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) 2013 by Delphix. All rights reserved.
26 /* Portions Copyright 2007 Jeremy Teo */
29 #include <sys/types.h>
30 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/sysmacros.h>
34 #include <sys/resource.h>
35 #include <sys/mntent.h>
36 #include <sys/mkdev.h>
37 #include <sys/u8_textprep.h>
38 #include <sys/dsl_dataset.h>
40 #include <sys/vfs_opreg.h>
41 #include <sys/vnode.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
47 #include <sys/atomic.h>
49 #include "fs/fs_subr.h"
50 #include <sys/zfs_dir.h>
51 #include <sys/zfs_acl.h>
52 #include <sys/zfs_ioctl.h>
53 #include <sys/zfs_rlock.h>
54 #include <sys/zfs_fuid.h>
55 #include <sys/zfs_vnops.h>
56 #include <sys/zfs_ctldir.h>
57 #include <sys/dnode.h>
58 #include <sys/fs/zfs.h>
59 #include <sys/kidmap.h>
64 #include <sys/refcount.h>
67 #include <sys/zfs_znode.h>
69 #include <sys/zfs_sa.h>
70 #include <sys/zfs_stat.h>
73 #include "zfs_comutil.h"
76 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
77 * turned on when DEBUG is also defined.
84 #define ZNODE_STAT_ADD(stat) ((stat)++)
86 #define ZNODE_STAT_ADD(stat) /* nothing */
87 #endif /* ZNODE_STATS */
90 * Functions needed for userland (ie: libzpool) are not put under
91 * #ifdef_KERNEL; the rest of the functions have dependencies
92 * (such as VFS logic) that will not compile easily in userland.
96 static kmem_cache_t
*znode_cache
= NULL
;
100 zfs_znode_cache_constructor(void *buf
, void *arg
, int kmflags
)
104 inode_init_once(ZTOI(zp
));
105 list_link_init(&zp
->z_link_node
);
107 mutex_init(&zp
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
108 rw_init(&zp
->z_parent_lock
, NULL
, RW_DEFAULT
, NULL
);
109 rw_init(&zp
->z_name_lock
, NULL
, RW_DEFAULT
, NULL
);
110 mutex_init(&zp
->z_acl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
111 rw_init(&zp
->z_xattr_lock
, NULL
, RW_DEFAULT
, NULL
);
113 mutex_init(&zp
->z_range_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
114 avl_create(&zp
->z_range_avl
, zfs_range_compare
,
115 sizeof (rl_t
), offsetof(rl_t
, r_node
));
117 zp
->z_dirlocks
= NULL
;
118 zp
->z_acl_cached
= NULL
;
119 zp
->z_xattr_cached
= NULL
;
120 zp
->z_xattr_parent
= NULL
;
127 zfs_znode_cache_destructor(void *buf
, void *arg
)
131 ASSERT(!list_link_active(&zp
->z_link_node
));
132 mutex_destroy(&zp
->z_lock
);
133 rw_destroy(&zp
->z_parent_lock
);
134 rw_destroy(&zp
->z_name_lock
);
135 mutex_destroy(&zp
->z_acl_lock
);
136 rw_destroy(&zp
->z_xattr_lock
);
137 avl_destroy(&zp
->z_range_avl
);
138 mutex_destroy(&zp
->z_range_lock
);
140 ASSERT(zp
->z_dirlocks
== NULL
);
141 ASSERT(zp
->z_acl_cached
== NULL
);
142 ASSERT(zp
->z_xattr_cached
== NULL
);
143 ASSERT(zp
->z_xattr_parent
== NULL
);
152 ASSERT(znode_cache
== NULL
);
153 znode_cache
= kmem_cache_create("zfs_znode_cache",
154 sizeof (znode_t
), 0, zfs_znode_cache_constructor
,
155 zfs_znode_cache_destructor
, NULL
, NULL
, NULL
, KMC_KMEM
);
165 kmem_cache_destroy(znode_cache
);
170 zfs_create_share_dir(zfs_sb_t
*zsb
, dmu_tx_t
*tx
)
172 #ifdef HAVE_SMB_SHARE
173 zfs_acl_ids_t acl_ids
;
180 vattr
.va_mask
= AT_MODE
|AT_UID
|AT_GID
|AT_TYPE
;
181 vattr
.va_mode
= S_IFDIR
| 0555;
182 vattr
.va_uid
= crgetuid(kcred
);
183 vattr
.va_gid
= crgetgid(kcred
);
185 sharezp
= kmem_cache_alloc(znode_cache
, KM_PUSHPAGE
);
186 sharezp
->z_moved
= 0;
187 sharezp
->z_unlinked
= 0;
188 sharezp
->z_atime_dirty
= 0;
189 sharezp
->z_zfsvfs
= zfsvfs
;
190 sharezp
->z_is_sa
= zfsvfs
->z_use_sa
;
196 VERIFY(0 == zfs_acl_ids_create(sharezp
, IS_ROOT_NODE
, &vattr
,
197 kcred
, NULL
, &acl_ids
));
198 zfs_mknode(sharezp
, &vattr
, tx
, kcred
, IS_ROOT_NODE
, &zp
, &acl_ids
);
199 ASSERT3P(zp
, ==, sharezp
);
200 ASSERT(!vn_in_dnlc(ZTOV(sharezp
))); /* not valid to move */
201 POINTER_INVALIDATE(&sharezp
->z_zfsvfs
);
202 error
= zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
203 ZFS_SHARES_DIR
, 8, 1, &sharezp
->z_id
, tx
);
204 zfsvfs
->z_shares_dir
= sharezp
->z_id
;
206 zfs_acl_ids_free(&acl_ids
);
207 // ZTOV(sharezp)->v_count = 0;
208 sa_handle_destroy(sharezp
->z_sa_hdl
);
209 kmem_cache_free(znode_cache
, sharezp
);
214 #endif /* HAVE_SMB_SHARE */
218 zfs_znode_sa_init(zfs_sb_t
*zsb
, znode_t
*zp
,
219 dmu_buf_t
*db
, dmu_object_type_t obj_type
, sa_handle_t
*sa_hdl
)
221 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb
, zp
->z_id
)));
223 mutex_enter(&zp
->z_lock
);
225 ASSERT(zp
->z_sa_hdl
== NULL
);
226 ASSERT(zp
->z_acl_cached
== NULL
);
227 if (sa_hdl
== NULL
) {
228 VERIFY(0 == sa_handle_get_from_db(zsb
->z_os
, db
, zp
,
229 SA_HDL_SHARED
, &zp
->z_sa_hdl
));
231 zp
->z_sa_hdl
= sa_hdl
;
232 sa_set_userp(sa_hdl
, zp
);
235 zp
->z_is_sa
= (obj_type
== DMU_OT_SA
) ? B_TRUE
: B_FALSE
;
237 mutex_exit(&zp
->z_lock
);
241 zfs_znode_dmu_fini(znode_t
*zp
)
243 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp
), zp
->z_id
)) ||
245 RW_WRITE_HELD(&ZTOZSB(zp
)->z_teardown_inactive_lock
));
247 sa_handle_destroy(zp
->z_sa_hdl
);
252 * Called by new_inode() to allocate a new inode.
255 zfs_inode_alloc(struct super_block
*sb
, struct inode
**ip
)
259 zp
= kmem_cache_alloc(znode_cache
, KM_PUSHPAGE
);
266 * Called in multiple places when an inode should be destroyed.
269 zfs_inode_destroy(struct inode
*ip
)
271 znode_t
*zp
= ITOZ(ip
);
272 zfs_sb_t
*zsb
= ZTOZSB(zp
);
274 if (zfsctl_is_node(ip
))
275 zfsctl_inode_destroy(ip
);
277 mutex_enter(&zsb
->z_znodes_lock
);
278 if (list_link_active(&zp
->z_link_node
)) {
279 list_remove(&zsb
->z_all_znodes
, zp
);
282 mutex_exit(&zsb
->z_znodes_lock
);
284 if (zp
->z_acl_cached
) {
285 zfs_acl_free(zp
->z_acl_cached
);
286 zp
->z_acl_cached
= NULL
;
289 if (zp
->z_xattr_cached
) {
290 nvlist_free(zp
->z_xattr_cached
);
291 zp
->z_xattr_cached
= NULL
;
294 if (zp
->z_xattr_parent
) {
295 iput(ZTOI(zp
->z_xattr_parent
));
296 zp
->z_xattr_parent
= NULL
;
299 kmem_cache_free(znode_cache
, zp
);
303 zfs_inode_set_ops(zfs_sb_t
*zsb
, struct inode
*ip
)
307 switch (ip
->i_mode
& S_IFMT
) {
309 ip
->i_op
= &zpl_inode_operations
;
310 ip
->i_fop
= &zpl_file_operations
;
311 ip
->i_mapping
->a_ops
= &zpl_address_space_operations
;
315 ip
->i_op
= &zpl_dir_inode_operations
;
316 ip
->i_fop
= &zpl_dir_file_operations
;
317 ITOZ(ip
)->z_zn_prefetch
= B_TRUE
;
321 ip
->i_op
= &zpl_symlink_inode_operations
;
325 * rdev is only stored in a SA only for device files.
329 VERIFY(sa_lookup(ITOZ(ip
)->z_sa_hdl
, SA_ZPL_RDEV(zsb
),
330 &rdev
, sizeof (rdev
)) == 0);
334 init_special_inode(ip
, ip
->i_mode
, rdev
);
335 ip
->i_op
= &zpl_special_inode_operations
;
339 printk("ZFS: Invalid mode: 0x%x\n", ip
->i_mode
);
345 * Construct a znode+inode and initialize.
347 * This does not do a call to dmu_set_user() that is
348 * up to the caller to do, in case you don't want to
352 zfs_znode_alloc(zfs_sb_t
*zsb
, dmu_buf_t
*db
, int blksz
,
353 dmu_object_type_t obj_type
, uint64_t obj
, sa_handle_t
*hdl
,
360 sa_bulk_attr_t bulk
[9];
365 ip
= new_inode(zsb
->z_sb
);
370 ASSERT(zp
->z_dirlocks
== NULL
);
371 ASSERT3P(zp
->z_acl_cached
, ==, NULL
);
372 ASSERT3P(zp
->z_xattr_cached
, ==, NULL
);
373 ASSERT3P(zp
->z_xattr_parent
, ==, NULL
);
377 zp
->z_atime_dirty
= 0;
379 zp
->z_id
= db
->db_object
;
381 zp
->z_seq
= 0x7A4653;
383 zp
->z_is_zvol
= B_FALSE
;
384 zp
->z_is_mapped
= B_FALSE
;
385 zp
->z_is_ctldir
= B_FALSE
;
386 zp
->z_is_stale
= B_FALSE
;
388 zfs_znode_sa_init(zsb
, zp
, db
, obj_type
, hdl
);
390 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
, &mode
, 8);
391 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zsb
), NULL
, &zp
->z_gen
, 8);
392 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
, &zp
->z_size
, 8);
393 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
, &zp
->z_links
, 8);
394 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
396 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_PARENT(zsb
), NULL
,
398 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zsb
), NULL
,
400 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zsb
), NULL
, &zp
->z_uid
, 8);
401 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zsb
), NULL
, &zp
->z_gid
, 8);
403 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
) != 0 || zp
->z_gen
== 0) {
405 sa_handle_destroy(zp
->z_sa_hdl
);
413 * xattr znodes hold a reference on their unique parent
415 if (dip
&& zp
->z_pflags
& ZFS_XATTR
) {
417 zp
->z_xattr_parent
= ITOZ(dip
);
421 zfs_inode_update(zp
);
422 zfs_inode_set_ops(zsb
, ip
);
425 * The only way insert_inode_locked() can fail is if the ip->i_ino
426 * number is already hashed for this super block. This can never
427 * happen because the inode numbers map 1:1 with the object numbers.
429 * The one exception is rolling back a mounted file system, but in
430 * this case all the active inode are unhashed during the rollback.
432 VERIFY3S(insert_inode_locked(ip
), ==, 0);
434 mutex_enter(&zsb
->z_znodes_lock
);
435 list_insert_tail(&zsb
->z_all_znodes
, zp
);
438 mutex_exit(&zsb
->z_znodes_lock
);
440 unlock_new_inode(ip
);
444 unlock_new_inode(ip
);
450 * Update the embedded inode given the znode. We should work toward
451 * eliminating this function as soon as possible by removing values
452 * which are duplicated between the znode and inode. If the generic
453 * inode has the correct field it should be used, and the ZFS code
454 * updated to access the inode. This can be done incrementally.
457 zfs_inode_update(znode_t
*zp
)
462 uint64_t atime
[2], mtime
[2], ctime
[2];
468 /* Skip .zfs control nodes which do not exist on disk. */
469 if (zfsctl_is_node(ip
))
472 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ATIME(zsb
), &atime
, 16);
473 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MTIME(zsb
), &mtime
, 16);
474 sa_lookup(zp
->z_sa_hdl
, SA_ZPL_CTIME(zsb
), &ctime
, 16);
476 spin_lock(&ip
->i_lock
);
477 ip
->i_generation
= zp
->z_gen
;
478 ip
->i_uid
= SUID_TO_KUID(zp
->z_uid
);
479 ip
->i_gid
= SGID_TO_KGID(zp
->z_gid
);
480 set_nlink(ip
, zp
->z_links
);
481 ip
->i_mode
= zp
->z_mode
;
482 ip
->i_blkbits
= SPA_MINBLOCKSHIFT
;
483 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &blksize
,
484 (u_longlong_t
*)&ip
->i_blocks
);
486 ZFS_TIME_DECODE(&ip
->i_atime
, atime
);
487 ZFS_TIME_DECODE(&ip
->i_mtime
, mtime
);
488 ZFS_TIME_DECODE(&ip
->i_ctime
, ctime
);
490 i_size_write(ip
, zp
->z_size
);
491 spin_unlock(&ip
->i_lock
);
494 static uint64_t empty_xattr
;
495 static uint64_t pad
[4];
496 static zfs_acl_phys_t acl_phys
;
498 * Create a new DMU object to hold a zfs znode.
500 * IN: dzp - parent directory for new znode
501 * vap - file attributes for new znode
502 * tx - dmu transaction id for zap operations
503 * cr - credentials of caller
505 * IS_ROOT_NODE - new object will be root
506 * IS_XATTR - new object is an attribute
507 * bonuslen - length of bonus buffer
508 * setaclp - File/Dir initial ACL
509 * fuidp - Tracks fuid allocation.
511 * OUT: zpp - allocated znode
515 zfs_mknode(znode_t
*dzp
, vattr_t
*vap
, dmu_tx_t
*tx
, cred_t
*cr
,
516 uint_t flag
, znode_t
**zpp
, zfs_acl_ids_t
*acl_ids
)
518 uint64_t crtime
[2], atime
[2], mtime
[2], ctime
[2];
519 uint64_t mode
, size
, links
, parent
, pflags
;
520 uint64_t dzp_pflags
= 0;
522 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
529 dmu_object_type_t obj_type
;
530 sa_bulk_attr_t
*sa_attrs
;
532 zfs_acl_locator_cb_t locate
= { 0 };
535 obj
= vap
->va_nodeid
;
536 now
= vap
->va_ctime
; /* see zfs_replay_create() */
537 gen
= vap
->va_nblocks
; /* ditto */
541 gen
= dmu_tx_get_txg(tx
);
544 obj_type
= zsb
->z_use_sa
? DMU_OT_SA
: DMU_OT_ZNODE
;
545 bonuslen
= (obj_type
== DMU_OT_SA
) ?
546 DN_MAX_BONUSLEN
: ZFS_OLD_ZNODE_PHYS_SIZE
;
549 * Create a new DMU object.
552 * There's currently no mechanism for pre-reading the blocks that will
553 * be needed to allocate a new object, so we accept the small chance
554 * that there will be an i/o error and we will fail one of the
557 if (S_ISDIR(vap
->va_mode
)) {
559 err
= zap_create_claim_norm(zsb
->z_os
, obj
,
560 zsb
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
561 obj_type
, bonuslen
, tx
);
564 obj
= zap_create_norm(zsb
->z_os
,
565 zsb
->z_norm
, DMU_OT_DIRECTORY_CONTENTS
,
566 obj_type
, bonuslen
, tx
);
570 err
= dmu_object_claim(zsb
->z_os
, obj
,
571 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
572 obj_type
, bonuslen
, tx
);
575 obj
= dmu_object_alloc(zsb
->z_os
,
576 DMU_OT_PLAIN_FILE_CONTENTS
, 0,
577 obj_type
, bonuslen
, tx
);
581 ZFS_OBJ_HOLD_ENTER(zsb
, obj
);
582 VERIFY(0 == sa_buf_hold(zsb
->z_os
, obj
, NULL
, &db
));
585 * If this is the root, fix up the half-initialized parent pointer
586 * to reference the just-allocated physical data area.
588 if (flag
& IS_ROOT_NODE
) {
591 dzp_pflags
= dzp
->z_pflags
;
595 * If parent is an xattr, so am I.
597 if (dzp_pflags
& ZFS_XATTR
) {
601 if (zsb
->z_use_fuids
)
602 pflags
= ZFS_ARCHIVE
| ZFS_AV_MODIFIED
;
606 if (S_ISDIR(vap
->va_mode
)) {
607 size
= 2; /* contents ("." and "..") */
608 links
= (flag
& (IS_ROOT_NODE
| IS_XATTR
)) ? 2 : 1;
613 if (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))
617 mode
= acl_ids
->z_mode
;
622 * No execs denied will be deterimed when zfs_mode_compute() is called.
624 pflags
|= acl_ids
->z_aclp
->z_hints
&
625 (ZFS_ACL_TRIVIAL
|ZFS_INHERIT_ACE
|ZFS_ACL_AUTO_INHERIT
|
626 ZFS_ACL_DEFAULTED
|ZFS_ACL_PROTECTED
);
628 ZFS_TIME_ENCODE(&now
, crtime
);
629 ZFS_TIME_ENCODE(&now
, ctime
);
631 if (vap
->va_mask
& ATTR_ATIME
) {
632 ZFS_TIME_ENCODE(&vap
->va_atime
, atime
);
634 ZFS_TIME_ENCODE(&now
, atime
);
637 if (vap
->va_mask
& ATTR_MTIME
) {
638 ZFS_TIME_ENCODE(&vap
->va_mtime
, mtime
);
640 ZFS_TIME_ENCODE(&now
, mtime
);
643 /* Now add in all of the "SA" attributes */
644 VERIFY(0 == sa_handle_get_from_db(zsb
->z_os
, db
, NULL
, SA_HDL_SHARED
,
648 * Setup the array of attributes to be replaced/set on the new file
650 * order for DMU_OT_ZNODE is critical since it needs to be constructed
651 * in the old znode_phys_t format. Don't change this ordering
653 sa_attrs
= kmem_alloc(sizeof (sa_bulk_attr_t
) * ZPL_END
, KM_PUSHPAGE
);
655 if (obj_type
== DMU_OT_ZNODE
) {
656 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zsb
),
658 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zsb
),
660 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zsb
),
662 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zsb
),
664 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zsb
),
666 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zsb
),
668 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zsb
),
670 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zsb
),
673 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MODE(zsb
),
675 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_SIZE(zsb
),
677 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GEN(zsb
),
679 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zsb
),
680 NULL
, &acl_ids
->z_fuid
, 8);
681 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zsb
),
682 NULL
, &acl_ids
->z_fgid
, 8);
683 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PARENT(zsb
),
685 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zsb
),
687 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ATIME(zsb
),
689 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_MTIME(zsb
),
691 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CTIME(zsb
),
693 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_CRTIME(zsb
),
697 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_LINKS(zsb
), NULL
, &links
, 8);
699 if (obj_type
== DMU_OT_ZNODE
) {
700 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_XATTR(zsb
), NULL
,
703 if (obj_type
== DMU_OT_ZNODE
||
704 (S_ISBLK(vap
->va_mode
) || S_ISCHR(vap
->va_mode
))) {
705 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_RDEV(zsb
),
708 if (obj_type
== DMU_OT_ZNODE
) {
709 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_FLAGS(zsb
),
711 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_UID(zsb
), NULL
,
712 &acl_ids
->z_fuid
, 8);
713 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_GID(zsb
), NULL
,
714 &acl_ids
->z_fgid
, 8);
715 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_PAD(zsb
), NULL
, pad
,
716 sizeof (uint64_t) * 4);
717 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_ZNODE_ACL(zsb
), NULL
,
718 &acl_phys
, sizeof (zfs_acl_phys_t
));
719 } else if (acl_ids
->z_aclp
->z_version
>= ZFS_ACL_VERSION_FUID
) {
720 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_COUNT(zsb
), NULL
,
721 &acl_ids
->z_aclp
->z_acl_count
, 8);
722 locate
.cb_aclp
= acl_ids
->z_aclp
;
723 SA_ADD_BULK_ATTR(sa_attrs
, cnt
, SA_ZPL_DACL_ACES(zsb
),
724 zfs_acl_data_locator
, &locate
,
725 acl_ids
->z_aclp
->z_acl_bytes
);
726 mode
= zfs_mode_compute(mode
, acl_ids
->z_aclp
, &pflags
,
727 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
730 VERIFY(sa_replace_all_by_template(sa_hdl
, sa_attrs
, cnt
, tx
) == 0);
732 if (!(flag
& IS_ROOT_NODE
)) {
733 *zpp
= zfs_znode_alloc(zsb
, db
, 0, obj_type
, obj
, sa_hdl
,
735 VERIFY(*zpp
!= NULL
);
739 * If we are creating the root node, the "parent" we
740 * passed in is the znode for the root.
744 (*zpp
)->z_sa_hdl
= sa_hdl
;
747 (*zpp
)->z_pflags
= pflags
;
748 (*zpp
)->z_mode
= mode
;
750 if (obj_type
== DMU_OT_ZNODE
||
751 acl_ids
->z_aclp
->z_version
< ZFS_ACL_VERSION_FUID
) {
752 err
= zfs_aclset_common(*zpp
, acl_ids
->z_aclp
, cr
, tx
);
755 kmem_free(sa_attrs
, sizeof (sa_bulk_attr_t
) * ZPL_END
);
756 ZFS_OBJ_HOLD_EXIT(zsb
, obj
);
760 * Update in-core attributes. It is assumed the caller will be doing an
761 * sa_bulk_update to push the changes out.
764 zfs_xvattr_set(znode_t
*zp
, xvattr_t
*xvap
, dmu_tx_t
*tx
)
768 xoap
= xva_getxoptattr(xvap
);
771 if (XVA_ISSET_REQ(xvap
, XAT_CREATETIME
)) {
773 ZFS_TIME_ENCODE(&xoap
->xoa_createtime
, times
);
774 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_CRTIME(ZTOZSB(zp
)),
775 ×
, sizeof (times
), tx
);
776 XVA_SET_RTN(xvap
, XAT_CREATETIME
);
778 if (XVA_ISSET_REQ(xvap
, XAT_READONLY
)) {
779 ZFS_ATTR_SET(zp
, ZFS_READONLY
, xoap
->xoa_readonly
,
781 XVA_SET_RTN(xvap
, XAT_READONLY
);
783 if (XVA_ISSET_REQ(xvap
, XAT_HIDDEN
)) {
784 ZFS_ATTR_SET(zp
, ZFS_HIDDEN
, xoap
->xoa_hidden
,
786 XVA_SET_RTN(xvap
, XAT_HIDDEN
);
788 if (XVA_ISSET_REQ(xvap
, XAT_SYSTEM
)) {
789 ZFS_ATTR_SET(zp
, ZFS_SYSTEM
, xoap
->xoa_system
,
791 XVA_SET_RTN(xvap
, XAT_SYSTEM
);
793 if (XVA_ISSET_REQ(xvap
, XAT_ARCHIVE
)) {
794 ZFS_ATTR_SET(zp
, ZFS_ARCHIVE
, xoap
->xoa_archive
,
796 XVA_SET_RTN(xvap
, XAT_ARCHIVE
);
798 if (XVA_ISSET_REQ(xvap
, XAT_IMMUTABLE
)) {
799 ZFS_ATTR_SET(zp
, ZFS_IMMUTABLE
, xoap
->xoa_immutable
,
801 XVA_SET_RTN(xvap
, XAT_IMMUTABLE
);
803 if (XVA_ISSET_REQ(xvap
, XAT_NOUNLINK
)) {
804 ZFS_ATTR_SET(zp
, ZFS_NOUNLINK
, xoap
->xoa_nounlink
,
806 XVA_SET_RTN(xvap
, XAT_NOUNLINK
);
808 if (XVA_ISSET_REQ(xvap
, XAT_APPENDONLY
)) {
809 ZFS_ATTR_SET(zp
, ZFS_APPENDONLY
, xoap
->xoa_appendonly
,
811 XVA_SET_RTN(xvap
, XAT_APPENDONLY
);
813 if (XVA_ISSET_REQ(xvap
, XAT_NODUMP
)) {
814 ZFS_ATTR_SET(zp
, ZFS_NODUMP
, xoap
->xoa_nodump
,
816 XVA_SET_RTN(xvap
, XAT_NODUMP
);
818 if (XVA_ISSET_REQ(xvap
, XAT_OPAQUE
)) {
819 ZFS_ATTR_SET(zp
, ZFS_OPAQUE
, xoap
->xoa_opaque
,
821 XVA_SET_RTN(xvap
, XAT_OPAQUE
);
823 if (XVA_ISSET_REQ(xvap
, XAT_AV_QUARANTINED
)) {
824 ZFS_ATTR_SET(zp
, ZFS_AV_QUARANTINED
,
825 xoap
->xoa_av_quarantined
, zp
->z_pflags
, tx
);
826 XVA_SET_RTN(xvap
, XAT_AV_QUARANTINED
);
828 if (XVA_ISSET_REQ(xvap
, XAT_AV_MODIFIED
)) {
829 ZFS_ATTR_SET(zp
, ZFS_AV_MODIFIED
, xoap
->xoa_av_modified
,
831 XVA_SET_RTN(xvap
, XAT_AV_MODIFIED
);
833 if (XVA_ISSET_REQ(xvap
, XAT_AV_SCANSTAMP
)) {
834 zfs_sa_set_scanstamp(zp
, xvap
, tx
);
835 XVA_SET_RTN(xvap
, XAT_AV_SCANSTAMP
);
837 if (XVA_ISSET_REQ(xvap
, XAT_REPARSE
)) {
838 ZFS_ATTR_SET(zp
, ZFS_REPARSE
, xoap
->xoa_reparse
,
840 XVA_SET_RTN(xvap
, XAT_REPARSE
);
842 if (XVA_ISSET_REQ(xvap
, XAT_OFFLINE
)) {
843 ZFS_ATTR_SET(zp
, ZFS_OFFLINE
, xoap
->xoa_offline
,
845 XVA_SET_RTN(xvap
, XAT_OFFLINE
);
847 if (XVA_ISSET_REQ(xvap
, XAT_SPARSE
)) {
848 ZFS_ATTR_SET(zp
, ZFS_SPARSE
, xoap
->xoa_sparse
,
850 XVA_SET_RTN(xvap
, XAT_SPARSE
);
855 zfs_zget(zfs_sb_t
*zsb
, uint64_t obj_num
, znode_t
**zpp
)
857 dmu_object_info_t doi
;
866 ZFS_OBJ_HOLD_ENTER(zsb
, obj_num
);
868 err
= sa_buf_hold(zsb
->z_os
, obj_num
, NULL
, &db
);
870 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
874 dmu_object_info_from_db(db
, &doi
);
875 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
876 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
877 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
878 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
879 sa_buf_rele(db
, NULL
);
880 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
881 return (SET_ERROR(EINVAL
));
884 hdl
= dmu_buf_get_user(db
);
886 zp
= sa_get_userdata(hdl
);
890 * Since "SA" does immediate eviction we
891 * should never find a sa handle that doesn't
892 * know about the znode.
895 ASSERT3P(zp
, !=, NULL
);
897 mutex_enter(&zp
->z_lock
);
898 ASSERT3U(zp
->z_id
, ==, obj_num
);
899 if (zp
->z_unlinked
) {
900 err
= SET_ERROR(ENOENT
);
903 * If igrab() returns NULL the VFS has independently
904 * determined the inode should be evicted and has
905 * called iput_final() to start the eviction process.
906 * The SA handle is still valid but because the VFS
907 * requires that the eviction succeed we must drop
908 * our locks and references to allow the eviction to
909 * complete. The zfs_zget() may then be retried.
911 * This unlikely case could be optimized by registering
912 * a sops->drop_inode() callback. The callback would
913 * need to detect the active SA hold thereby informing
914 * the VFS that this inode should not be evicted.
916 if (igrab(ZTOI(zp
)) == NULL
) {
917 mutex_exit(&zp
->z_lock
);
918 sa_buf_rele(db
, NULL
);
919 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
925 sa_buf_rele(db
, NULL
);
926 mutex_exit(&zp
->z_lock
);
927 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
932 * Not found create new znode/vnode but only if file exists.
934 * There is a small window where zfs_vget() could
935 * find this object while a file create is still in
936 * progress. This is checked for in zfs_znode_alloc()
938 * if zfs_znode_alloc() fails it will drop the hold on the
941 zp
= zfs_znode_alloc(zsb
, db
, doi
.doi_data_block_size
,
942 doi
.doi_bonus_type
, obj_num
, NULL
, NULL
);
944 err
= SET_ERROR(ENOENT
);
948 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
953 zfs_rezget(znode_t
*zp
)
955 zfs_sb_t
*zsb
= ZTOZSB(zp
);
956 dmu_object_info_t doi
;
958 uint64_t obj_num
= zp
->z_id
;
960 sa_bulk_attr_t bulk
[8];
965 ZFS_OBJ_HOLD_ENTER(zsb
, obj_num
);
967 mutex_enter(&zp
->z_acl_lock
);
968 if (zp
->z_acl_cached
) {
969 zfs_acl_free(zp
->z_acl_cached
);
970 zp
->z_acl_cached
= NULL
;
972 mutex_exit(&zp
->z_acl_lock
);
974 rw_enter(&zp
->z_xattr_lock
, RW_WRITER
);
975 if (zp
->z_xattr_cached
) {
976 nvlist_free(zp
->z_xattr_cached
);
977 zp
->z_xattr_cached
= NULL
;
980 if (zp
->z_xattr_parent
) {
981 iput(ZTOI(zp
->z_xattr_parent
));
982 zp
->z_xattr_parent
= NULL
;
984 rw_exit(&zp
->z_xattr_lock
);
986 ASSERT(zp
->z_sa_hdl
== NULL
);
987 err
= sa_buf_hold(zsb
->z_os
, obj_num
, NULL
, &db
);
989 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
993 dmu_object_info_from_db(db
, &doi
);
994 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
995 (doi
.doi_bonus_type
!= DMU_OT_ZNODE
||
996 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
997 doi
.doi_bonus_size
< sizeof (znode_phys_t
)))) {
998 sa_buf_rele(db
, NULL
);
999 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1000 return (SET_ERROR(EINVAL
));
1003 zfs_znode_sa_init(zsb
, zp
, db
, doi
.doi_bonus_type
, NULL
);
1005 /* reload cached values */
1006 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GEN(zsb
), NULL
,
1007 &gen
, sizeof (gen
));
1008 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
), NULL
,
1009 &zp
->z_size
, sizeof (zp
->z_size
));
1010 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_LINKS(zsb
), NULL
,
1011 &zp
->z_links
, sizeof (zp
->z_links
));
1012 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
1013 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1014 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ATIME(zsb
), NULL
,
1015 &zp
->z_atime
, sizeof (zp
->z_atime
));
1016 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_UID(zsb
), NULL
,
1017 &zp
->z_uid
, sizeof (zp
->z_uid
));
1018 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_GID(zsb
), NULL
,
1019 &zp
->z_gid
, sizeof (zp
->z_gid
));
1020 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
,
1021 &mode
, sizeof (mode
));
1023 if (sa_bulk_lookup(zp
->z_sa_hdl
, bulk
, count
)) {
1024 zfs_znode_dmu_fini(zp
);
1025 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1026 return (SET_ERROR(EIO
));
1031 if (gen
!= zp
->z_gen
) {
1032 zfs_znode_dmu_fini(zp
);
1033 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1034 return (SET_ERROR(EIO
));
1037 zp
->z_unlinked
= (zp
->z_links
== 0);
1038 zp
->z_blksz
= doi
.doi_data_block_size
;
1039 zfs_inode_update(zp
);
1041 ZFS_OBJ_HOLD_EXIT(zsb
, obj_num
);
1047 zfs_znode_delete(znode_t
*zp
, dmu_tx_t
*tx
)
1049 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1050 objset_t
*os
= zsb
->z_os
;
1051 uint64_t obj
= zp
->z_id
;
1052 uint64_t acl_obj
= zfs_external_acl(zp
);
1054 ZFS_OBJ_HOLD_ENTER(zsb
, obj
);
1056 VERIFY(!zp
->z_is_sa
);
1057 VERIFY(0 == dmu_object_free(os
, acl_obj
, tx
));
1059 VERIFY(0 == dmu_object_free(os
, obj
, tx
));
1060 zfs_znode_dmu_fini(zp
);
1061 ZFS_OBJ_HOLD_EXIT(zsb
, obj
);
1065 zfs_zinactive(znode_t
*zp
)
1067 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1068 uint64_t z_id
= zp
->z_id
;
1069 boolean_t drop_mutex
= 0;
1071 ASSERT(zp
->z_sa_hdl
);
1074 * Don't allow a zfs_zget() while were trying to release this znode.
1076 * Linux allows direct memory reclaim which means that any KM_SLEEP
1077 * allocation may trigger inode eviction. This can lead to a deadlock
1078 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
1079 * zfs_zinactive() call path. To avoid this deadlock the process
1080 * must not reacquire the mutex when it is already holding it.
1082 if (!ZFS_OBJ_HOLD_OWNED(zsb
, z_id
)) {
1083 ZFS_OBJ_HOLD_ENTER(zsb
, z_id
);
1087 mutex_enter(&zp
->z_lock
);
1090 * If this was the last reference to a file with no links,
1091 * remove the file from the file system.
1093 if (zp
->z_unlinked
) {
1094 mutex_exit(&zp
->z_lock
);
1097 ZFS_OBJ_HOLD_EXIT(zsb
, z_id
);
1103 mutex_exit(&zp
->z_lock
);
1104 zfs_znode_dmu_fini(zp
);
1107 ZFS_OBJ_HOLD_EXIT(zsb
, z_id
);
1111 zfs_compare_timespec(struct timespec
*t1
, struct timespec
*t2
)
1113 if (t1
->tv_sec
< t2
->tv_sec
)
1116 if (t1
->tv_sec
> t2
->tv_sec
)
1119 return (t1
->tv_nsec
- t2
->tv_nsec
);
1123 * Determine whether the znode's atime must be updated. The logic mostly
1124 * duplicates the Linux kernel's relatime_need_update() functionality.
1125 * This function is only called if the underlying filesystem actually has
1126 * atime updates enabled.
1128 static inline boolean_t
1129 zfs_atime_need_update(znode_t
*zp
, timestruc_t
*now
)
1131 if (!ZTOZSB(zp
)->z_relatime
)
1135 * In relatime mode, only update the atime if the previous atime
1136 * is earlier than either the ctime or mtime or if at least a day
1137 * has passed since the last update of atime.
1139 if (zfs_compare_timespec(&ZTOI(zp
)->i_mtime
, &ZTOI(zp
)->i_atime
) >= 0)
1142 if (zfs_compare_timespec(&ZTOI(zp
)->i_ctime
, &ZTOI(zp
)->i_atime
) >= 0)
1145 if ((long)now
->tv_sec
- ZTOI(zp
)->i_atime
.tv_sec
>= 24*60*60)
1152 * Prepare to update znode time stamps.
1154 * IN: zp - znode requiring timestamp update
1155 * flag - ATTR_MTIME, ATTR_CTIME, ATTR_ATIME flags
1156 * have_tx - true of caller is creating a new txg
1158 * OUT: zp - new atime (via underlying inode's i_atime)
1162 * NOTE: The arguments are somewhat redundant. The following condition
1165 * have_tx == !(flag & ATTR_ATIME)
1168 zfs_tstamp_update_setup(znode_t
*zp
, uint_t flag
, uint64_t mtime
[2],
1169 uint64_t ctime
[2], boolean_t have_tx
)
1173 ASSERT(have_tx
== !(flag
& ATTR_ATIME
));
1177 * NOTE: The following test intentionally does not update z_atime_dirty
1178 * in the case where an ATIME update has been requested but for which
1179 * the update is omitted due to relatime logic. The rationale being
1180 * that if the flag was set somewhere else, we should leave it alone
1183 if (flag
& ATTR_ATIME
) {
1184 if (zfs_atime_need_update(zp
, &now
)) {
1185 ZFS_TIME_ENCODE(&now
, zp
->z_atime
);
1186 ZTOI(zp
)->i_atime
.tv_sec
= zp
->z_atime
[0];
1187 ZTOI(zp
)->i_atime
.tv_nsec
= zp
->z_atime
[1];
1188 zp
->z_atime_dirty
= 1;
1191 zp
->z_atime_dirty
= 0;
1195 if (flag
& ATTR_MTIME
) {
1196 ZFS_TIME_ENCODE(&now
, mtime
);
1197 if (ZTOZSB(zp
)->z_use_fuids
) {
1198 zp
->z_pflags
|= (ZFS_ARCHIVE
|
1203 if (flag
& ATTR_CTIME
) {
1204 ZFS_TIME_ENCODE(&now
, ctime
);
1205 if (ZTOZSB(zp
)->z_use_fuids
)
1206 zp
->z_pflags
|= ZFS_ARCHIVE
;
1211 * Grow the block size for a file.
1213 * IN: zp - znode of file to free data in.
1214 * size - requested block size
1215 * tx - open transaction.
1217 * NOTE: this function assumes that the znode is write locked.
1220 zfs_grow_blocksize(znode_t
*zp
, uint64_t size
, dmu_tx_t
*tx
)
1225 if (size
<= zp
->z_blksz
)
1228 * If the file size is already greater than the current blocksize,
1229 * we will not grow. If there is more than one block in a file,
1230 * the blocksize cannot change.
1232 if (zp
->z_blksz
&& zp
->z_size
> zp
->z_blksz
)
1235 error
= dmu_object_set_blocksize(ZTOZSB(zp
)->z_os
, zp
->z_id
,
1238 if (error
== ENOTSUP
)
1242 /* What blocksize did we actually get? */
1243 dmu_object_size_from_db(sa_get_db(zp
->z_sa_hdl
), &zp
->z_blksz
, &dummy
);
1247 * Increase the file length
1249 * IN: zp - znode of file to free data in.
1250 * end - new end-of-file
1252 * RETURN: 0 on success, error code on failure
1255 zfs_extend(znode_t
*zp
, uint64_t end
)
1257 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1264 * We will change zp_size, lock the whole file.
1266 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1269 * Nothing to do if file already at desired length.
1271 if (end
<= zp
->z_size
) {
1272 zfs_range_unlock(rl
);
1275 tx
= dmu_tx_create(zsb
->z_os
);
1276 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1277 zfs_sa_upgrade_txholds(tx
, zp
);
1278 if (end
> zp
->z_blksz
&&
1279 (!ISP2(zp
->z_blksz
) || zp
->z_blksz
< zsb
->z_max_blksz
)) {
1281 * We are growing the file past the current block size.
1283 if (zp
->z_blksz
> ZTOZSB(zp
)->z_max_blksz
) {
1284 ASSERT(!ISP2(zp
->z_blksz
));
1285 newblksz
= MIN(end
, SPA_MAXBLOCKSIZE
);
1287 newblksz
= MIN(end
, ZTOZSB(zp
)->z_max_blksz
);
1289 dmu_tx_hold_write(tx
, zp
->z_id
, 0, newblksz
);
1294 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1297 zfs_range_unlock(rl
);
1302 zfs_grow_blocksize(zp
, newblksz
, tx
);
1306 VERIFY(0 == sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(ZTOZSB(zp
)),
1307 &zp
->z_size
, sizeof (zp
->z_size
), tx
));
1309 zfs_range_unlock(rl
);
1317 * Free space in a file.
1319 * IN: zp - znode of file to free data in.
1320 * off - start of section to free.
1321 * len - length of section to free.
1323 * RETURN: 0 on success, error code on failure
1326 zfs_free_range(znode_t
*zp
, uint64_t off
, uint64_t len
)
1328 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1333 * Lock the range being freed.
1335 rl
= zfs_range_lock(zp
, off
, len
, RL_WRITER
);
1338 * Nothing to do if file already at desired length.
1340 if (off
>= zp
->z_size
) {
1341 zfs_range_unlock(rl
);
1345 if (off
+ len
> zp
->z_size
)
1346 len
= zp
->z_size
- off
;
1348 error
= dmu_free_long_range(zsb
->z_os
, zp
->z_id
, off
, len
);
1350 zfs_range_unlock(rl
);
1358 * IN: zp - znode of file to free data in.
1359 * end - new end-of-file.
1361 * RETURN: 0 on success, error code on failure
1364 zfs_trunc(znode_t
*zp
, uint64_t end
)
1366 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1370 sa_bulk_attr_t bulk
[2];
1374 * We will change zp_size, lock the whole file.
1376 rl
= zfs_range_lock(zp
, 0, UINT64_MAX
, RL_WRITER
);
1379 * Nothing to do if file already at desired length.
1381 if (end
>= zp
->z_size
) {
1382 zfs_range_unlock(rl
);
1386 error
= dmu_free_long_range(zsb
->z_os
, zp
->z_id
, end
, -1);
1388 zfs_range_unlock(rl
);
1392 tx
= dmu_tx_create(zsb
->z_os
);
1393 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1394 zfs_sa_upgrade_txholds(tx
, zp
);
1395 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
1397 if (error
== ERESTART
) {
1403 zfs_range_unlock(rl
);
1408 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_SIZE(zsb
),
1409 NULL
, &zp
->z_size
, sizeof (zp
->z_size
));
1412 zp
->z_pflags
&= ~ZFS_SPARSE
;
1413 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
1414 NULL
, &zp
->z_pflags
, 8);
1416 VERIFY(sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
) == 0);
1420 zfs_range_unlock(rl
);
1426 * Free space in a file
1428 * IN: zp - znode of file to free data in.
1429 * off - start of range
1430 * len - end of range (0 => EOF)
1431 * flag - current file open mode flags.
1432 * log - TRUE if this action should be logged
1434 * RETURN: 0 on success, error code on failure
1437 zfs_freesp(znode_t
*zp
, uint64_t off
, uint64_t len
, int flag
, boolean_t log
)
1439 struct inode
*ip
= ZTOI(zp
);
1441 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1442 zilog_t
*zilog
= zsb
->z_log
;
1444 uint64_t mtime
[2], ctime
[2];
1445 sa_bulk_attr_t bulk
[3];
1449 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_MODE(zsb
), &mode
,
1450 sizeof (mode
))) != 0)
1453 if (off
> zp
->z_size
) {
1454 error
= zfs_extend(zp
, off
+len
);
1455 if (error
== 0 && log
)
1462 * Check for any locks in the region to be freed.
1464 if (ip
->i_flock
&& mandatory_lock(ip
)) {
1465 uint64_t length
= (len
? len
: zp
->z_size
- off
);
1466 if (!lock_may_write(ip
, off
, length
))
1467 return (SET_ERROR(EAGAIN
));
1471 error
= zfs_trunc(zp
, off
);
1473 if ((error
= zfs_free_range(zp
, off
, len
)) == 0 &&
1474 off
+ len
> zp
->z_size
)
1475 error
= zfs_extend(zp
, off
+len
);
1480 tx
= dmu_tx_create(zsb
->z_os
);
1481 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
1482 zfs_sa_upgrade_txholds(tx
, zp
);
1483 error
= dmu_tx_assign(tx
, TXG_WAIT
);
1489 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MTIME(zsb
), NULL
, mtime
, 16);
1490 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
, ctime
, 16);
1491 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
),
1492 NULL
, &zp
->z_pflags
, 8);
1493 zfs_tstamp_update_setup(zp
, CONTENT_MODIFIED
, mtime
, ctime
, B_TRUE
);
1494 error
= sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
);
1497 zfs_log_truncate(zilog
, tx
, TX_TRUNCATE
, zp
, off
, len
);
1500 zfs_inode_update(zp
);
1505 zfs_create_fs(objset_t
*os
, cred_t
*cr
, nvlist_t
*zplprops
, dmu_tx_t
*tx
)
1507 struct super_block
*sb
;
1509 uint64_t moid
, obj
, sa_obj
, version
;
1510 uint64_t sense
= ZFS_CASE_SENSITIVE
;
1515 znode_t
*rootzp
= NULL
;
1518 zfs_acl_ids_t acl_ids
;
1521 * First attempt to create master node.
1524 * In an empty objset, there are no blocks to read and thus
1525 * there can be no i/o errors (which we assert below).
1527 moid
= MASTER_NODE_OBJ
;
1528 error
= zap_create_claim(os
, moid
, DMU_OT_MASTER_NODE
,
1529 DMU_OT_NONE
, 0, tx
);
1533 * Set starting attributes.
1535 version
= zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)));
1537 while ((elem
= nvlist_next_nvpair(zplprops
, elem
)) != NULL
) {
1538 /* For the moment we expect all zpl props to be uint64_ts */
1542 ASSERT(nvpair_type(elem
) == DATA_TYPE_UINT64
);
1543 VERIFY(nvpair_value_uint64(elem
, &val
) == 0);
1544 name
= nvpair_name(elem
);
1545 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_VERSION
)) == 0) {
1549 error
= zap_update(os
, moid
, name
, 8, 1, &val
, tx
);
1552 if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_NORMALIZE
)) == 0)
1554 else if (strcmp(name
, zfs_prop_to_name(ZFS_PROP_CASE
)) == 0)
1557 ASSERT(version
!= 0);
1558 error
= zap_update(os
, moid
, ZPL_VERSION_STR
, 8, 1, &version
, tx
);
1561 * Create zap object used for SA attribute registration
1564 if (version
>= ZPL_VERSION_SA
) {
1565 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
1566 DMU_OT_NONE
, 0, tx
);
1567 error
= zap_add(os
, moid
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
1573 * Create a delete queue.
1575 obj
= zap_create(os
, DMU_OT_UNLINKED_SET
, DMU_OT_NONE
, 0, tx
);
1577 error
= zap_add(os
, moid
, ZFS_UNLINKED_SET
, 8, 1, &obj
, tx
);
1581 * Create root znode. Create minimal znode/inode/zsb/sb
1582 * to allow zfs_mknode to work.
1584 vattr
.va_mask
= ATTR_MODE
|ATTR_UID
|ATTR_GID
;
1585 vattr
.va_mode
= S_IFDIR
|0755;
1586 vattr
.va_uid
= crgetuid(cr
);
1587 vattr
.va_gid
= crgetgid(cr
);
1589 rootzp
= kmem_cache_alloc(znode_cache
, KM_PUSHPAGE
);
1590 rootzp
->z_moved
= 0;
1591 rootzp
->z_unlinked
= 0;
1592 rootzp
->z_atime_dirty
= 0;
1593 rootzp
->z_is_sa
= USE_SA(version
, os
);
1595 zsb
= kmem_zalloc(sizeof (zfs_sb_t
), KM_PUSHPAGE
| KM_NODEBUG
);
1597 zsb
->z_parent
= zsb
;
1598 zsb
->z_version
= version
;
1599 zsb
->z_use_fuids
= USE_FUIDS(version
, os
);
1600 zsb
->z_use_sa
= USE_SA(version
, os
);
1603 sb
= kmem_zalloc(sizeof (struct super_block
), KM_PUSHPAGE
);
1604 sb
->s_fs_info
= zsb
;
1606 ZTOI(rootzp
)->i_sb
= sb
;
1608 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
1609 &zsb
->z_attr_table
);
1614 * Fold case on file systems that are always or sometimes case
1617 if (sense
== ZFS_CASE_INSENSITIVE
|| sense
== ZFS_CASE_MIXED
)
1618 zsb
->z_norm
|= U8_TEXTPREP_TOUPPER
;
1620 mutex_init(&zsb
->z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1621 list_create(&zsb
->z_all_znodes
, sizeof (znode_t
),
1622 offsetof(znode_t
, z_link_node
));
1624 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1625 mutex_init(&zsb
->z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
1627 VERIFY(0 == zfs_acl_ids_create(rootzp
, IS_ROOT_NODE
, &vattr
,
1628 cr
, NULL
, &acl_ids
));
1629 zfs_mknode(rootzp
, &vattr
, tx
, cr
, IS_ROOT_NODE
, &zp
, &acl_ids
);
1630 ASSERT3P(zp
, ==, rootzp
);
1631 error
= zap_add(os
, moid
, ZFS_ROOT_OBJ
, 8, 1, &rootzp
->z_id
, tx
);
1633 zfs_acl_ids_free(&acl_ids
);
1635 atomic_set(&ZTOI(rootzp
)->i_count
, 0);
1636 sa_handle_destroy(rootzp
->z_sa_hdl
);
1637 kmem_cache_free(znode_cache
, rootzp
);
1640 * Create shares directory
1642 error
= zfs_create_share_dir(zsb
, tx
);
1645 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1646 mutex_destroy(&zsb
->z_hold_mtx
[i
]);
1648 kmem_free(sb
, sizeof (struct super_block
));
1649 kmem_free(zsb
, sizeof (zfs_sb_t
));
1651 #endif /* _KERNEL */
1654 zfs_sa_setup(objset_t
*osp
, sa_attr_type_t
**sa_table
)
1656 uint64_t sa_obj
= 0;
1659 error
= zap_lookup(osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
1660 if (error
!= 0 && error
!= ENOENT
)
1663 error
= sa_setup(osp
, sa_obj
, zfs_attr_table
, ZPL_END
, sa_table
);
1668 zfs_grab_sa_handle(objset_t
*osp
, uint64_t obj
, sa_handle_t
**hdlp
,
1669 dmu_buf_t
**db
, void *tag
)
1671 dmu_object_info_t doi
;
1674 if ((error
= sa_buf_hold(osp
, obj
, tag
, db
)) != 0)
1677 dmu_object_info_from_db(*db
, &doi
);
1678 if ((doi
.doi_bonus_type
!= DMU_OT_SA
&&
1679 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) ||
1680 (doi
.doi_bonus_type
== DMU_OT_ZNODE
&&
1681 doi
.doi_bonus_size
< sizeof (znode_phys_t
))) {
1682 sa_buf_rele(*db
, tag
);
1683 return (SET_ERROR(ENOTSUP
));
1686 error
= sa_handle_get(osp
, obj
, NULL
, SA_HDL_PRIVATE
, hdlp
);
1688 sa_buf_rele(*db
, tag
);
1696 zfs_release_sa_handle(sa_handle_t
*hdl
, dmu_buf_t
*db
, void *tag
)
1698 sa_handle_destroy(hdl
);
1699 sa_buf_rele(db
, tag
);
1703 * Given an object number, return its parent object number and whether
1704 * or not the object is an extended attribute directory.
1707 zfs_obj_to_pobj(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
, uint64_t *pobjp
,
1713 sa_bulk_attr_t bulk
[3];
1717 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_PARENT
], NULL
,
1718 &parent
, sizeof (parent
));
1719 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_FLAGS
], NULL
,
1720 &pflags
, sizeof (pflags
));
1721 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1722 &mode
, sizeof (mode
));
1724 if ((error
= sa_bulk_lookup(hdl
, bulk
, count
)) != 0)
1728 *is_xattrdir
= ((pflags
& ZFS_XATTR
) != 0) && S_ISDIR(mode
);
1734 * Given an object number, return some zpl level statistics
1737 zfs_obj_to_stats_impl(sa_handle_t
*hdl
, sa_attr_type_t
*sa_table
,
1740 sa_bulk_attr_t bulk
[4];
1743 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_MODE
], NULL
,
1744 &sb
->zs_mode
, sizeof (sb
->zs_mode
));
1745 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_GEN
], NULL
,
1746 &sb
->zs_gen
, sizeof (sb
->zs_gen
));
1747 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_LINKS
], NULL
,
1748 &sb
->zs_links
, sizeof (sb
->zs_links
));
1749 SA_ADD_BULK_ATTR(bulk
, count
, sa_table
[ZPL_CTIME
], NULL
,
1750 &sb
->zs_ctime
, sizeof (sb
->zs_ctime
));
1752 return (sa_bulk_lookup(hdl
, bulk
, count
));
1756 zfs_obj_to_path_impl(objset_t
*osp
, uint64_t obj
, sa_handle_t
*hdl
,
1757 sa_attr_type_t
*sa_table
, char *buf
, int len
)
1759 sa_handle_t
*sa_hdl
;
1760 sa_handle_t
*prevhdl
= NULL
;
1761 dmu_buf_t
*prevdb
= NULL
;
1762 dmu_buf_t
*sa_db
= NULL
;
1763 char *path
= buf
+ len
- 1;
1771 char component
[MAXNAMELEN
+ 2];
1773 int is_xattrdir
= 0;
1776 zfs_release_sa_handle(prevhdl
, prevdb
, FTAG
);
1778 if ((error
= zfs_obj_to_pobj(sa_hdl
, sa_table
, &pobj
,
1779 &is_xattrdir
)) != 0)
1790 (void) sprintf(component
+ 1, "<xattrdir>");
1792 error
= zap_value_search(osp
, pobj
, obj
,
1793 ZFS_DIRENT_OBJ(-1ULL), component
+ 1);
1798 complen
= strlen(component
);
1800 ASSERT(path
>= buf
);
1801 bcopy(component
, path
, complen
);
1804 if (sa_hdl
!= hdl
) {
1808 error
= zfs_grab_sa_handle(osp
, obj
, &sa_hdl
, &sa_db
, FTAG
);
1816 if (sa_hdl
!= NULL
&& sa_hdl
!= hdl
) {
1817 ASSERT(sa_db
!= NULL
);
1818 zfs_release_sa_handle(sa_hdl
, sa_db
, FTAG
);
1822 (void) memmove(buf
, path
, buf
+ len
- path
);
1828 zfs_obj_to_path(objset_t
*osp
, uint64_t obj
, char *buf
, int len
)
1830 sa_attr_type_t
*sa_table
;
1835 error
= zfs_sa_setup(osp
, &sa_table
);
1839 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
1843 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
1845 zfs_release_sa_handle(hdl
, db
, FTAG
);
1850 zfs_obj_to_stats(objset_t
*osp
, uint64_t obj
, zfs_stat_t
*sb
,
1853 char *path
= buf
+ len
- 1;
1854 sa_attr_type_t
*sa_table
;
1861 error
= zfs_sa_setup(osp
, &sa_table
);
1865 error
= zfs_grab_sa_handle(osp
, obj
, &hdl
, &db
, FTAG
);
1869 error
= zfs_obj_to_stats_impl(hdl
, sa_table
, sb
);
1871 zfs_release_sa_handle(hdl
, db
, FTAG
);
1875 error
= zfs_obj_to_path_impl(osp
, obj
, hdl
, sa_table
, buf
, len
);
1877 zfs_release_sa_handle(hdl
, db
, FTAG
);
1881 #if defined(_KERNEL) && defined(HAVE_SPL)
1882 EXPORT_SYMBOL(zfs_create_fs
);
1883 EXPORT_SYMBOL(zfs_obj_to_path
);