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 by Delphix. All rights reserved.
26 /* Portions Copyright 2010 Robert Milkowski */
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
33 #include <sys/pathname.h>
34 #include <sys/vnode.h>
36 #include <sys/vfs_opreg.h>
37 #include <sys/mntent.h>
38 #include <sys/mount.h>
39 #include <sys/cmn_err.h>
40 #include "fs/fs_subr.h"
41 #include <sys/zfs_znode.h>
42 #include <sys/zfs_dir.h>
44 #include <sys/fs/zfs.h>
46 #include <sys/dsl_prop.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_deleg.h>
52 #include <sys/varargs.h>
53 #include <sys/policy.h>
54 #include <sys/atomic.h>
55 #include <sys/mkdev.h>
56 #include <sys/modctl.h>
57 #include <sys/refstr.h>
58 #include <sys/zfs_ioctl.h>
59 #include <sys/zfs_ctldir.h>
60 #include <sys/zfs_fuid.h>
61 #include <sys/bootconf.h>
62 #include <sys/sunddi.h>
64 #include <sys/dmu_objset.h>
65 #include <sys/spa_boot.h>
67 #include "zfs_comutil.h"
70 vfsops_t
*zfs_vfsops
= NULL
;
71 static major_t zfs_major
;
72 static minor_t zfs_minor
;
73 static kmutex_t zfs_dev_mtx
;
75 extern int sys_shutdown
;
77 static int zfs_mount(vfs_t
*vfsp
, vnode_t
*mvp
, struct mounta
*uap
, cred_t
*cr
);
78 static int zfs_umount(vfs_t
*vfsp
, int fflag
, cred_t
*cr
);
79 static int zfs_mountroot(vfs_t
*vfsp
, enum whymountroot
);
80 static int zfs_root(vfs_t
*vfsp
, vnode_t
**vpp
);
81 static int zfs_statvfs(vfs_t
*vfsp
, struct statvfs64
*statp
);
82 static int zfs_vget(vfs_t
*vfsp
, vnode_t
**vpp
, fid_t
*fidp
);
83 static void zfs_freevfs(vfs_t
*vfsp
);
85 static const fs_operation_def_t zfs_vfsops_template
[] = {
86 VFSNAME_MOUNT
, { .vfs_mount
= zfs_mount
},
87 VFSNAME_MOUNTROOT
, { .vfs_mountroot
= zfs_mountroot
},
88 VFSNAME_UNMOUNT
, { .vfs_unmount
= zfs_umount
},
89 VFSNAME_ROOT
, { .vfs_root
= zfs_root
},
90 VFSNAME_STATVFS
, { .vfs_statvfs
= zfs_statvfs
},
91 VFSNAME_SYNC
, { .vfs_sync
= zfs_sync
},
92 VFSNAME_VGET
, { .vfs_vget
= zfs_vget
},
93 VFSNAME_FREEVFS
, { .vfs_freevfs
= zfs_freevfs
},
97 static const fs_operation_def_t zfs_vfsops_eio_template
[] = {
98 VFSNAME_FREEVFS
, { .vfs_freevfs
= zfs_freevfs
},
103 * We need to keep a count of active fs's.
104 * This is necessary to prevent our module
105 * from being unloaded after a umount -f
107 static uint32_t zfs_active_fs_count
= 0;
109 static char *noatime_cancel
[] = { MNTOPT_ATIME
, NULL
};
110 static char *atime_cancel
[] = { MNTOPT_NOATIME
, NULL
};
111 static char *noxattr_cancel
[] = { MNTOPT_XATTR
, NULL
};
112 static char *xattr_cancel
[] = { MNTOPT_NOXATTR
, NULL
};
115 * MO_DEFAULT is not used since the default value is determined
116 * by the equivalent property.
118 static mntopt_t mntopts
[] = {
119 { MNTOPT_NOXATTR
, noxattr_cancel
, NULL
, 0, NULL
},
120 { MNTOPT_XATTR
, xattr_cancel
, NULL
, 0, NULL
},
121 { MNTOPT_NOATIME
, noatime_cancel
, NULL
, 0, NULL
},
122 { MNTOPT_ATIME
, atime_cancel
, NULL
, 0, NULL
}
125 static mntopts_t zfs_mntopts
= {
126 sizeof (mntopts
) / sizeof (mntopt_t
),
132 zfs_sync(vfs_t
*vfsp
, short flag
, cred_t
*cr
)
135 * Data integrity is job one. We don't want a compromised kernel
136 * writing to the storage pool, so we never sync during panic.
142 * SYNC_ATTR is used by fsflush() to force old filesystems like UFS
143 * to sync metadata, which they would otherwise cache indefinitely.
144 * Semantically, the only requirement is that the sync be initiated.
145 * The DMU syncs out txgs frequently, so there's nothing to do.
147 if (flag
& SYNC_ATTR
)
152 * Sync a specific filesystem.
154 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
158 dp
= dmu_objset_pool(zfsvfs
->z_os
);
161 * If the system is shutting down, then skip any
162 * filesystems which may exist on a suspended pool.
164 if (sys_shutdown
&& spa_suspended(dp
->dp_spa
)) {
169 if (zfsvfs
->z_log
!= NULL
)
170 zil_commit(zfsvfs
->z_log
, 0);
175 * Sync all ZFS filesystems. This is what happens when you
176 * run sync(1M). Unlike other filesystems, ZFS honors the
177 * request by waiting for all pools to commit all dirty data.
186 zfs_create_unique_device(dev_t
*dev
)
191 ASSERT3U(zfs_minor
, <=, MAXMIN32
);
192 minor_t start
= zfs_minor
;
194 mutex_enter(&zfs_dev_mtx
);
195 if (zfs_minor
>= MAXMIN32
) {
197 * If we're still using the real major
198 * keep out of /dev/zfs and /dev/zvol minor
199 * number space. If we're using a getudev()'ed
200 * major number, we can use all of its minors.
202 if (zfs_major
== ddi_name_to_major(ZFS_DRIVER
))
203 zfs_minor
= ZFS_MIN_MINOR
;
209 *dev
= makedevice(zfs_major
, zfs_minor
);
210 mutex_exit(&zfs_dev_mtx
);
211 } while (vfs_devismounted(*dev
) && zfs_minor
!= start
);
212 if (zfs_minor
== start
) {
214 * We are using all ~262,000 minor numbers for the
215 * current major number. Create a new major number.
217 if ((new_major
= getudev()) == (major_t
)-1) {
219 "zfs_mount: Can't get unique major "
223 mutex_enter(&zfs_dev_mtx
);
224 zfs_major
= new_major
;
227 mutex_exit(&zfs_dev_mtx
);
231 /* CONSTANTCONDITION */
238 atime_changed_cb(void *arg
, uint64_t newval
)
240 zfsvfs_t
*zfsvfs
= arg
;
242 if (newval
== TRUE
) {
243 zfsvfs
->z_atime
= TRUE
;
244 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NOATIME
);
245 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_ATIME
, NULL
, 0);
247 zfsvfs
->z_atime
= FALSE
;
248 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_ATIME
);
249 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NOATIME
, NULL
, 0);
254 xattr_changed_cb(void *arg
, uint64_t newval
)
256 zfsvfs_t
*zfsvfs
= arg
;
258 if (newval
== TRUE
) {
259 /* XXX locking on vfs_flag? */
260 zfsvfs
->z_vfs
->vfs_flag
|= VFS_XATTR
;
261 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NOXATTR
);
262 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_XATTR
, NULL
, 0);
264 /* XXX locking on vfs_flag? */
265 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_XATTR
;
266 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_XATTR
);
267 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NOXATTR
, NULL
, 0);
272 blksz_changed_cb(void *arg
, uint64_t newval
)
274 zfsvfs_t
*zfsvfs
= arg
;
276 if (newval
< SPA_MINBLOCKSIZE
||
277 newval
> SPA_MAXBLOCKSIZE
|| !ISP2(newval
))
278 newval
= SPA_MAXBLOCKSIZE
;
280 zfsvfs
->z_max_blksz
= newval
;
281 zfsvfs
->z_vfs
->vfs_bsize
= newval
;
285 readonly_changed_cb(void *arg
, uint64_t newval
)
287 zfsvfs_t
*zfsvfs
= arg
;
290 /* XXX locking on vfs_flag? */
291 zfsvfs
->z_vfs
->vfs_flag
|= VFS_RDONLY
;
292 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_RW
);
293 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_RO
, NULL
, 0);
295 /* XXX locking on vfs_flag? */
296 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_RDONLY
;
297 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_RO
);
298 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_RW
, NULL
, 0);
303 devices_changed_cb(void *arg
, uint64_t newval
)
305 zfsvfs_t
*zfsvfs
= arg
;
307 if (newval
== FALSE
) {
308 zfsvfs
->z_vfs
->vfs_flag
|= VFS_NODEVICES
;
309 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_DEVICES
);
310 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NODEVICES
, NULL
, 0);
312 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_NODEVICES
;
313 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NODEVICES
);
314 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_DEVICES
, NULL
, 0);
319 setuid_changed_cb(void *arg
, uint64_t newval
)
321 zfsvfs_t
*zfsvfs
= arg
;
323 if (newval
== FALSE
) {
324 zfsvfs
->z_vfs
->vfs_flag
|= VFS_NOSETUID
;
325 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_SETUID
);
326 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NOSETUID
, NULL
, 0);
328 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_NOSETUID
;
329 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NOSETUID
);
330 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_SETUID
, NULL
, 0);
335 exec_changed_cb(void *arg
, uint64_t newval
)
337 zfsvfs_t
*zfsvfs
= arg
;
339 if (newval
== FALSE
) {
340 zfsvfs
->z_vfs
->vfs_flag
|= VFS_NOEXEC
;
341 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_EXEC
);
342 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NOEXEC
, NULL
, 0);
344 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_NOEXEC
;
345 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NOEXEC
);
346 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_EXEC
, NULL
, 0);
351 * The nbmand mount option can be changed at mount time.
352 * We can't allow it to be toggled on live file systems or incorrect
353 * behavior may be seen from cifs clients
355 * This property isn't registered via dsl_prop_register(), but this callback
356 * will be called when a file system is first mounted
359 nbmand_changed_cb(void *arg
, uint64_t newval
)
361 zfsvfs_t
*zfsvfs
= arg
;
362 if (newval
== FALSE
) {
363 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NBMAND
);
364 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NONBMAND
, NULL
, 0);
366 vfs_clearmntopt(zfsvfs
->z_vfs
, MNTOPT_NONBMAND
);
367 vfs_setmntopt(zfsvfs
->z_vfs
, MNTOPT_NBMAND
, NULL
, 0);
372 snapdir_changed_cb(void *arg
, uint64_t newval
)
374 zfsvfs_t
*zfsvfs
= arg
;
376 zfsvfs
->z_show_ctldir
= newval
;
380 vscan_changed_cb(void *arg
, uint64_t newval
)
382 zfsvfs_t
*zfsvfs
= arg
;
384 zfsvfs
->z_vscan
= newval
;
388 acl_mode_changed_cb(void *arg
, uint64_t newval
)
390 zfsvfs_t
*zfsvfs
= arg
;
392 zfsvfs
->z_acl_mode
= newval
;
396 acl_inherit_changed_cb(void *arg
, uint64_t newval
)
398 zfsvfs_t
*zfsvfs
= arg
;
400 zfsvfs
->z_acl_inherit
= newval
;
404 zfs_register_callbacks(vfs_t
*vfsp
)
406 struct dsl_dataset
*ds
= NULL
;
408 zfsvfs_t
*zfsvfs
= NULL
;
410 int readonly
, do_readonly
= B_FALSE
;
411 int setuid
, do_setuid
= B_FALSE
;
412 int exec
, do_exec
= B_FALSE
;
413 int devices
, do_devices
= B_FALSE
;
414 int xattr
, do_xattr
= B_FALSE
;
415 int atime
, do_atime
= B_FALSE
;
419 zfsvfs
= vfsp
->vfs_data
;
424 * The act of registering our callbacks will destroy any mount
425 * options we may have. In order to enable temporary overrides
426 * of mount options, we stash away the current values and
427 * restore them after we register the callbacks.
429 if (vfs_optionisset(vfsp
, MNTOPT_RO
, NULL
) ||
430 !spa_writeable(dmu_objset_spa(os
))) {
432 do_readonly
= B_TRUE
;
433 } else if (vfs_optionisset(vfsp
, MNTOPT_RW
, NULL
)) {
435 do_readonly
= B_TRUE
;
437 if (vfs_optionisset(vfsp
, MNTOPT_NOSUID
, NULL
)) {
443 if (vfs_optionisset(vfsp
, MNTOPT_NODEVICES
, NULL
)) {
446 } else if (vfs_optionisset(vfsp
, MNTOPT_DEVICES
, NULL
)) {
451 if (vfs_optionisset(vfsp
, MNTOPT_NOSETUID
, NULL
)) {
454 } else if (vfs_optionisset(vfsp
, MNTOPT_SETUID
, NULL
)) {
459 if (vfs_optionisset(vfsp
, MNTOPT_NOEXEC
, NULL
)) {
462 } else if (vfs_optionisset(vfsp
, MNTOPT_EXEC
, NULL
)) {
466 if (vfs_optionisset(vfsp
, MNTOPT_NOXATTR
, NULL
)) {
469 } else if (vfs_optionisset(vfsp
, MNTOPT_XATTR
, NULL
)) {
473 if (vfs_optionisset(vfsp
, MNTOPT_NOATIME
, NULL
)) {
476 } else if (vfs_optionisset(vfsp
, MNTOPT_ATIME
, NULL
)) {
482 * nbmand is a special property. It can only be changed at
485 * This is weird, but it is documented to only be changeable
488 if (vfs_optionisset(vfsp
, MNTOPT_NONBMAND
, NULL
)) {
490 } else if (vfs_optionisset(vfsp
, MNTOPT_NBMAND
, NULL
)) {
493 char osname
[MAXNAMELEN
];
495 dmu_objset_name(os
, osname
);
496 if (error
= dsl_prop_get_integer(osname
, "nbmand", &nbmand
,
503 * Register property callbacks.
505 * It would probably be fine to just check for i/o error from
506 * the first prop_register(), but I guess I like to go
509 ds
= dmu_objset_ds(os
);
510 error
= dsl_prop_register(ds
, "atime", atime_changed_cb
, zfsvfs
);
511 error
= error
? error
: dsl_prop_register(ds
,
512 "xattr", xattr_changed_cb
, zfsvfs
);
513 error
= error
? error
: dsl_prop_register(ds
,
514 "recordsize", blksz_changed_cb
, zfsvfs
);
515 error
= error
? error
: dsl_prop_register(ds
,
516 "readonly", readonly_changed_cb
, zfsvfs
);
517 error
= error
? error
: dsl_prop_register(ds
,
518 "devices", devices_changed_cb
, zfsvfs
);
519 error
= error
? error
: dsl_prop_register(ds
,
520 "setuid", setuid_changed_cb
, zfsvfs
);
521 error
= error
? error
: dsl_prop_register(ds
,
522 "exec", exec_changed_cb
, zfsvfs
);
523 error
= error
? error
: dsl_prop_register(ds
,
524 "snapdir", snapdir_changed_cb
, zfsvfs
);
525 error
= error
? error
: dsl_prop_register(ds
,
526 "aclmode", acl_mode_changed_cb
, zfsvfs
);
527 error
= error
? error
: dsl_prop_register(ds
,
528 "aclinherit", acl_inherit_changed_cb
, zfsvfs
);
529 error
= error
? error
: dsl_prop_register(ds
,
530 "vscan", vscan_changed_cb
, zfsvfs
);
535 * Invoke our callbacks to restore temporary mount options.
538 readonly_changed_cb(zfsvfs
, readonly
);
540 setuid_changed_cb(zfsvfs
, setuid
);
542 exec_changed_cb(zfsvfs
, exec
);
544 devices_changed_cb(zfsvfs
, devices
);
546 xattr_changed_cb(zfsvfs
, xattr
);
548 atime_changed_cb(zfsvfs
, atime
);
550 nbmand_changed_cb(zfsvfs
, nbmand
);
556 * We may attempt to unregister some callbacks that are not
557 * registered, but this is OK; it will simply return ENOMSG,
558 * which we will ignore.
560 (void) dsl_prop_unregister(ds
, "atime", atime_changed_cb
, zfsvfs
);
561 (void) dsl_prop_unregister(ds
, "xattr", xattr_changed_cb
, zfsvfs
);
562 (void) dsl_prop_unregister(ds
, "recordsize", blksz_changed_cb
, zfsvfs
);
563 (void) dsl_prop_unregister(ds
, "readonly", readonly_changed_cb
, zfsvfs
);
564 (void) dsl_prop_unregister(ds
, "devices", devices_changed_cb
, zfsvfs
);
565 (void) dsl_prop_unregister(ds
, "setuid", setuid_changed_cb
, zfsvfs
);
566 (void) dsl_prop_unregister(ds
, "exec", exec_changed_cb
, zfsvfs
);
567 (void) dsl_prop_unregister(ds
, "snapdir", snapdir_changed_cb
, zfsvfs
);
568 (void) dsl_prop_unregister(ds
, "aclmode", acl_mode_changed_cb
, zfsvfs
);
569 (void) dsl_prop_unregister(ds
, "aclinherit", acl_inherit_changed_cb
,
571 (void) dsl_prop_unregister(ds
, "vscan", vscan_changed_cb
, zfsvfs
);
577 zfs_space_delta_cb(dmu_object_type_t bonustype
, void *data
,
578 uint64_t *userp
, uint64_t *groupp
)
580 znode_phys_t
*znp
= data
;
584 * Is it a valid type of object to track?
586 if (bonustype
!= DMU_OT_ZNODE
&& bonustype
!= DMU_OT_SA
)
590 * If we have a NULL data pointer
591 * then assume the id's aren't changing and
592 * return EEXIST to the dmu to let it know to
598 if (bonustype
== DMU_OT_ZNODE
) {
599 *userp
= znp
->zp_uid
;
600 *groupp
= znp
->zp_gid
;
604 ASSERT(bonustype
== DMU_OT_SA
);
605 hdrsize
= sa_hdrsize(data
);
608 *userp
= *((uint64_t *)((uintptr_t)data
+ hdrsize
+
610 *groupp
= *((uint64_t *)((uintptr_t)data
+ hdrsize
+
614 * This should only happen for newly created
615 * files that haven't had the znode data filled
626 fuidstr_to_sid(zfsvfs_t
*zfsvfs
, const char *fuidstr
,
627 char *domainbuf
, int buflen
, uid_t
*ridp
)
632 fuid
= strtonum(fuidstr
, NULL
);
634 domain
= zfs_fuid_find_by_idx(zfsvfs
, FUID_INDEX(fuid
));
636 (void) strlcpy(domainbuf
, domain
, buflen
);
639 *ridp
= FUID_RID(fuid
);
643 zfs_userquota_prop_to_obj(zfsvfs_t
*zfsvfs
, zfs_userquota_prop_t type
)
646 case ZFS_PROP_USERUSED
:
647 return (DMU_USERUSED_OBJECT
);
648 case ZFS_PROP_GROUPUSED
:
649 return (DMU_GROUPUSED_OBJECT
);
650 case ZFS_PROP_USERQUOTA
:
651 return (zfsvfs
->z_userquota_obj
);
652 case ZFS_PROP_GROUPQUOTA
:
653 return (zfsvfs
->z_groupquota_obj
);
659 zfs_userspace_many(zfsvfs_t
*zfsvfs
, zfs_userquota_prop_t type
,
660 uint64_t *cookiep
, void *vbuf
, uint64_t *bufsizep
)
665 zfs_useracct_t
*buf
= vbuf
;
668 if (!dmu_objset_userspace_present(zfsvfs
->z_os
))
671 obj
= zfs_userquota_prop_to_obj(zfsvfs
, type
);
677 for (zap_cursor_init_serialized(&zc
, zfsvfs
->z_os
, obj
, *cookiep
);
678 (error
= zap_cursor_retrieve(&zc
, &za
)) == 0;
679 zap_cursor_advance(&zc
)) {
680 if ((uintptr_t)buf
- (uintptr_t)vbuf
+ sizeof (zfs_useracct_t
) >
684 fuidstr_to_sid(zfsvfs
, za
.za_name
,
685 buf
->zu_domain
, sizeof (buf
->zu_domain
), &buf
->zu_rid
);
687 buf
->zu_space
= za
.za_first_integer
;
693 ASSERT3U((uintptr_t)buf
- (uintptr_t)vbuf
, <=, *bufsizep
);
694 *bufsizep
= (uintptr_t)buf
- (uintptr_t)vbuf
;
695 *cookiep
= zap_cursor_serialize(&zc
);
696 zap_cursor_fini(&zc
);
701 * buf must be big enough (eg, 32 bytes)
704 id_to_fuidstr(zfsvfs_t
*zfsvfs
, const char *domain
, uid_t rid
,
705 char *buf
, boolean_t addok
)
710 if (domain
&& domain
[0]) {
711 domainid
= zfs_fuid_find_by_domain(zfsvfs
, domain
, NULL
, addok
);
715 fuid
= FUID_ENCODE(domainid
, rid
);
716 (void) sprintf(buf
, "%llx", (longlong_t
)fuid
);
721 zfs_userspace_one(zfsvfs_t
*zfsvfs
, zfs_userquota_prop_t type
,
722 const char *domain
, uint64_t rid
, uint64_t *valp
)
730 if (!dmu_objset_userspace_present(zfsvfs
->z_os
))
733 obj
= zfs_userquota_prop_to_obj(zfsvfs
, type
);
737 err
= id_to_fuidstr(zfsvfs
, domain
, rid
, buf
, B_FALSE
);
741 err
= zap_lookup(zfsvfs
->z_os
, obj
, buf
, 8, 1, valp
);
748 zfs_set_userquota(zfsvfs_t
*zfsvfs
, zfs_userquota_prop_t type
,
749 const char *domain
, uint64_t rid
, uint64_t quota
)
755 boolean_t fuid_dirtied
;
757 if (type
!= ZFS_PROP_USERQUOTA
&& type
!= ZFS_PROP_GROUPQUOTA
)
760 if (zfsvfs
->z_version
< ZPL_VERSION_USERSPACE
)
763 objp
= (type
== ZFS_PROP_USERQUOTA
) ? &zfsvfs
->z_userquota_obj
:
764 &zfsvfs
->z_groupquota_obj
;
766 err
= id_to_fuidstr(zfsvfs
, domain
, rid
, buf
, B_TRUE
);
769 fuid_dirtied
= zfsvfs
->z_fuid_dirty
;
771 tx
= dmu_tx_create(zfsvfs
->z_os
);
772 dmu_tx_hold_zap(tx
, *objp
? *objp
: DMU_NEW_OBJECT
, B_TRUE
, NULL
);
774 dmu_tx_hold_zap(tx
, MASTER_NODE_OBJ
, B_TRUE
,
775 zfs_userquota_prop_prefixes
[type
]);
778 zfs_fuid_txhold(zfsvfs
, tx
);
779 err
= dmu_tx_assign(tx
, TXG_WAIT
);
785 mutex_enter(&zfsvfs
->z_lock
);
787 *objp
= zap_create(zfsvfs
->z_os
, DMU_OT_USERGROUP_QUOTA
,
789 VERIFY(0 == zap_add(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
790 zfs_userquota_prop_prefixes
[type
], 8, 1, objp
, tx
));
792 mutex_exit(&zfsvfs
->z_lock
);
795 err
= zap_remove(zfsvfs
->z_os
, *objp
, buf
, tx
);
799 err
= zap_update(zfsvfs
->z_os
, *objp
, buf
, 8, 1, "a
, tx
);
803 zfs_fuid_sync(zfsvfs
, tx
);
809 zfs_fuid_overquota(zfsvfs_t
*zfsvfs
, boolean_t isgroup
, uint64_t fuid
)
812 uint64_t used
, quota
, usedobj
, quotaobj
;
815 usedobj
= isgroup
? DMU_GROUPUSED_OBJECT
: DMU_USERUSED_OBJECT
;
816 quotaobj
= isgroup
? zfsvfs
->z_groupquota_obj
: zfsvfs
->z_userquota_obj
;
818 if (quotaobj
== 0 || zfsvfs
->z_replay
)
821 (void) sprintf(buf
, "%llx", (longlong_t
)fuid
);
822 err
= zap_lookup(zfsvfs
->z_os
, quotaobj
, buf
, 8, 1, "a
);
826 err
= zap_lookup(zfsvfs
->z_os
, usedobj
, buf
, 8, 1, &used
);
829 return (used
>= quota
);
833 zfs_owner_overquota(zfsvfs_t
*zfsvfs
, znode_t
*zp
, boolean_t isgroup
)
838 quotaobj
= isgroup
? zfsvfs
->z_groupquota_obj
: zfsvfs
->z_userquota_obj
;
840 fuid
= isgroup
? zp
->z_gid
: zp
->z_uid
;
842 if (quotaobj
== 0 || zfsvfs
->z_replay
)
845 return (zfs_fuid_overquota(zfsvfs
, isgroup
, fuid
));
849 zfsvfs_create(const char *osname
, zfsvfs_t
**zfvp
)
857 zfsvfs
= kmem_zalloc(sizeof (zfsvfs_t
), KM_SLEEP
);
860 * We claim to always be readonly so we can open snapshots;
861 * other ZPL code will prevent us from writing to snapshots.
863 error
= dmu_objset_own(osname
, DMU_OST_ZFS
, B_TRUE
, zfsvfs
, &os
);
865 kmem_free(zfsvfs
, sizeof (zfsvfs_t
));
870 * Initialize the zfs-specific filesystem structure.
871 * Should probably make this a kmem cache, shuffle fields,
872 * and just bzero up to z_hold_mtx[].
874 zfsvfs
->z_vfs
= NULL
;
875 zfsvfs
->z_parent
= zfsvfs
;
876 zfsvfs
->z_max_blksz
= SPA_MAXBLOCKSIZE
;
877 zfsvfs
->z_show_ctldir
= ZFS_SNAPDIR_VISIBLE
;
880 error
= zfs_get_zplprop(os
, ZFS_PROP_VERSION
, &zfsvfs
->z_version
);
883 } else if (zfsvfs
->z_version
>
884 zfs_zpl_version_map(spa_version(dmu_objset_spa(os
)))) {
885 (void) printf("Can't mount a version %lld file system "
886 "on a version %lld pool\n. Pool must be upgraded to mount "
887 "this file system.", (u_longlong_t
)zfsvfs
->z_version
,
888 (u_longlong_t
)spa_version(dmu_objset_spa(os
)));
892 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_NORMALIZE
, &zval
)) != 0)
894 zfsvfs
->z_norm
= (int)zval
;
896 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_UTF8ONLY
, &zval
)) != 0)
898 zfsvfs
->z_utf8
= (zval
!= 0);
900 if ((error
= zfs_get_zplprop(os
, ZFS_PROP_CASE
, &zval
)) != 0)
902 zfsvfs
->z_case
= (uint_t
)zval
;
905 * Fold case on file systems that are always or sometimes case
908 if (zfsvfs
->z_case
== ZFS_CASE_INSENSITIVE
||
909 zfsvfs
->z_case
== ZFS_CASE_MIXED
)
910 zfsvfs
->z_norm
|= U8_TEXTPREP_TOUPPER
;
912 zfsvfs
->z_use_fuids
= USE_FUIDS(zfsvfs
->z_version
, zfsvfs
->z_os
);
913 zfsvfs
->z_use_sa
= USE_SA(zfsvfs
->z_version
, zfsvfs
->z_os
);
915 if (zfsvfs
->z_use_sa
) {
916 /* should either have both of these objects or none */
917 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
, 8, 1,
923 * Pre SA versions file systems should never touch
924 * either the attribute registration or layout objects.
929 error
= sa_setup(os
, sa_obj
, zfs_attr_table
, ZPL_END
,
930 &zfsvfs
->z_attr_table
);
934 if (zfsvfs
->z_version
>= ZPL_VERSION_SA
)
935 sa_register_update_callback(os
, zfs_sa_upgrade
);
937 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_ROOT_OBJ
, 8, 1,
941 ASSERT(zfsvfs
->z_root
!= 0);
943 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_UNLINKED_SET
, 8, 1,
944 &zfsvfs
->z_unlinkedobj
);
948 error
= zap_lookup(os
, MASTER_NODE_OBJ
,
949 zfs_userquota_prop_prefixes
[ZFS_PROP_USERQUOTA
],
950 8, 1, &zfsvfs
->z_userquota_obj
);
951 if (error
&& error
!= ENOENT
)
954 error
= zap_lookup(os
, MASTER_NODE_OBJ
,
955 zfs_userquota_prop_prefixes
[ZFS_PROP_GROUPQUOTA
],
956 8, 1, &zfsvfs
->z_groupquota_obj
);
957 if (error
&& error
!= ENOENT
)
960 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_FUID_TABLES
, 8, 1,
961 &zfsvfs
->z_fuid_obj
);
962 if (error
&& error
!= ENOENT
)
965 error
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_SHARES_DIR
, 8, 1,
966 &zfsvfs
->z_shares_dir
);
967 if (error
&& error
!= ENOENT
)
970 mutex_init(&zfsvfs
->z_znodes_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
971 mutex_init(&zfsvfs
->z_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
972 list_create(&zfsvfs
->z_all_znodes
, sizeof (znode_t
),
973 offsetof(znode_t
, z_link_node
));
974 rrw_init(&zfsvfs
->z_teardown_lock
);
975 rw_init(&zfsvfs
->z_teardown_inactive_lock
, NULL
, RW_DEFAULT
, NULL
);
976 rw_init(&zfsvfs
->z_fuid_lock
, NULL
, RW_DEFAULT
, NULL
);
977 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
978 mutex_init(&zfsvfs
->z_hold_mtx
[i
], NULL
, MUTEX_DEFAULT
, NULL
);
984 dmu_objset_disown(os
, zfsvfs
);
986 kmem_free(zfsvfs
, sizeof (zfsvfs_t
));
991 zfsvfs_setup(zfsvfs_t
*zfsvfs
, boolean_t mounting
)
995 error
= zfs_register_callbacks(zfsvfs
->z_vfs
);
1000 * Set the objset user_ptr to track its zfsvfs.
1002 mutex_enter(&zfsvfs
->z_os
->os_user_ptr_lock
);
1003 dmu_objset_set_user(zfsvfs
->z_os
, zfsvfs
);
1004 mutex_exit(&zfsvfs
->z_os
->os_user_ptr_lock
);
1006 zfsvfs
->z_log
= zil_open(zfsvfs
->z_os
, zfs_get_data
);
1009 * If we are not mounting (ie: online recv), then we don't
1010 * have to worry about replaying the log as we blocked all
1011 * operations out since we closed the ZIL.
1017 * During replay we remove the read only flag to
1018 * allow replays to succeed.
1020 readonly
= zfsvfs
->z_vfs
->vfs_flag
& VFS_RDONLY
;
1022 zfsvfs
->z_vfs
->vfs_flag
&= ~VFS_RDONLY
;
1024 zfs_unlinked_drain(zfsvfs
);
1027 * Parse and replay the intent log.
1029 * Because of ziltest, this must be done after
1030 * zfs_unlinked_drain(). (Further note: ziltest
1031 * doesn't use readonly mounts, where
1032 * zfs_unlinked_drain() isn't called.) This is because
1033 * ziltest causes spa_sync() to think it's committed,
1034 * but actually it is not, so the intent log contains
1035 * many txg's worth of changes.
1037 * In particular, if object N is in the unlinked set in
1038 * the last txg to actually sync, then it could be
1039 * actually freed in a later txg and then reallocated
1040 * in a yet later txg. This would write a "create
1041 * object N" record to the intent log. Normally, this
1042 * would be fine because the spa_sync() would have
1043 * written out the fact that object N is free, before
1044 * we could write the "create object N" intent log
1047 * But when we are in ziltest mode, we advance the "open
1048 * txg" without actually spa_sync()-ing the changes to
1049 * disk. So we would see that object N is still
1050 * allocated and in the unlinked set, and there is an
1051 * intent log record saying to allocate it.
1053 if (spa_writeable(dmu_objset_spa(zfsvfs
->z_os
))) {
1054 if (zil_replay_disable
) {
1055 zil_destroy(zfsvfs
->z_log
, B_FALSE
);
1057 zfsvfs
->z_replay
= B_TRUE
;
1058 zil_replay(zfsvfs
->z_os
, zfsvfs
,
1060 zfsvfs
->z_replay
= B_FALSE
;
1063 zfsvfs
->z_vfs
->vfs_flag
|= readonly
; /* restore readonly bit */
1070 zfsvfs_free(zfsvfs_t
*zfsvfs
)
1073 extern krwlock_t zfsvfs_lock
; /* in zfs_znode.c */
1076 * This is a barrier to prevent the filesystem from going away in
1077 * zfs_znode_move() until we can safely ensure that the filesystem is
1078 * not unmounted. We consider the filesystem valid before the barrier
1079 * and invalid after the barrier.
1081 rw_enter(&zfsvfs_lock
, RW_READER
);
1082 rw_exit(&zfsvfs_lock
);
1084 zfs_fuid_destroy(zfsvfs
);
1086 mutex_destroy(&zfsvfs
->z_znodes_lock
);
1087 mutex_destroy(&zfsvfs
->z_lock
);
1088 list_destroy(&zfsvfs
->z_all_znodes
);
1089 rrw_destroy(&zfsvfs
->z_teardown_lock
);
1090 rw_destroy(&zfsvfs
->z_teardown_inactive_lock
);
1091 rw_destroy(&zfsvfs
->z_fuid_lock
);
1092 for (i
= 0; i
!= ZFS_OBJ_MTX_SZ
; i
++)
1093 mutex_destroy(&zfsvfs
->z_hold_mtx
[i
]);
1094 kmem_free(zfsvfs
, sizeof (zfsvfs_t
));
1098 zfs_set_fuid_feature(zfsvfs_t
*zfsvfs
)
1100 zfsvfs
->z_use_fuids
= USE_FUIDS(zfsvfs
->z_version
, zfsvfs
->z_os
);
1101 if (zfsvfs
->z_vfs
) {
1102 if (zfsvfs
->z_use_fuids
) {
1103 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_XVATTR
);
1104 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_SYSATTR_VIEWS
);
1105 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_ACEMASKONACCESS
);
1106 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_ACLONCREATE
);
1107 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_ACCESS_FILTER
);
1108 vfs_set_feature(zfsvfs
->z_vfs
, VFSFT_REPARSE
);
1110 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_XVATTR
);
1111 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_SYSATTR_VIEWS
);
1112 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_ACEMASKONACCESS
);
1113 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_ACLONCREATE
);
1114 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_ACCESS_FILTER
);
1115 vfs_clear_feature(zfsvfs
->z_vfs
, VFSFT_REPARSE
);
1118 zfsvfs
->z_use_sa
= USE_SA(zfsvfs
->z_version
, zfsvfs
->z_os
);
1122 zfs_domount(vfs_t
*vfsp
, char *osname
)
1125 uint64_t recordsize
, fsid_guid
;
1132 error
= zfsvfs_create(osname
, &zfsvfs
);
1135 zfsvfs
->z_vfs
= vfsp
;
1137 /* Initialize the generic filesystem structure. */
1138 vfsp
->vfs_bcount
= 0;
1139 vfsp
->vfs_data
= NULL
;
1141 if (zfs_create_unique_device(&mount_dev
) == -1) {
1145 ASSERT(vfs_devismounted(mount_dev
) == 0);
1147 if (error
= dsl_prop_get_integer(osname
, "recordsize", &recordsize
,
1151 vfsp
->vfs_dev
= mount_dev
;
1152 vfsp
->vfs_fstype
= zfsfstype
;
1153 vfsp
->vfs_bsize
= recordsize
;
1154 vfsp
->vfs_flag
|= VFS_NOTRUNC
;
1155 vfsp
->vfs_data
= zfsvfs
;
1158 * The fsid is 64 bits, composed of an 8-bit fs type, which
1159 * separates our fsid from any other filesystem types, and a
1160 * 56-bit objset unique ID. The objset unique ID is unique to
1161 * all objsets open on this system, provided by unique_create().
1162 * The 8-bit fs type must be put in the low bits of fsid[1]
1163 * because that's where other Solaris filesystems put it.
1165 fsid_guid
= dmu_objset_fsid_guid(zfsvfs
->z_os
);
1166 ASSERT((fsid_guid
& ~((1ULL<<56)-1)) == 0);
1167 vfsp
->vfs_fsid
.val
[0] = fsid_guid
;
1168 vfsp
->vfs_fsid
.val
[1] = ((fsid_guid
>>32) << 8) |
1172 * Set features for file system.
1174 zfs_set_fuid_feature(zfsvfs
);
1175 if (zfsvfs
->z_case
== ZFS_CASE_INSENSITIVE
) {
1176 vfs_set_feature(vfsp
, VFSFT_DIRENTFLAGS
);
1177 vfs_set_feature(vfsp
, VFSFT_CASEINSENSITIVE
);
1178 vfs_set_feature(vfsp
, VFSFT_NOCASESENSITIVE
);
1179 } else if (zfsvfs
->z_case
== ZFS_CASE_MIXED
) {
1180 vfs_set_feature(vfsp
, VFSFT_DIRENTFLAGS
);
1181 vfs_set_feature(vfsp
, VFSFT_CASEINSENSITIVE
);
1183 vfs_set_feature(vfsp
, VFSFT_ZEROCOPY_SUPPORTED
);
1185 if (dmu_objset_is_snapshot(zfsvfs
->z_os
)) {
1188 atime_changed_cb(zfsvfs
, B_FALSE
);
1189 readonly_changed_cb(zfsvfs
, B_TRUE
);
1190 if (error
= dsl_prop_get_integer(osname
, "xattr", &pval
, NULL
))
1192 xattr_changed_cb(zfsvfs
, pval
);
1193 zfsvfs
->z_issnap
= B_TRUE
;
1194 zfsvfs
->z_os
->os_sync
= ZFS_SYNC_DISABLED
;
1196 mutex_enter(&zfsvfs
->z_os
->os_user_ptr_lock
);
1197 dmu_objset_set_user(zfsvfs
->z_os
, zfsvfs
);
1198 mutex_exit(&zfsvfs
->z_os
->os_user_ptr_lock
);
1200 error
= zfsvfs_setup(zfsvfs
, B_TRUE
);
1203 if (!zfsvfs
->z_issnap
)
1204 zfsctl_create(zfsvfs
);
1207 dmu_objset_disown(zfsvfs
->z_os
, zfsvfs
);
1208 zfsvfs_free(zfsvfs
);
1210 atomic_add_32(&zfs_active_fs_count
, 1);
1217 zfs_unregister_callbacks(zfsvfs_t
*zfsvfs
)
1219 objset_t
*os
= zfsvfs
->z_os
;
1220 struct dsl_dataset
*ds
;
1223 * Unregister properties.
1225 if (!dmu_objset_is_snapshot(os
)) {
1226 ds
= dmu_objset_ds(os
);
1227 VERIFY(dsl_prop_unregister(ds
, "atime", atime_changed_cb
,
1230 VERIFY(dsl_prop_unregister(ds
, "xattr", xattr_changed_cb
,
1233 VERIFY(dsl_prop_unregister(ds
, "recordsize", blksz_changed_cb
,
1236 VERIFY(dsl_prop_unregister(ds
, "readonly", readonly_changed_cb
,
1239 VERIFY(dsl_prop_unregister(ds
, "devices", devices_changed_cb
,
1242 VERIFY(dsl_prop_unregister(ds
, "setuid", setuid_changed_cb
,
1245 VERIFY(dsl_prop_unregister(ds
, "exec", exec_changed_cb
,
1248 VERIFY(dsl_prop_unregister(ds
, "snapdir", snapdir_changed_cb
,
1251 VERIFY(dsl_prop_unregister(ds
, "aclmode", acl_mode_changed_cb
,
1254 VERIFY(dsl_prop_unregister(ds
, "aclinherit",
1255 acl_inherit_changed_cb
, zfsvfs
) == 0);
1257 VERIFY(dsl_prop_unregister(ds
, "vscan",
1258 vscan_changed_cb
, zfsvfs
) == 0);
1263 * Convert a decimal digit string to a uint64_t integer.
1266 str_to_uint64(char *str
, uint64_t *objnum
)
1271 if (*str
< '0' || *str
> '9')
1274 num
= num
*10 + *str
++ - '0';
1282 * The boot path passed from the boot loader is in the form of
1283 * "rootpool-name/root-filesystem-object-number'. Convert this
1284 * string to a dataset name: "rootpool-name/root-filesystem-name".
1287 zfs_parse_bootfs(char *bpath
, char *outpath
)
1293 if (*bpath
== 0 || *bpath
== '/')
1296 (void) strcpy(outpath
, bpath
);
1298 slashp
= strchr(bpath
, '/');
1300 /* if no '/', just return the pool name */
1301 if (slashp
== NULL
) {
1305 /* if not a number, just return the root dataset name */
1306 if (str_to_uint64(slashp
+1, &objnum
)) {
1311 error
= dsl_dsobj_to_dsname(bpath
, objnum
, outpath
);
1318 * zfs_check_global_label:
1319 * Check that the hex label string is appropriate for the dataset
1320 * being mounted into the global_zone proper.
1322 * Return an error if the hex label string is not default or
1323 * admin_low/admin_high. For admin_low labels, the corresponding
1324 * dataset must be readonly.
1327 zfs_check_global_label(const char *dsname
, const char *hexsl
)
1329 if (strcasecmp(hexsl
, ZFS_MLSLABEL_DEFAULT
) == 0)
1331 if (strcasecmp(hexsl
, ADMIN_HIGH
) == 0)
1333 if (strcasecmp(hexsl
, ADMIN_LOW
) == 0) {
1334 /* must be readonly */
1337 if (dsl_prop_get_integer(dsname
,
1338 zfs_prop_to_name(ZFS_PROP_READONLY
), &rdonly
, NULL
))
1340 return (rdonly
? 0 : EACCES
);
1346 * zfs_mount_label_policy:
1347 * Determine whether the mount is allowed according to MAC check.
1348 * by comparing (where appropriate) label of the dataset against
1349 * the label of the zone being mounted into. If the dataset has
1350 * no label, create one.
1353 * 0 : access allowed
1354 * >0 : error code, such as EACCES
1357 zfs_mount_label_policy(vfs_t
*vfsp
, char *osname
)
1360 zone_t
*mntzone
= NULL
;
1361 ts_label_t
*mnt_tsl
;
1364 char ds_hexsl
[MAXNAMELEN
];
1366 retv
= EACCES
; /* assume the worst */
1369 * Start by getting the dataset label if it exists.
1371 error
= dsl_prop_get(osname
, zfs_prop_to_name(ZFS_PROP_MLSLABEL
),
1372 1, sizeof (ds_hexsl
), &ds_hexsl
, NULL
);
1377 * If labeling is NOT enabled, then disallow the mount of datasets
1378 * which have a non-default label already. No other label checks
1381 if (!is_system_labeled()) {
1382 if (strcasecmp(ds_hexsl
, ZFS_MLSLABEL_DEFAULT
) == 0)
1388 * Get the label of the mountpoint. If mounting into the global
1389 * zone (i.e. mountpoint is not within an active zone and the
1390 * zoned property is off), the label must be default or
1391 * admin_low/admin_high only; no other checks are needed.
1393 mntzone
= zone_find_by_any_path(refstr_value(vfsp
->vfs_mntpt
), B_FALSE
);
1394 if (mntzone
->zone_id
== GLOBAL_ZONEID
) {
1399 if (dsl_prop_get_integer(osname
,
1400 zfs_prop_to_name(ZFS_PROP_ZONED
), &zoned
, NULL
))
1403 return (zfs_check_global_label(osname
, ds_hexsl
));
1406 * This is the case of a zone dataset being mounted
1407 * initially, before the zone has been fully created;
1408 * allow this mount into global zone.
1413 mnt_tsl
= mntzone
->zone_slabel
;
1414 ASSERT(mnt_tsl
!= NULL
);
1415 label_hold(mnt_tsl
);
1416 mnt_sl
= label2bslabel(mnt_tsl
);
1418 if (strcasecmp(ds_hexsl
, ZFS_MLSLABEL_DEFAULT
) == 0) {
1420 * The dataset doesn't have a real label, so fabricate one.
1424 if (l_to_str_internal(mnt_sl
, &str
) == 0 &&
1425 dsl_prop_set(osname
, zfs_prop_to_name(ZFS_PROP_MLSLABEL
),
1426 ZPROP_SRC_LOCAL
, 1, strlen(str
) + 1, str
) == 0)
1429 kmem_free(str
, strlen(str
) + 1);
1430 } else if (hexstr_to_label(ds_hexsl
, &ds_sl
) == 0) {
1432 * Now compare labels to complete the MAC check. If the
1433 * labels are equal then allow access. If the mountpoint
1434 * label dominates the dataset label, allow readonly access.
1435 * Otherwise, access is denied.
1437 if (blequal(mnt_sl
, &ds_sl
))
1439 else if (bldominates(mnt_sl
, &ds_sl
)) {
1440 vfs_setmntopt(vfsp
, MNTOPT_RO
, NULL
, 0);
1445 label_rele(mnt_tsl
);
1451 zfs_mountroot(vfs_t
*vfsp
, enum whymountroot why
)
1454 static int zfsrootdone
= 0;
1455 zfsvfs_t
*zfsvfs
= NULL
;
1464 * The filesystem that we mount as root is defined in the
1465 * boot property "zfs-bootfs" with a format of
1466 * "poolname/root-dataset-objnum".
1468 if (why
== ROOT_INIT
) {
1472 * the process of doing a spa_load will require the
1473 * clock to be set before we could (for example) do
1474 * something better by looking at the timestamp on
1475 * an uberblock, so just set it to -1.
1479 if ((zfs_bootfs
= spa_get_bootprop("zfs-bootfs")) == NULL
) {
1480 cmn_err(CE_NOTE
, "spa_get_bootfs: can not get "
1484 zfs_devid
= spa_get_bootprop("diskdevid");
1485 error
= spa_import_rootpool(rootfs
.bo_name
, zfs_devid
);
1487 spa_free_bootprop(zfs_devid
);
1489 spa_free_bootprop(zfs_bootfs
);
1490 cmn_err(CE_NOTE
, "spa_import_rootpool: error %d",
1494 if (error
= zfs_parse_bootfs(zfs_bootfs
, rootfs
.bo_name
)) {
1495 spa_free_bootprop(zfs_bootfs
);
1496 cmn_err(CE_NOTE
, "zfs_parse_bootfs: error %d",
1501 spa_free_bootprop(zfs_bootfs
);
1503 if (error
= vfs_lock(vfsp
))
1506 if (error
= zfs_domount(vfsp
, rootfs
.bo_name
)) {
1507 cmn_err(CE_NOTE
, "zfs_domount: error %d", error
);
1511 zfsvfs
= (zfsvfs_t
*)vfsp
->vfs_data
;
1513 if (error
= zfs_zget(zfsvfs
, zfsvfs
->z_root
, &zp
)) {
1514 cmn_err(CE_NOTE
, "zfs_zget: error %d", error
);
1519 mutex_enter(&vp
->v_lock
);
1520 vp
->v_flag
|= VROOT
;
1521 mutex_exit(&vp
->v_lock
);
1525 * Leave rootvp held. The root file system is never unmounted.
1528 vfs_add((struct vnode
*)0, vfsp
,
1529 (vfsp
->vfs_flag
& VFS_RDONLY
) ? MS_RDONLY
: 0);
1533 } else if (why
== ROOT_REMOUNT
) {
1534 readonly_changed_cb(vfsp
->vfs_data
, B_FALSE
);
1535 vfsp
->vfs_flag
|= VFS_REMOUNT
;
1537 /* refresh mount options */
1538 zfs_unregister_callbacks(vfsp
->vfs_data
);
1539 return (zfs_register_callbacks(vfsp
));
1541 } else if (why
== ROOT_UNMOUNT
) {
1542 zfs_unregister_callbacks((zfsvfs_t
*)vfsp
->vfs_data
);
1543 (void) zfs_sync(vfsp
, 0, 0);
1548 * if "why" is equal to anything else other than ROOT_INIT,
1549 * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it.
1556 zfs_mount(vfs_t
*vfsp
, vnode_t
*mvp
, struct mounta
*uap
, cred_t
*cr
)
1561 uio_seg_t fromspace
= (uap
->flags
& MS_SYSSPACE
) ?
1562 UIO_SYSSPACE
: UIO_USERSPACE
;
1565 if (mvp
->v_type
!= VDIR
)
1568 mutex_enter(&mvp
->v_lock
);
1569 if ((uap
->flags
& MS_REMOUNT
) == 0 &&
1570 (uap
->flags
& MS_OVERLAY
) == 0 &&
1571 (mvp
->v_count
!= 1 || (mvp
->v_flag
& VROOT
))) {
1572 mutex_exit(&mvp
->v_lock
);
1575 mutex_exit(&mvp
->v_lock
);
1578 * ZFS does not support passing unparsed data in via MS_DATA.
1579 * Users should use the MS_OPTIONSTR interface; this means
1580 * that all option parsing is already done and the options struct
1581 * can be interrogated.
1583 if ((uap
->flags
& MS_DATA
) && uap
->datalen
> 0)
1587 * Get the objset name (the "special" mount argument).
1589 if (error
= pn_get(uap
->spec
, fromspace
, &spn
))
1592 osname
= spn
.pn_path
;
1595 * Check for mount privilege?
1597 * If we don't have privilege then see if
1598 * we have local permission to allow it
1600 error
= secpolicy_fs_mount(cr
, mvp
, vfsp
);
1602 if (dsl_deleg_access(osname
, ZFS_DELEG_PERM_MOUNT
, cr
) == 0) {
1606 * Make sure user is the owner of the mount point
1607 * or has sufficient privileges.
1610 vattr
.va_mask
= AT_UID
;
1612 if (VOP_GETATTR(mvp
, &vattr
, 0, cr
, NULL
)) {
1616 if (secpolicy_vnode_owner(cr
, vattr
.va_uid
) != 0 &&
1617 VOP_ACCESS(mvp
, VWRITE
, 0, cr
, NULL
) != 0) {
1620 secpolicy_fs_mount_clearopts(cr
, vfsp
);
1627 * Refuse to mount a filesystem if we are in a local zone and the
1628 * dataset is not visible.
1630 if (!INGLOBALZONE(curproc
) &&
1631 (!zone_dataset_visible(osname
, &canwrite
) || !canwrite
)) {
1636 error
= zfs_mount_label_policy(vfsp
, osname
);
1641 * When doing a remount, we simply refresh our temporary properties
1642 * according to those options set in the current VFS options.
1644 if (uap
->flags
& MS_REMOUNT
) {
1645 /* refresh mount options */
1646 zfs_unregister_callbacks(vfsp
->vfs_data
);
1647 error
= zfs_register_callbacks(vfsp
);
1651 error
= zfs_domount(vfsp
, osname
);
1654 * Add an extra VFS_HOLD on our parent vfs so that it can't
1655 * disappear due to a forced unmount.
1657 if (error
== 0 && ((zfsvfs_t
*)vfsp
->vfs_data
)->z_issnap
)
1658 VFS_HOLD(mvp
->v_vfsp
);
1666 zfs_statvfs(vfs_t
*vfsp
, struct statvfs64
*statp
)
1668 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
1670 uint64_t refdbytes
, availbytes
, usedobjs
, availobjs
;
1674 dmu_objset_space(zfsvfs
->z_os
,
1675 &refdbytes
, &availbytes
, &usedobjs
, &availobjs
);
1678 * The underlying storage pool actually uses multiple block sizes.
1679 * We report the fragsize as the smallest block size we support,
1680 * and we report our blocksize as the filesystem's maximum blocksize.
1682 statp
->f_frsize
= 1UL << SPA_MINBLOCKSHIFT
;
1683 statp
->f_bsize
= zfsvfs
->z_max_blksz
;
1686 * The following report "total" blocks of various kinds in the
1687 * file system, but reported in terms of f_frsize - the
1691 statp
->f_blocks
= (refdbytes
+ availbytes
) >> SPA_MINBLOCKSHIFT
;
1692 statp
->f_bfree
= availbytes
>> SPA_MINBLOCKSHIFT
;
1693 statp
->f_bavail
= statp
->f_bfree
; /* no root reservation */
1696 * statvfs() should really be called statufs(), because it assumes
1697 * static metadata. ZFS doesn't preallocate files, so the best
1698 * we can do is report the max that could possibly fit in f_files,
1699 * and that minus the number actually used in f_ffree.
1700 * For f_ffree, report the smaller of the number of object available
1701 * and the number of blocks (each object will take at least a block).
1703 statp
->f_ffree
= MIN(availobjs
, statp
->f_bfree
);
1704 statp
->f_favail
= statp
->f_ffree
; /* no "root reservation" */
1705 statp
->f_files
= statp
->f_ffree
+ usedobjs
;
1707 (void) cmpldev(&d32
, vfsp
->vfs_dev
);
1708 statp
->f_fsid
= d32
;
1711 * We're a zfs filesystem.
1713 (void) strcpy(statp
->f_basetype
, vfssw
[vfsp
->vfs_fstype
].vsw_name
);
1715 statp
->f_flag
= vf_to_stf(vfsp
->vfs_flag
);
1717 statp
->f_namemax
= ZFS_MAXNAMELEN
;
1720 * We have all of 32 characters to stuff a string here.
1721 * Is there anything useful we could/should provide?
1723 bzero(statp
->f_fstr
, sizeof (statp
->f_fstr
));
1730 zfs_root(vfs_t
*vfsp
, vnode_t
**vpp
)
1732 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
1738 error
= zfs_zget(zfsvfs
, zfsvfs
->z_root
, &rootzp
);
1740 *vpp
= ZTOV(rootzp
);
1747 * Teardown the zfsvfs::z_os.
1749 * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
1750 * and 'z_teardown_inactive_lock' held.
1753 zfsvfs_teardown(zfsvfs_t
*zfsvfs
, boolean_t unmounting
)
1757 rrw_enter(&zfsvfs
->z_teardown_lock
, RW_WRITER
, FTAG
);
1761 * We purge the parent filesystem's vfsp as the parent
1762 * filesystem and all of its snapshots have their vnode's
1763 * v_vfsp set to the parent's filesystem's vfsp. Note,
1764 * 'z_parent' is self referential for non-snapshots.
1766 (void) dnlc_purge_vfsp(zfsvfs
->z_parent
->z_vfs
, 0);
1770 * Close the zil. NB: Can't close the zil while zfs_inactive
1771 * threads are blocked as zil_close can call zfs_inactive.
1773 if (zfsvfs
->z_log
) {
1774 zil_close(zfsvfs
->z_log
);
1775 zfsvfs
->z_log
= NULL
;
1778 rw_enter(&zfsvfs
->z_teardown_inactive_lock
, RW_WRITER
);
1781 * If we are not unmounting (ie: online recv) and someone already
1782 * unmounted this file system while we were doing the switcheroo,
1783 * or a reopen of z_os failed then just bail out now.
1785 if (!unmounting
&& (zfsvfs
->z_unmounted
|| zfsvfs
->z_os
== NULL
)) {
1786 rw_exit(&zfsvfs
->z_teardown_inactive_lock
);
1787 rrw_exit(&zfsvfs
->z_teardown_lock
, FTAG
);
1792 * At this point there are no vops active, and any new vops will
1793 * fail with EIO since we have z_teardown_lock for writer (only
1794 * relavent for forced unmount).
1796 * Release all holds on dbufs.
1798 mutex_enter(&zfsvfs
->z_znodes_lock
);
1799 for (zp
= list_head(&zfsvfs
->z_all_znodes
); zp
!= NULL
;
1800 zp
= list_next(&zfsvfs
->z_all_znodes
, zp
))
1802 ASSERT(ZTOV(zp
)->v_count
> 0);
1803 zfs_znode_dmu_fini(zp
);
1805 mutex_exit(&zfsvfs
->z_znodes_lock
);
1808 * If we are unmounting, set the unmounted flag and let new vops
1809 * unblock. zfs_inactive will have the unmounted behavior, and all
1810 * other vops will fail with EIO.
1813 zfsvfs
->z_unmounted
= B_TRUE
;
1814 rrw_exit(&zfsvfs
->z_teardown_lock
, FTAG
);
1815 rw_exit(&zfsvfs
->z_teardown_inactive_lock
);
1819 * z_os will be NULL if there was an error in attempting to reopen
1820 * zfsvfs, so just return as the properties had already been
1821 * unregistered and cached data had been evicted before.
1823 if (zfsvfs
->z_os
== NULL
)
1827 * Unregister properties.
1829 zfs_unregister_callbacks(zfsvfs
);
1834 if (dmu_objset_is_dirty_anywhere(zfsvfs
->z_os
))
1835 if (!(zfsvfs
->z_vfs
->vfs_flag
& VFS_RDONLY
))
1836 txg_wait_synced(dmu_objset_pool(zfsvfs
->z_os
), 0);
1837 (void) dmu_objset_evict_dbufs(zfsvfs
->z_os
);
1844 zfs_umount(vfs_t
*vfsp
, int fflag
, cred_t
*cr
)
1846 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
1850 ret
= secpolicy_fs_unmount(cr
, vfsp
);
1852 if (dsl_deleg_access((char *)refstr_value(vfsp
->vfs_resource
),
1853 ZFS_DELEG_PERM_MOUNT
, cr
))
1858 * We purge the parent filesystem's vfsp as the parent filesystem
1859 * and all of its snapshots have their vnode's v_vfsp set to the
1860 * parent's filesystem's vfsp. Note, 'z_parent' is self
1861 * referential for non-snapshots.
1863 (void) dnlc_purge_vfsp(zfsvfs
->z_parent
->z_vfs
, 0);
1866 * Unmount any snapshots mounted under .zfs before unmounting the
1869 if (zfsvfs
->z_ctldir
!= NULL
&&
1870 (ret
= zfsctl_umount_snapshots(vfsp
, fflag
, cr
)) != 0) {
1874 if (!(fflag
& MS_FORCE
)) {
1876 * Check the number of active vnodes in the file system.
1877 * Our count is maintained in the vfs structure, but the
1878 * number is off by 1 to indicate a hold on the vfs
1881 * The '.zfs' directory maintains a reference of its
1882 * own, and any active references underneath are
1883 * reflected in the vnode count.
1885 if (zfsvfs
->z_ctldir
== NULL
) {
1886 if (vfsp
->vfs_count
> 1)
1889 if (vfsp
->vfs_count
> 2 ||
1890 zfsvfs
->z_ctldir
->v_count
> 1)
1895 vfsp
->vfs_flag
|= VFS_UNMOUNTED
;
1897 VERIFY(zfsvfs_teardown(zfsvfs
, B_TRUE
) == 0);
1901 * z_os will be NULL if there was an error in
1902 * attempting to reopen zfsvfs.
1906 * Unset the objset user_ptr.
1908 mutex_enter(&os
->os_user_ptr_lock
);
1909 dmu_objset_set_user(os
, NULL
);
1910 mutex_exit(&os
->os_user_ptr_lock
);
1913 * Finally release the objset
1915 dmu_objset_disown(os
, zfsvfs
);
1919 * We can now safely destroy the '.zfs' directory node.
1921 if (zfsvfs
->z_ctldir
!= NULL
)
1922 zfsctl_destroy(zfsvfs
);
1928 zfs_vget(vfs_t
*vfsp
, vnode_t
**vpp
, fid_t
*fidp
)
1930 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
1932 uint64_t object
= 0;
1933 uint64_t fid_gen
= 0;
1942 if (fidp
->fid_len
== LONG_FID_LEN
) {
1943 zfid_long_t
*zlfid
= (zfid_long_t
*)fidp
;
1944 uint64_t objsetid
= 0;
1945 uint64_t setgen
= 0;
1947 for (i
= 0; i
< sizeof (zlfid
->zf_setid
); i
++)
1948 objsetid
|= ((uint64_t)zlfid
->zf_setid
[i
]) << (8 * i
);
1950 for (i
= 0; i
< sizeof (zlfid
->zf_setgen
); i
++)
1951 setgen
|= ((uint64_t)zlfid
->zf_setgen
[i
]) << (8 * i
);
1955 err
= zfsctl_lookup_objset(vfsp
, objsetid
, &zfsvfs
);
1961 if (fidp
->fid_len
== SHORT_FID_LEN
|| fidp
->fid_len
== LONG_FID_LEN
) {
1962 zfid_short_t
*zfid
= (zfid_short_t
*)fidp
;
1964 for (i
= 0; i
< sizeof (zfid
->zf_object
); i
++)
1965 object
|= ((uint64_t)zfid
->zf_object
[i
]) << (8 * i
);
1967 for (i
= 0; i
< sizeof (zfid
->zf_gen
); i
++)
1968 fid_gen
|= ((uint64_t)zfid
->zf_gen
[i
]) << (8 * i
);
1974 /* A zero fid_gen means we are in the .zfs control directories */
1976 (object
== ZFSCTL_INO_ROOT
|| object
== ZFSCTL_INO_SNAPDIR
)) {
1977 *vpp
= zfsvfs
->z_ctldir
;
1978 ASSERT(*vpp
!= NULL
);
1979 if (object
== ZFSCTL_INO_SNAPDIR
) {
1980 VERIFY(zfsctl_root_lookup(*vpp
, "snapshot", vpp
, NULL
,
1981 0, NULL
, NULL
, NULL
, NULL
, NULL
) == 0);
1989 gen_mask
= -1ULL >> (64 - 8 * i
);
1991 dprintf("getting %llu [%u mask %llx]\n", object
, fid_gen
, gen_mask
);
1992 if (err
= zfs_zget(zfsvfs
, object
, &zp
)) {
1996 (void) sa_lookup(zp
->z_sa_hdl
, SA_ZPL_GEN(zfsvfs
), &zp_gen
,
1998 zp_gen
= zp_gen
& gen_mask
;
2001 if (zp
->z_unlinked
|| zp_gen
!= fid_gen
) {
2002 dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen
, fid_gen
);
2014 * Block out VOPs and close zfsvfs_t::z_os
2016 * Note, if successful, then we return with the 'z_teardown_lock' and
2017 * 'z_teardown_inactive_lock' write held.
2020 zfs_suspend_fs(zfsvfs_t
*zfsvfs
)
2024 if ((error
= zfsvfs_teardown(zfsvfs
, B_FALSE
)) != 0)
2026 dmu_objset_disown(zfsvfs
->z_os
, zfsvfs
);
2032 * Reopen zfsvfs_t::z_os and release VOPs.
2035 zfs_resume_fs(zfsvfs_t
*zfsvfs
, const char *osname
)
2039 ASSERT(RRW_WRITE_HELD(&zfsvfs
->z_teardown_lock
));
2040 ASSERT(RW_WRITE_HELD(&zfsvfs
->z_teardown_inactive_lock
));
2042 err
= dmu_objset_own(osname
, DMU_OST_ZFS
, B_FALSE
, zfsvfs
,
2045 zfsvfs
->z_os
= NULL
;
2048 uint64_t sa_obj
= 0;
2051 * Make sure version hasn't changed
2054 err
= zfs_get_zplprop(zfsvfs
->z_os
, ZFS_PROP_VERSION
,
2055 &zfsvfs
->z_version
);
2060 err
= zap_lookup(zfsvfs
->z_os
, MASTER_NODE_OBJ
,
2061 ZFS_SA_ATTRS
, 8, 1, &sa_obj
);
2063 if (err
&& zfsvfs
->z_version
>= ZPL_VERSION_SA
)
2066 if ((err
= sa_setup(zfsvfs
->z_os
, sa_obj
,
2067 zfs_attr_table
, ZPL_END
, &zfsvfs
->z_attr_table
)) != 0)
2070 if (zfsvfs
->z_version
>= ZPL_VERSION_SA
)
2071 sa_register_update_callback(zfsvfs
->z_os
,
2074 VERIFY(zfsvfs_setup(zfsvfs
, B_FALSE
) == 0);
2076 zfs_set_fuid_feature(zfsvfs
);
2079 * Attempt to re-establish all the active znodes with
2080 * their dbufs. If a zfs_rezget() fails, then we'll let
2081 * any potential callers discover that via ZFS_ENTER_VERIFY_VP
2082 * when they try to use their znode.
2084 mutex_enter(&zfsvfs
->z_znodes_lock
);
2085 for (zp
= list_head(&zfsvfs
->z_all_znodes
); zp
;
2086 zp
= list_next(&zfsvfs
->z_all_znodes
, zp
)) {
2087 (void) zfs_rezget(zp
);
2089 mutex_exit(&zfsvfs
->z_znodes_lock
);
2093 /* release the VOPs */
2094 rw_exit(&zfsvfs
->z_teardown_inactive_lock
);
2095 rrw_exit(&zfsvfs
->z_teardown_lock
, FTAG
);
2099 * Since we couldn't reopen zfsvfs::z_os, or
2100 * setup the sa framework force unmount this file system.
2102 if (vn_vfswlock(zfsvfs
->z_vfs
->vfs_vnodecovered
) == 0)
2103 (void) dounmount(zfsvfs
->z_vfs
, MS_FORCE
, CRED());
2109 zfs_freevfs(vfs_t
*vfsp
)
2111 zfsvfs_t
*zfsvfs
= vfsp
->vfs_data
;
2114 * If this is a snapshot, we have an extra VFS_HOLD on our parent
2115 * from zfs_mount(). Release it here. If we came through
2116 * zfs_mountroot() instead, we didn't grab an extra hold, so
2117 * skip the VFS_RELE for rootvfs.
2119 if (zfsvfs
->z_issnap
&& (vfsp
!= rootvfs
))
2120 VFS_RELE(zfsvfs
->z_parent
->z_vfs
);
2122 zfsvfs_free(zfsvfs
);
2124 atomic_add_32(&zfs_active_fs_count
, -1);
2128 * VFS_INIT() initialization. Note that there is no VFS_FINI(),
2129 * so we can't safely do any non-idempotent initialization here.
2130 * Leave that to zfs_init() and zfs_fini(), which are called
2131 * from the module's _init() and _fini() entry points.
2135 zfs_vfsinit(int fstype
, char *name
)
2142 * Setup vfsops and vnodeops tables.
2144 error
= vfs_setfsops(fstype
, zfs_vfsops_template
, &zfs_vfsops
);
2146 cmn_err(CE_WARN
, "zfs: bad vfs ops template");
2149 error
= zfs_create_op_tables();
2151 zfs_remove_op_tables();
2152 cmn_err(CE_WARN
, "zfs: bad vnode ops template");
2153 (void) vfs_freevfsops_by_type(zfsfstype
);
2157 mutex_init(&zfs_dev_mtx
, NULL
, MUTEX_DEFAULT
, NULL
);
2160 * Unique major number for all zfs mounts.
2161 * If we run out of 32-bit minors, we'll getudev() another major.
2163 zfs_major
= ddi_name_to_major(ZFS_DRIVER
);
2164 zfs_minor
= ZFS_MIN_MINOR
;
2173 * Initialize .zfs directory structures
2178 * Initialize znode cache, vnode ops, etc...
2182 dmu_objset_register_type(DMU_OST_ZFS
, zfs_space_delta_cb
);
2195 return (zfs_active_fs_count
!= 0);
2199 zfs_set_version(zfsvfs_t
*zfsvfs
, uint64_t newvers
)
2202 objset_t
*os
= zfsvfs
->z_os
;
2205 if (newvers
< ZPL_VERSION_INITIAL
|| newvers
> ZPL_VERSION
)
2208 if (newvers
< zfsvfs
->z_version
)
2211 if (zfs_spa_version_map(newvers
) >
2212 spa_version(dmu_objset_spa(zfsvfs
->z_os
)))
2215 tx
= dmu_tx_create(os
);
2216 dmu_tx_hold_zap(tx
, MASTER_NODE_OBJ
, B_FALSE
, ZPL_VERSION_STR
);
2217 if (newvers
>= ZPL_VERSION_SA
&& !zfsvfs
->z_use_sa
) {
2218 dmu_tx_hold_zap(tx
, MASTER_NODE_OBJ
, B_TRUE
,
2220 dmu_tx_hold_zap(tx
, DMU_NEW_OBJECT
, FALSE
, NULL
);
2222 error
= dmu_tx_assign(tx
, TXG_WAIT
);
2228 error
= zap_update(os
, MASTER_NODE_OBJ
, ZPL_VERSION_STR
,
2229 8, 1, &newvers
, tx
);
2236 if (newvers
>= ZPL_VERSION_SA
&& !zfsvfs
->z_use_sa
) {
2239 ASSERT3U(spa_version(dmu_objset_spa(zfsvfs
->z_os
)), >=,
2241 sa_obj
= zap_create(os
, DMU_OT_SA_MASTER_NODE
,
2242 DMU_OT_NONE
, 0, tx
);
2244 error
= zap_add(os
, MASTER_NODE_OBJ
,
2245 ZFS_SA_ATTRS
, 8, 1, &sa_obj
, tx
);
2246 ASSERT3U(error
, ==, 0);
2248 VERIFY(0 == sa_set_sa_object(os
, sa_obj
));
2249 sa_register_update_callback(os
, zfs_sa_upgrade
);
2252 spa_history_log_internal_ds(dmu_objset_ds(os
), "upgrade", tx
,
2253 "from %llu to %llu", zfsvfs
->z_version
, newvers
);
2257 zfsvfs
->z_version
= newvers
;
2259 zfs_set_fuid_feature(zfsvfs
);
2265 * Read a property stored within the master node.
2268 zfs_get_zplprop(objset_t
*os
, zfs_prop_t prop
, uint64_t *value
)
2274 * Look up the file system's value for the property. For the
2275 * version property, we look up a slightly different string.
2277 if (prop
== ZFS_PROP_VERSION
)
2278 pname
= ZPL_VERSION_STR
;
2280 pname
= zfs_prop_to_name(prop
);
2283 error
= zap_lookup(os
, MASTER_NODE_OBJ
, pname
, 8, 1, value
);
2285 if (error
== ENOENT
) {
2286 /* No value set, use the default value */
2288 case ZFS_PROP_VERSION
:
2289 *value
= ZPL_VERSION
;
2291 case ZFS_PROP_NORMALIZE
:
2292 case ZFS_PROP_UTF8ONLY
:
2296 *value
= ZFS_CASE_SENSITIVE
;
2306 static vfsdef_t vfw
= {
2310 VSW_HASPROTO
|VSW_CANRWRO
|VSW_CANREMOUNT
|VSW_VOLATILEDEV
|VSW_STATS
|
2315 struct modlfs zfs_modlfs
= {
2316 &mod_fsops
, "ZFS filesystem version " SPA_VERSION_STRING
, &vfw