znode_cache
, KM_PUSHPAGE
);
263 * Called in multiple places when an inode should be destroyed.
266 zfs_inode_destroy(struct inode
*ip
)
268 znode_t
*zp
= ITOZ(ip
);
269 zfs_sb_t
*zsb
= ZTOZSB(zp
);
271 if (zfsctl_is_node(ip
))
272 zfsctl_inode_destroy(ip
);
274 mutex_enter(&zsb
->z_znodes_lock
);
275 list_remove(&zsb
->z_all_znodes
, zp
);
277 mutex_exit(&zsb
->z_znodes_lock
);
279 if (zp
->z_acl_cached
) {
280 zfs_acl_free(zp
->z_acl_cached
);
281 zp
->z_acl_cached
= NULL
;
284 if (zp
->z_xattr_cached
) {
285 nvlist_free(zp
->z_xattr_cached
);
286 zp
->z_xattr_cached
= NULL
;
289 kmem_cache_free(znode_cache
, zp
);
293 zfs_inode_set_ops(zfs_sb_t
*zsb
, struct inode
*ip
)
297 switch (ip
->i_mode
& S_IFMT
) {
299 ip
->i_op
= &zpl_inode_operations
;
300 ip
->i_fop
= &zpl_file_operations
;
301 ip
->i_mapping
->a_ops
= &zpl_address_space_operations
;
305 ip
->i_op
= &zpl_dir_inode_operations
;
306 ip
->i_fop
= &zpl_dir_file_operations
;
307 ITOZ(ip
)->z_zn_prefetch
= B_TRUE
;
311 ip
->i_op
= &zpl_symlink_inode_operations
;
315 * rdev is only stored in a SA only for device files.
319 VERIFY(sa_lookup(ITOZ(ip
)->z_sa_hdl
, SA_ZPL_RDEV(zsb
),
320 &rdev
, sizeof (rdev
)) == 0);
324 init_special_inode(ip
, ip
->i_mode
, rdev
);
325 ip
->i_op
= &zpl_special_inode_operations
;
329 printk("ZFS: Invalid mode: 0x%x\n", ip
->i_mode
);
335 * Construct a znode+inode and initialize.
337 * This does not do a call to dmu_set_user() that is
338 * up to the caller to do, in case you don't want to
342 zfs_znode_alloc(zfs_sb_t
*zsb
, dmu_buf_t
*db
, int blksz
,
343 dmu_object_type_t obj_type
, uint64_t obj
, sa_handle_t
*hdl
,
344 struct dentry
*dentry
, struct inode
*dip
)
349 sa_bulk_attr_t bulk
[9];
354 ip
= new_inode(zsb
->z_sb
);
359 ASSERT(zp
->z_dirlocks
== NULL
);
360 ASSERT3P(zp
->z_acl_cached
, ==, NULL
);
361 ASSERT3P(zp
->z_xattr_cached
, ==, NULL
);
365 zp
->z_atime_dirty
= 0;
367 zp
->z_id
= db
->db_object
;
369 zp
->z_seq
= 0x7A4653;
371 zp
->z_is_zvol
= B_FALSE
;
372 zp
->z_is_mapped
= B_FALSE
;
373 zp
->z_is_ctldir
= B_FALSE
;
375 zfs_znode_sa_init(zsb
, zp
, db
, obj_type
, hdl
);
377 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
, &zp
->z_mode
, 8);
378 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zsb
), NULL
, &zp
->z_gen
, 8);
379 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
, &zp
->z_size
, 8);
380 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
, &zp
->z_links
, 8);
381 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
383 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zsb
), NULL
,
385 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zsb
), NULL
,
387 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zsb
), NULL
, &zp
->z_uid
, 8);
388 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zsb
), NULL
, &zp
->z_gid
, 8);
390 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
) != 0 || zp
->z_gen
== 0) {
392 sa_handle_destroy(zp
->z_sa_hdl
);
398 zfs_inode_update(zp
);
399 zfs_inode_set_ops(zsb
, ip
);
401 if (insert_inode_locked(ip
))
405 if (zpl_xattr_security_init(ip
, dip
, &dentry
->d_name
))
408 d_instantiate(dentry
, ip
);
411 mutex_enter(&zsb
->z_znodes_lock
);
412 list_insert_tail(&zsb
->z_all_znodes
, zp
);
415 mutex_exit(&zsb
->z_znodes_lock
);
417 unlock_new_inode(ip
);
421 unlock_new_inode(ip
);
427 * Update the embedded inode given the znode. We should work toward
428 * eliminating this function as soon as possible by removing values
429 * which are duplicated between the znode and inode. If the generic
430 * inode has the correct field it should be used, and the ZFS code
431 * updated to access the inode. This can be done incrementally.
434 zfs_inode_update(znode_t
*zp
)
439 uint64_t atime
[2], mtime
[2], ctime
[2];
445 /* Skip .zfs control nodes which do not exist on disk. */
446 if (zfsctl_is_node(ip
))
449 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ATIME(zsb
), &atime
, 16);
450 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MTIME(zsb
), &mtime
, 16);
451 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_CTIME(zsb
), &ctime
, 16);
453 spin_lock(&ip
->i_lock
);
454 ip
->i_generation
= zp
->z_gen
;
455 ip
->i_uid
= zp
->z_uid
;
456 ip
->i_gid
= zp
->z_gid
;
457 set_nlink(ip
, zp
->z_links
);
458 ip
->i_mode
= zp
->z_mode
;
459 ip
->i_blkbits
= SPA_MINBLOCKSHIFT
;
460 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &blksize
,
461 (u_longlong_t
*)&ip
->i_blocks
);
463 ZFS_TIME_DECODE(&ip
->i_atime
, atime
);
464 ZFS_TIME_DECODE(&ip
->i_mtime
, mtime
);
465 ZFS_TIME_DECODE(&ip
->i_ctime
, ctime
);
467 i_size_write(ip
, zp
->z_size
);
468 spin_unlock(&ip
->i_lock
);
471 static uint64_t empty_xattr
;
472 static uint64_t pad
[4];
473 static zfs_acl_phys_t acl_phys
;
475 * Create a new DMU object to hold a zfs znode.
477 * IN: dzp - parent directory for new znode
478 * vap - file attributes for new znode
479 * tx - dmu transaction id for zap operations
480 * cr - credentials of caller
482 * IS_ROOT_NODE - new object will be root
483 * IS_XATTR - new object is an attribute
484 * bonuslen - length of bonus buffer
485 * setaclp - File/Dir initial ACL
486 * fuidp - Tracks fuid allocation.
488 * OUT: zpp - allocated znode
492 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
493 uint_t flag
, znode_t
**zpp
, zfs_acl_ids_t
*acl_ids
)
495 uint64_t crtime
[2], atime
[2], mtime
[2], ctime
[2];
496 uint64_t mode
, size
, links
, parent
, pflags
;
497 uint64_t dzp_pflags
= 0;
499 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
506 dmu_object_type_t obj_type
;
507 sa_bulk_attr_t
*sa_attrs
;
509 zfs_acl_locator_cb_t locate
= { 0 };
512 obj
= vap
->va_nodeid
;
513 now
= vap
->va_ctime
; /* see zfs_replay_create() */
514 gen
= vap
->va_nblocks
; /* ditto */
518 gen
= dmu_tx_get_txg(tx
);
521 obj_type
= zsb
->z_use_sa
? DMU_OT_SA
: DMU_OT_ZNODE
;
522 bonuslen
= (obj_type
== DMU_OT_SA
) ?
523 DN_MAX_BONUSLEN
: ZFS_OLD_ZNODE_PHYS_SIZE
;
526 * Create a new DMU object.
529 * There's currently no mechanism for pre-reading the blocks that will
530 * be needed to allocate a new object, so we accept the small chance
531 * that there will be an i/o error and we will fail one of the
534 if (S_ISDIR(vap
->va_mode
)) {
536 err
= zap_create_claim_norm(zsb
->z_os
, obj
,
537 zsb
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
538 obj_type
, bonuslen
, tx
);
539 ASSERT3U(err
, ==, 0);
541 obj
= zap_create_norm(zsb
->z_os
,
542 zsb
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
543 obj_type
, bonuslen
, tx
);
547 err
= dmu_object_claim(zsb
->z_os
, obj
,
548 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
549 obj_type
, bonuslen
, tx
);
550 ASSERT3U(err
, ==, 0);
552 obj
= dmu_object_alloc(zsb
->z_os
,
553 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
554 obj_type
, bonuslen
, tx
);
558 ZFS_OBJ_HOLD_ENTER(zsb
, obj
);
559 VERIFY(0 == sa_buf_hold(zsb
->z_os
, obj
, NULL
, &db
));
562 * If this is the root, fix up the half-initialized parent pointer
563 * to reference the just-allocated physical data area.
565 if (flag
& IS_ROOT_NODE
) {
568 dzp_pflags
= dzp
->z_pflags
;
572 * If parent is an xattr, so am I.
574 if (dzp_pflags
& ZFS_XATTR
) {
578 if (zsb
->z_use_fuids
)
579 pflags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
583 if (S_ISDIR(vap
->va_mode
)) {
584 size
= 2; /* contents ("." and "..") */
585 links
= (flag
& (IS_ROOT_NODE
| IS_XATTR
)) ? 2 : 1;
590 if (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))
594 mode
= acl_ids
->z_mode
;
599 * No execs denied will be deterimed when zfs_mode_compute() is called.
601 pflags
|= acl_ids
->z_aclp
->z_hints
&
602 (ZFS_ACL_TRIVIAL
|ZFS_INHERIT_ACE
|ZFS_ACL_AUTO_INHERIT
|
603 ZFS_ACL_DEFAULTED
|ZFS_ACL_PROTECTED
);
605 ZFS_TIME_ENCODE(&now
, crtime
);
606 ZFS_TIME_ENCODE(&now
, ctime
);
608 if (vap
->va_mask
& ATTR_ATIME
) {
609 ZFS_TIME_ENCODE(&vap
->va_atime
, atime
);
611 ZFS_TIME_ENCODE(&now
, atime
);
614 if (vap
->va_mask
& ATTR_MTIME
) {
615 ZFS_TIME_ENCODE(&vap
->va_mtime
, mtime
);
617 ZFS_TIME_ENCODE(&now
, mtime
);
620 /* Now add in all of the "SA" attributes */
621 VERIFY(0 == sa_handle_get_from_db(zsb
->z_os
, db
, NULL
, SA_HDL_SHARED
,
625 * Setup the array of attributes to be replaced/set on the new file
627 * order for DMU_OT_ZNODE is critical since it needs to be constructed
628 * in the old znode_phys_t format. Don't change this ordering
630 sa_attrs
= kmem_alloc(sizeof(sa_bulk_attr_t
) * ZPL_END
, KM_SLEEP
);
632 if (obj_type
== DMU_OT_ZNODE
) {
633 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zsb
),
635 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zsb
),
637 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zsb
),
639 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zsb
),
641 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zsb
),
643 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zsb
),
645 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zsb
),
647 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zsb
),
650 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zsb
),
652 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zsb
),
654 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zsb
),
656 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zsb
),
657 NULL
, &acl_ids
->z_fuid
, 8);
658 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zsb
),
659 NULL
, &acl_ids
->z_fgid
, 8);
660 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zsb
),
662 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zsb
),
664 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zsb
),
666 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zsb
),
668 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zsb
),
670 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zsb
),
674 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zsb
), NULL
, &links
, 8);
676 if (obj_type
== DMU_OT_ZNODE
) {
677 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zsb
), NULL
,
680 if (obj_type
== DMU_OT_ZNODE
||
681 (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))) {
682 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zsb
),
685 if (obj_type
== DMU_OT_ZNODE
) {
686 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zsb
),
688 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zsb
), NULL
,
689 &acl_ids
->z_fuid
, 8);
690 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zsb
), NULL
,
691 &acl_ids
->z_fgid
, 8);
692 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zsb
), NULL
, pad
,
693 sizeof (uint64_t) * 4);
694 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zsb
), NULL
,
695 &acl_phys
, sizeof (zfs_acl_phys_t
));
696 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
697 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zsb
), NULL
,
698 &acl_ids
->z_aclp
->z_acl_count
, 8);
699 locate
.cb_aclp
= acl_ids
->z_aclp
;
700 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zsb
),
701 zfs_acl_data_locator
, &locate
,
702 acl_ids
->z_aclp
->z_acl_bytes
);
703 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
704 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
707 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
709 if (!(flag
& IS_ROOT_NODE
)) {
710 *zpp
= zfs_znode_alloc(zsb
, db
, 0, obj_type
, obj
, sa_hdl
,
711 vap
->va_dentry
, ZTOI(dzp
));
712 ASSERT(*zpp
!= NULL
);
716 * If we are creating the root node, the "parent" we
717 * passed in is the znode for the root.
721 (*zpp
)->z_sa_hdl
= sa_hdl
;
724 (*zpp
)->z_pflags
= pflags
;
725 (*zpp
)->z_mode
= mode
;
727 if (obj_type
== DMU_OT_ZNODE
||
728 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
729 err
= zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
);
730 ASSERT3S(err
, ==, 0);
732 kmem_free(sa_attrs
, sizeof(sa_bulk_attr_t
) * ZPL_END
);
733 ZFS_OBJ_HOLD_EXIT(zsb
, obj
);
737 * zfs_xvattr_set only updates the in-core attributes
738 * it is assumed the caller will be doing an sa_bulk_update
739 * to push the changes out
742 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
746 xoap
= xva_getxoptattr(xvap
);
749 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
751 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
752 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(ZTOZSB(zp
)),
753 ×
, sizeof (times
), tx
);
754 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
756 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
757 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
759 XVA_SET_RTN(xvap
, XAT_READONLY
);
761 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
762 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
764 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
766 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
767 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
769 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
771 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
772 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
774 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
776 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
777 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
779 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
781 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
782 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
784 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
786 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
787 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
789 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
791 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
792 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
794 XVA_SET_RTN(xvap
, XAT_NODUMP
);
796 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
797 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
799 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
801 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
802 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
803 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
804 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
806 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
807 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
809 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
811 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
812 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
813 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
815 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
816 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
818 XVA_SET_RTN(xvap
, XAT_REPARSE
);
820 if (XVA_ISSET_REQ(xvap
, XAT_OFFLINE
)) {
821 ZFS_ATTR_SET(zp
, ZFS_OFFLINE
, xoap
->xoa_offline
,
823 XVA_SET_RTN(xvap
, XAT_OFFLINE
);
825 if (XVA_ISSET_REQ(xvap
, XAT_SPARSE
)) {
826 ZFS_ATTR_SET(zp
, ZFS_SPARSE
, xoap
->xoa_sparse
,
828 XVA_SET_RTN(xvap
, XAT_SPARSE
);
833 zfs_zget(zfs_sb_t
*zsb
, uint64_t obj_num
, znode_t
**zpp
)
835 dmu_object_info_t doi
;
845 ip
= ilookup(zsb
->z_sb
, obj_num
);
847 ZFS_OBJ_HOLD_ENTER(zsb
, obj_num
);
849 err
= sa_buf_hold(zsb
->z_os
, obj_num
, NULL
, &db
);
851 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
856 dmu_object_info_from_db(db
, &doi
);
857 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
858 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
859 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
860 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
861 sa_buf_rele(db
, NULL
);
862 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
867 hdl
= dmu_buf_get_user(db
);
871 * ilookup returned NULL, which means
872 * the znode is dying - but the SA handle isn't
873 * quite dead yet, we need to drop any locks
874 * we're holding, re-schedule the task and try again.
876 sa_buf_rele(db
, NULL
);
877 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
883 zp
= sa_get_userdata(hdl
);
886 * Since "SA" does immediate eviction we
887 * should never find a sa handle that doesn't
888 * know about the znode.
891 ASSERT3P(zp
, !=, NULL
);
893 mutex_enter(&zp
->z_lock
);
894 ASSERT3U(zp
->z_id
, ==, obj_num
);
895 if (zp
->z_unlinked
) {
902 sa_buf_rele(db
, NULL
);
903 mutex_exit(&zp
->z_lock
);
904 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
909 ASSERT3P(ip
, ==, NULL
);
912 * Not found create new znode/vnode but only if file exists.
914 * There is a small window where zfs_vget() could
915 * find this object while a file create is still in
916 * progress. This is checked for in zfs_znode_alloc()
918 * if zfs_znode_alloc() fails it will drop the hold on the
921 zp
= zfs_znode_alloc(zsb
, db
, doi
.doi_data_block_size
,
922 doi
.doi_bonus_type
, obj_num
, NULL
, NULL
, NULL
);
928 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
933 zfs_rezget(znode_t
*zp
)
935 zfs_sb_t
*zsb
= ZTOZSB(zp
);
936 dmu_object_info_t doi
;
938 uint64_t obj_num
= zp
->z_id
;
940 sa_bulk_attr_t bulk
[8];
945 ZFS_OBJ_HOLD_ENTER(zsb
, obj_num
);
947 mutex_enter(&zp
->z_acl_lock
);
948 if (zp
->z_acl_cached
) {
949 zfs_acl_free(zp
->z_acl_cached
);
950 zp
->z_acl_cached
= NULL
;
953 mutex_exit(&zp
->z_acl_lock
);
954 ASSERT(zp
->z_sa_hdl
== NULL
);
955 err
= sa_buf_hold(zsb
->z_os
, obj_num
, NULL
, &db
);
957 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
961 dmu_object_info_from_db(db
, &doi
);
962 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
963 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
964 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
965 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
966 sa_buf_rele(db
, NULL
);
967 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
971 zfs_znode_sa_init(zsb
, zp
, db
, doi
.doi_bonus_type
, NULL
);
973 /* reload cached values */
974 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zsb
), NULL
,
976 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
,
977 &zp
->z_size
, sizeof (zp
->z_size
));
978 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
979 &zp
->z_links
, sizeof (zp
->z_links
));
980 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
981 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
982 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zsb
), NULL
,
983 &zp
->z_atime
, sizeof (zp
->z_atime
));
984 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zsb
), NULL
,
985 &zp
->z_uid
, sizeof (zp
->z_uid
));
986 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zsb
), NULL
,
987 &zp
->z_gid
, sizeof (zp
->z_gid
));
988 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
,
989 &mode
, sizeof (mode
));
991 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
992 zfs_znode_dmu_fini(zp
);
993 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
999 if (gen
!= zp
->z_gen
) {
1000 zfs_znode_dmu_fini(zp
);
1001 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1005 zp
->z_unlinked
= (zp
->z_links
== 0);
1006 zp
->z_blksz
= doi
.doi_data_block_size
;
1008 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1014 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1016 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1017 objset_t
*os
= zsb
->z_os
;
1018 uint64_t obj
= zp
->z_id
;
1019 uint64_t acl_obj
= zfs_external_acl(zp
);
1021 ZFS_OBJ_HOLD_ENTER(zsb
, obj
);
1023 VERIFY(!zp
->z_is_sa
);
1024 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1026 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1027 zfs_znode_dmu_fini(zp
);
1028 ZFS_OBJ_HOLD_EXIT(zsb
, obj
);
1032 zfs_zinactive(znode_t
*zp
)
1034 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1035 uint64_t z_id
= zp
->z_id
;
1036 boolean_t drop_mutex
= 0;
1038 ASSERT(zp
->z_sa_hdl
);
1041 * Don't allow a zfs_zget() while were trying to release this znode.
1043 * Linux allows direct memory reclaim which means that any KM_SLEEP
1044 * allocation may trigger inode eviction. This can lead to a deadlock
1045 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
1046 * zfs_zinactive() call path. To avoid this deadlock the process
1047 * must not reacquire the mutex when it is already holding it.
1049 if (!ZFS_OBJ_HOLD_OWNED(zsb
, z_id
)) {
1050 ZFS_OBJ_HOLD_ENTER(zsb
, z_id
);
1054 mutex_enter(&zp
->z_lock
);
1057 * If this was the last reference to a file with no links,
1058 * remove the file from the file system.
1060 if (zp
->z_unlinked
) {
1061 mutex_exit(&zp
->z_lock
);
1064 ZFS_OBJ_HOLD_EXIT(zsb
, z_id
);
1070 mutex_exit(&zp
->z_lock
);
1071 zfs_znode_dmu_fini(zp
);
1074 ZFS_OBJ_HOLD_EXIT(zsb
, z_id
);
1078 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1079 uint64_t ctime
[2], boolean_t have_tx
)
1085 if (have_tx
) { /* will sa_bulk_update happen really soon? */
1086 zp
->z_atime_dirty
= 0;
1089 zp
->z_atime_dirty
= 1;
1092 if (flag
& ATTR_ATIME
) {
1093 ZFS_TIME_ENCODE(&now
, zp
->z_atime
);
1096 if (flag
& ATTR_MTIME
) {
1097 ZFS_TIME_ENCODE(&now
, mtime
);
1098 if (ZTOZSB(zp
)->z_use_fuids
) {
1099 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1104 if (flag
& ATTR_CTIME
) {
1105 ZFS_TIME_ENCODE(&now
, ctime
);
1106 if (ZTOZSB(zp
)->z_use_fuids
)
1107 zp
->z_pflags
|= ZFS_ARCHIVE
;
1112 * Grow the block size for a file.
1114 * IN: zp - znode of file to free data in.
1115 * size - requested block size
1116 * tx - open transaction.
1118 * NOTE: this function assumes that the znode is write locked.
1121 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1126 if (size
<= zp
->z_blksz
)
1129 * If the file size is already greater than the current blocksize,
1130 * we will not grow. If there is more than one block in a file,
1131 * the blocksize cannot change.
1133 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1136 error
= dmu_object_set_blocksize(ZTOZSB(zp
)->z_os
, zp
->z_id
,
1139 if (error
== ENOTSUP
)
1141 ASSERT3U(error
, ==, 0);
1143 /* What blocksize did we actually get? */
1144 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1148 * Increase the file length
1150 * IN: zp - znode of file to free data in.
1151 * end - new end-of-file
1153 * RETURN: 0 if success
1154 * error code if failure
1157 zfs_extend(znode_t
*zp
, uint64_t end
)
1159 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1166 * We will change zp_size, lock the whole file.
1168 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1171 * Nothing to do if file already at desired length.
1173 if (end
<= zp
->z_size
) {
1174 zfs_range_unlock(rl
);
1178 tx
= dmu_tx_create(zsb
->z_os
);
1179 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1180 zfs_sa_upgrade_txholds(tx
, zp
);
1181 if (end
> zp
->z_blksz
&&
1182 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zsb
->z_max_blksz
)) {
1184 * We are growing the file past the current block size.
1186 if (zp
->z_blksz
> ZTOZSB(zp
)->z_max_blksz
) {
1187 ASSERT(!ISP2(zp
->z_blksz
));
1188 newblksz
= MIN(end
, SPA_MAXBLOCKSIZE
);
1190 newblksz
= MIN(end
, ZTOZSB(zp
)->z_max_blksz
);
1192 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1197 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1199 if (error
== ERESTART
) {
1205 zfs_range_unlock(rl
);
1210 zfs_grow_blocksize(zp
, newblksz
, tx
);
1214 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(ZTOZSB(zp
)),
1215 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1217 zfs_range_unlock(rl
);
1225 * Free space in a file.
1227 * IN: zp - znode of file to free data in.
1228 * off - start of section to free.
1229 * len - length of section to free.
1231 * RETURN: 0 if success
1232 * error code if failure
1235 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1237 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1242 * Lock the range being freed.
1244 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1247 * Nothing to do if file already at desired length.
1249 if (off
>= zp
->z_size
) {
1250 zfs_range_unlock(rl
);
1254 if (off
+ len
> zp
->z_size
)
1255 len
= zp
->z_size
- off
;
1257 error
= dmu_free_long_range(zsb
->z_os
, zp
->z_id
, off
, len
);
1259 zfs_range_unlock(rl
);
1267 * IN: zp - znode of file to free data in.
1268 * end - new end-of-file.
1270 * RETURN: 0 if success
1271 * error code if failure
1274 zfs_trunc(znode_t
*zp
, uint64_t end
)
1276 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1280 sa_bulk_attr_t bulk
[2];
1284 * We will change zp_size, lock the whole file.
1286 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1289 * Nothing to do if file already at desired length.
1291 if (end
>= zp
->z_size
) {
1292 zfs_range_unlock(rl
);
1296 error
= dmu_free_long_range(zsb
->z_os
, zp
->z_id
, end
, -1);
1298 zfs_range_unlock(rl
);
1302 tx
= dmu_tx_create(zsb
->z_os
);
1303 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1304 zfs_sa_upgrade_txholds(tx
, zp
);
1305 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1307 if (error
== ERESTART
) {
1313 zfs_range_unlock(rl
);
1318 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
),
1319 NULL
, &zp
->z_size
, sizeof (zp
->z_size
));
1322 zp
->z_pflags
&= ~ZFS_SPARSE
;
1323 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
1324 NULL
, &zp
->z_pflags
, 8);
1326 VERIFY(sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
) == 0);
1330 zfs_range_unlock(rl
);
1336 * Free space in a file
1338 * IN: zp - znode of file to free data in.
1339 * off - start of range
1340 * len - end of range (0 => EOF)
1341 * flag - current file open mode flags.
1342 * log - TRUE if this action should be logged
1344 * RETURN: 0 if success
1345 * error code if failure
1348 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1350 struct inode
*ip
= ZTOI(zp
);
1352 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1353 zilog_t
*zilog
= zsb
->z_log
;
1355 uint64_t mtime
[2], ctime
[2];
1356 sa_bulk_attr_t bulk
[3];
1360 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zsb
), &mode
,
1361 sizeof (mode
))) != 0)
1364 if (off
> zp
->z_size
) {
1365 error
= zfs_extend(zp
, off
+len
);
1366 if (error
== 0 && log
)
1373 * Check for any locks in the region to be freed.
1375 if (ip
->i_flock
&& mandatory_lock(ip
)) {
1376 uint64_t length
= (len
? len
: zp
->z_size
- off
);
1377 if (!lock_may_write(ip
, off
, length
))
1382 error
= zfs_trunc(zp
, off
);
1384 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1385 off
+ len
> zp
->z_size
)
1386 error
= zfs_extend(zp
, off
+len
);
1391 tx
= dmu_tx_create(zsb
->z_os
);
1392 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1393 zfs_sa_upgrade_txholds(tx
, zp
);
1394 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1396 if (error
== ERESTART
) {
1405 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
), NULL
, mtime
, 16);
1406 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
, ctime
, 16);
1407 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
1408 NULL
, &zp
->z_pflags
, 8);
1409 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
1410 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1413 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1416 zfs_inode_update(zp
);
1421 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1423 struct super_block
*sb
;
1425 uint64_t moid
, obj
, sa_obj
, version
;
1426 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1431 znode_t
*rootzp
= NULL
;
1434 zfs_acl_ids_t acl_ids
;
1437 * First attempt to create master node.
1440 * In an empty objset, there are no blocks to read and thus
1441 * there can be no i/o errors (which we assert below).
1443 moid
= MASTER_NODE_OBJ
;
1444 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1445 DMU_OT_NONE
, 0, tx
);
1449 * Set starting attributes.
1451 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1453 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1454 /* For the moment we expect all zpl props to be uint64_ts */
1458 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1459 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1460 name
= nvpair_name(elem
);
1461 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1465 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1468 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1470 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1473 ASSERT(version
!= 0);
1474 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1477 * Create zap object used for SA attribute registration
1480 if (version
>= ZPL_VERSION_SA
) {
1481 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1482 DMU_OT_NONE
, 0, tx
);
1483 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1489 * Create a delete queue.
1491 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1493 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1497 * Create root znode. Create minimal znode/inode/zsb/sb
1498 * to allow zfs_mknode to work.
1500 vattr
.va_mask
= ATTR_MODE
|ATTR_UID
|ATTR_GID
;
1501 vattr
.va_mode
= S_IFDIR
|0755;
1502 vattr
.va_uid
= crgetuid(cr
);
1503 vattr
.va_gid
= crgetgid(cr
);
1505 rootzp
= kmem_cache_alloc(znode_cache
, KM_SLEEP
);
1506 rootzp
->z_moved
= 0;
1507 rootzp
->z_unlinked
= 0;
1508 rootzp
->z_atime_dirty
= 0;
1509 rootzp
->z_is_sa
= USE_SA(version
, os
);
1511 zsb
= kmem_zalloc(sizeof (zfs_sb_t
), KM_SLEEP
);
1513 zsb
->z_parent
= zsb
;
1514 zsb
->z_version
= version
;
1515 zsb
->z_use_fuids
= USE_FUIDS(version
, os
);
1516 zsb
->z_use_sa
= USE_SA(version
, os
);
1519 sb
= kmem_zalloc(sizeof (struct super_block
), KM_SLEEP
);
1520 sb
->s_fs_info
= zsb
;
1522 ZTOI(rootzp
)->i_sb
= sb
;
1524 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
1525 &zsb
->z_attr_table
);
1530 * Fold case on file systems that are always or sometimes case
1533 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1534 zsb
->z_norm
|= U8_TEXTPREP_TOUPPER
;
1536 mutex_init(&zsb
->z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1537 list_create(&zsb
->z_all_znodes
, sizeof (znode_t
),
1538 offsetof(znode_t
, z_link_node
));
1540 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1541 mutex_init(&zsb
->z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1543 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1544 cr
, NULL
, &acl_ids
));
1545 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1546 ASSERT3P(zp
, ==, rootzp
);
1547 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1549 zfs_acl_ids_free(&acl_ids
);
1551 atomic_set(&ZTOI(rootzp
)->i_count
, 0);
1552 sa_handle_destroy(rootzp
->z_sa_hdl
);
1553 kmem_cache_free(znode_cache
, rootzp
);
1556 * Create shares directory
1558 error
= zfs_create_share_dir(zsb
, tx
);
1561 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1562 mutex_destroy(&zsb
->z_hold_mtx
[i
]);
1564 kmem_free(sb
, sizeof (struct super_block
));
1565 kmem_free(zsb
, sizeof (zfs_sb_t
));
1567 #endif /* _KERNEL */
1570 zfs_sa_setup(objset_t
*osp
, sa_attr_type_t
**sa_table
)
1572 uint64_t sa_obj
= 0;
1575 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1576 if (error
!= 0 && error
!= ENOENT
)
1579 error
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
, sa_table
);
1584 zfs_grab_sa_handle(objset_t
*osp
, uint64_t obj
, sa_handle_t
**hdlp
,
1585 dmu_buf_t
**db
, void *tag
)
1587 dmu_object_info_t doi
;
1590 if ((error
= sa_buf_hold(osp
, obj
, tag
, db
)) != 0)
1593 dmu_object_info_from_db(*db
, &doi
);
1594 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1595 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
1596 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1597 doi
.doi_bonus_size
< sizeof (znode_phys_t
))) {
1598 sa_buf_rele(*db
, tag
);
1602 error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
, hdlp
);
1604 sa_buf_rele(*db
, tag
);
1612 zfs_release_sa_handle(sa_handle_t
*hdl
, dmu_buf_t
*db
, void *tag
)
1614 sa_handle_destroy(hdl
);
1615 sa_buf_rele(db
, tag
);
1619 * Given an object number, return its parent object number and whether
1620 * or not the object is an extended attribute directory.
1623 zfs_obj_to_pobj(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
, uint64_t *pobjp
,
1629 sa_bulk_attr_t bulk
[3];
1633 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
], NULL
,
1634 &parent
, sizeof (parent
));
1635 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
1636 &pflags
, sizeof (pflags
));
1637 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1638 &mode
, sizeof (mode
));
1640 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0)
1644 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
1650 * Given an object number, return some zpl level statistics
1653 zfs_obj_to_stats_impl(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
1656 sa_bulk_attr_t bulk
[4];
1659 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1660 &sb
->zs_mode
, sizeof (sb
->zs_mode
));
1661 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_GEN
], NULL
,
1662 &sb
->zs_gen
, sizeof (sb
->zs_gen
));
1663 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_LINKS
], NULL
,
1664 &sb
->zs_links
, sizeof (sb
->zs_links
));
1665 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_CTIME
], NULL
,
1666 &sb
->zs_ctime
, sizeof (sb
->zs_ctime
));
1668 return (sa_bulk_lookup(hdl
, bulk
, count
));
1672 zfs_obj_to_path_impl(objset_t
*osp
, uint64_t obj
, sa_handle_t
*hdl
,
1673 sa_attr_type_t
*sa_table
, char *buf
, int len
)
1675 sa_handle_t
*sa_hdl
;
1676 sa_handle_t
*prevhdl
= NULL
;
1677 dmu_buf_t
*prevdb
= NULL
;
1678 dmu_buf_t
*sa_db
= NULL
;
1679 char *path
= buf
+ len
- 1;
1687 char component
[MAXNAMELEN
+ 2];
1692 zfs_release_sa_handle(prevhdl
, prevdb
, FTAG
);
1694 if ((error
= zfs_obj_to_pobj(sa_hdl
, sa_table
, &pobj
,
1695 &is_xattrdir
)) != 0)
1706 (void) sprintf(component
+ 1, "<xattrdir>");
1708 error
= zap_value_search(osp
, pobj
, obj
,
1709 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
1714 complen
= strlen(component
);
1716 ASSERT(path
>= buf
);
1717 bcopy(component
, path
, complen
);
1720 if (sa_hdl
!= hdl
) {
1724 error
= zfs_grab_sa_handle(osp
, obj
, &sa_hdl
, &sa_db
, FTAG
);
1732 if (sa_hdl
!= NULL
&& sa_hdl
!= hdl
) {
1733 ASSERT(sa_db
!= NULL
);
1734 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
1738 (void) memmove(buf
, path
, buf
+ len
- path
);
1744 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
1746 sa_attr_type_t
*sa_table
;
1751 error
= zfs_sa_setup(osp
, &sa_table
);
1755 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
1759 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
1761 zfs_release_sa_handle(hdl
, db
, FTAG
);
1766 zfs_obj_to_stats(objset_t
*osp
, uint64_t obj
, zfs_stat_t
*sb
,
1769 char *path
= buf
+ len
- 1;
1770 sa_attr_type_t
*sa_table
;
1777 error
= zfs_sa_setup(osp
, &sa_table
);
1781 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
1785 error
= zfs_obj_to_stats_impl(hdl
, sa_table
, sb
);
1787 zfs_release_sa_handle(hdl
, db
, FTAG
);
1791 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
1793 zfs_release_sa_handle(hdl
, db
, FTAG
);
1797 #if defined(_KERNEL) && defined(HAVE_SPL)
1798 EXPORT_SYMBOL(zfs_create_fs
);
1799 EXPORT_SYMBOL(zfs_obj_to_path
);