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) 2011 Pawel Jakub Dawidek <pawel@dawidek.net>.
24 * All rights reserved.
27 /* Portions Copyright 2010 Robert Milkowski */
29 #include <sys/types.h>
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/kernel.h>
33 #include <sys/sysmacros.h>
36 #include <sys/vnode.h>
38 #include <sys/mntent.h>
39 #include <sys/mount.h>
40 #include <sys/cmn_err.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/sa_impl.h>
53 #include <sys/varargs.h>
54 #include <sys/policy.h>
55 #include <sys/atomic.h>
56 #include <sys/zfs_ioctl.h>
57 #include <sys/zfs_ctldir.h>
58 #include <sys/zfs_fuid.h>
59 #include <sys/sunddi.h>
61 #include <sys/dmu_objset.h>
62 #include <sys/spa_boot.h>
63 #include "zfs_comutil.h"
65 struct mtx zfs_debug_mtx;
66 MTX_SYSINIT(zfs_debug_mtx, &zfs_debug_mtx, "zfs_debug", MTX_DEF);
68 SYSCTL_NODE(_vfs, OID_AUTO, zfs, CTLFLAG_RW, 0, "ZFS file system");
71 SYSCTL_INT(_vfs_zfs, OID_AUTO, super_owner, CTLFLAG_RW, &zfs_super_owner, 0,
72 "File system owner can perform privileged operation on his file systems");
75 TUNABLE_INT("vfs.zfs.debug", &zfs_debug_level);
76 SYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RW, &zfs_debug_level, 0,
79 SYSCTL_NODE(_vfs_zfs, OID_AUTO, version, CTLFLAG_RD, 0, "ZFS versions");
80 static int zfs_version_acl = ZFS_ACL_VERSION;
81 SYSCTL_INT(_vfs_zfs_version, OID_AUTO, acl, CTLFLAG_RD, &zfs_version_acl, 0,
83 static int zfs_version_spa = SPA_VERSION;
84 SYSCTL_INT(_vfs_zfs_version, OID_AUTO, spa, CTLFLAG_RD, &zfs_version_spa, 0,
86 static int zfs_version_zpl = ZPL_VERSION;
87 SYSCTL_INT(_vfs_zfs_version, OID_AUTO, zpl, CTLFLAG_RD, &zfs_version_zpl, 0,
90 static int zfs_mount(vfs_t *vfsp);
91 static int zfs_umount(vfs_t *vfsp, int fflag);
92 static int zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp);
93 static int zfs_statfs(vfs_t *vfsp, struct statfs *statp);
94 static int zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp);
95 static int zfs_sync(vfs_t *vfsp, int waitfor);
96 static int zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp,
97 struct ucred **credanonp, int *numsecflavors, int **secflavors);
98 static int zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, vnode_t **vpp);
99 static void zfs_objset_close(zfsvfs_t *zfsvfs);
100 static void zfs_freevfs(vfs_t *vfsp);
102 static struct vfsops zfs_vfsops = {
103 .vfs_mount = zfs_mount,
104 .vfs_unmount = zfs_umount,
105 .vfs_root = zfs_root,
106 .vfs_statfs = zfs_statfs,
107 .vfs_vget = zfs_vget,
108 .vfs_sync = zfs_sync,
109 .vfs_checkexp = zfs_checkexp,
110 .vfs_fhtovp = zfs_fhtovp,
113 VFS_SET(zfs_vfsops, zfs, VFCF_JAIL | VFCF_DELEGADMIN);
116 * We need to keep a count of active fs's.
117 * This is necessary to prevent our module
118 * from being unloaded after a umount -f
120 static uint32_t zfs_active_fs_count = 0;
124 zfs_sync(vfs_t *vfsp, int waitfor)
128 * Data integrity is job one. We don't want a compromised kernel
129 * writing to the storage pool, so we never sync during panic.
136 * Sync a specific filesystem.
138 zfsvfs_t *zfsvfs = vfsp->vfs_data;
142 error = vfs_stdsync(vfsp, waitfor);
147 dp = dmu_objset_pool(zfsvfs->z_os);
150 * If the system is shutting down, then skip any
151 * filesystems which may exist on a suspended pool.
153 if (sys_shutdown && spa_suspended(dp->dp_spa)) {
158 if (zfsvfs->z_log != NULL)
159 zil_commit(zfsvfs->z_log, 0);
164 * Sync all ZFS filesystems. This is what happens when you
165 * run sync(1M). Unlike other filesystems, ZFS honors the
166 * request by waiting for all pools to commit all dirty data.
176 zfs_create_unique_device(dev_t *dev)
181 ASSERT3U(zfs_minor, <=, MAXMIN32);
182 minor_t start = zfs_minor;
184 mutex_enter(&zfs_dev_mtx);
185 if (zfs_minor >= MAXMIN32) {
187 * If we're still using the real major
188 * keep out of /dev/zfs and /dev/zvol minor
189 * number space. If we're using a getudev()'ed
190 * major number, we can use all of its minors.
192 if (zfs_major == ddi_name_to_major(ZFS_DRIVER))
193 zfs_minor = ZFS_MIN_MINOR;
199 *dev = makedevice(zfs_major, zfs_minor);
200 mutex_exit(&zfs_dev_mtx);
201 } while (vfs_devismounted(*dev) && zfs_minor != start);
202 if (zfs_minor == start) {
204 * We are using all ~262,000 minor numbers for the
205 * current major number. Create a new major number.
207 if ((new_major = getudev()) == (major_t)-1) {
209 "zfs_mount: Can't get unique major "
213 mutex_enter(&zfs_dev_mtx);
214 zfs_major = new_major;
217 mutex_exit(&zfs_dev_mtx);
221 /* CONSTANTCONDITION */
226 #endif /* !__FreeBSD__ */
229 atime_changed_cb(void *arg, uint64_t newval)
231 zfsvfs_t *zfsvfs = arg;
233 if (newval == TRUE) {
234 zfsvfs->z_atime = TRUE;
235 zfsvfs->z_vfs->vfs_flag &= ~MNT_NOATIME;
236 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME);
237 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0);
239 zfsvfs->z_atime = FALSE;
240 zfsvfs->z_vfs->vfs_flag |= MNT_NOATIME;
241 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME);
242 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0);
247 xattr_changed_cb(void *arg, uint64_t newval)
249 zfsvfs_t *zfsvfs = arg;
251 if (newval == TRUE) {
252 /* XXX locking on vfs_flag? */
254 zfsvfs->z_vfs->vfs_flag |= VFS_XATTR;
256 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR);
257 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0);
259 /* XXX locking on vfs_flag? */
261 zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR;
263 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR);
264 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0);
269 blksz_changed_cb(void *arg, uint64_t newval)
271 zfsvfs_t *zfsvfs = arg;
273 if (newval < SPA_MINBLOCKSIZE ||
274 newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
275 newval = SPA_MAXBLOCKSIZE;
277 zfsvfs->z_max_blksz = newval;
278 zfsvfs->z_vfs->mnt_stat.f_iosize = newval;
282 readonly_changed_cb(void *arg, uint64_t newval)
284 zfsvfs_t *zfsvfs = arg;
287 /* XXX locking on vfs_flag? */
288 zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY;
289 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW);
290 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0);
292 /* XXX locking on vfs_flag? */
293 zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
294 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO);
295 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0);
300 setuid_changed_cb(void *arg, uint64_t newval)
302 zfsvfs_t *zfsvfs = arg;
304 if (newval == FALSE) {
305 zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID;
306 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID);
307 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0);
309 zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID;
310 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID);
311 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0);
316 exec_changed_cb(void *arg, uint64_t newval)
318 zfsvfs_t *zfsvfs = arg;
320 if (newval == FALSE) {
321 zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC;
322 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC);
323 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0);
325 zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC;
326 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC);
327 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0);
332 * The nbmand mount option can be changed at mount time.
333 * We can't allow it to be toggled on live file systems or incorrect
334 * behavior may be seen from cifs clients
336 * This property isn't registered via dsl_prop_register(), but this callback
337 * will be called when a file system is first mounted
340 nbmand_changed_cb(void *arg, uint64_t newval)
342 zfsvfs_t *zfsvfs = arg;
343 if (newval == FALSE) {
344 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND);
345 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0);
347 vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND);
348 vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0);
353 snapdir_changed_cb(void *arg, uint64_t newval)
355 zfsvfs_t *zfsvfs = arg;
357 zfsvfs->z_show_ctldir = newval;
361 vscan_changed_cb(void *arg, uint64_t newval)
363 zfsvfs_t *zfsvfs = arg;
365 zfsvfs->z_vscan = newval;
369 acl_mode_changed_cb(void *arg, uint64_t newval)
371 zfsvfs_t *zfsvfs = arg;
373 zfsvfs->z_acl_mode = newval;
377 acl_inherit_changed_cb(void *arg, uint64_t newval)
379 zfsvfs_t *zfsvfs = arg;
381 zfsvfs->z_acl_inherit = newval;
385 zfs_register_callbacks(vfs_t *vfsp)
387 struct dsl_dataset *ds = NULL;
389 zfsvfs_t *zfsvfs = NULL;
391 int readonly, do_readonly = B_FALSE;
392 int setuid, do_setuid = B_FALSE;
393 int exec, do_exec = B_FALSE;
394 int xattr, do_xattr = B_FALSE;
395 int atime, do_atime = B_FALSE;
399 zfsvfs = vfsp->vfs_data;
404 * This function can be called for a snapshot when we update snapshot's
405 * mount point, which isn't really supported.
407 if (dmu_objset_is_snapshot(os))
411 * The act of registering our callbacks will destroy any mount
412 * options we may have. In order to enable temporary overrides
413 * of mount options, we stash away the current values and
414 * restore them after we register the callbacks.
416 if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) ||
417 !spa_writeable(dmu_objset_spa(os))) {
419 do_readonly = B_TRUE;
420 } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
422 do_readonly = B_TRUE;
424 if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
428 if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
431 } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
436 if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
439 } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
443 if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
446 } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
450 if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) {
453 } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) {
459 * nbmand is a special property. It can only be changed at
462 * This is weird, but it is documented to only be changeable
465 if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
467 } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) {
470 char osname[MAXNAMELEN];
472 dmu_objset_name(os, osname);
473 if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand,
480 * Register property callbacks.
482 * It would probably be fine to just check for i/o error from
483 * the first prop_register(), but I guess I like to go
486 ds = dmu_objset_ds(os);
487 error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs);
488 error = error ? error : dsl_prop_register(ds,
489 "xattr", xattr_changed_cb, zfsvfs);
490 error = error ? error : dsl_prop_register(ds,
491 "recordsize", blksz_changed_cb, zfsvfs);
492 error = error ? error : dsl_prop_register(ds,
493 "readonly", readonly_changed_cb, zfsvfs);
494 error = error ? error : dsl_prop_register(ds,
495 "setuid", setuid_changed_cb, zfsvfs);
496 error = error ? error : dsl_prop_register(ds,
497 "exec", exec_changed_cb, zfsvfs);
498 error = error ? error : dsl_prop_register(ds,
499 "snapdir", snapdir_changed_cb, zfsvfs);
500 error = error ? error : dsl_prop_register(ds,
501 "aclmode", acl_mode_changed_cb, zfsvfs);
502 error = error ? error : dsl_prop_register(ds,
503 "aclinherit", acl_inherit_changed_cb, zfsvfs);
504 error = error ? error : dsl_prop_register(ds,
505 "vscan", vscan_changed_cb, zfsvfs);
510 * Invoke our callbacks to restore temporary mount options.
513 readonly_changed_cb(zfsvfs, readonly);
515 setuid_changed_cb(zfsvfs, setuid);
517 exec_changed_cb(zfsvfs, exec);
519 xattr_changed_cb(zfsvfs, xattr);
521 atime_changed_cb(zfsvfs, atime);
523 nbmand_changed_cb(zfsvfs, nbmand);
529 * We may attempt to unregister some callbacks that are not
530 * registered, but this is OK; it will simply return ENOMSG,
531 * which we will ignore.
533 (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs);
534 (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs);
535 (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
536 (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
537 (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
538 (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
539 (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
540 (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
541 (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
543 (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs);
549 zfs_space_delta_cb(dmu_object_type_t bonustype, void *data,
550 uint64_t *userp, uint64_t *groupp)
555 * Is it a valid type of object to track?
557 if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
561 * If we have a NULL data pointer
562 * then assume the id's aren't changing and
563 * return EEXIST to the dmu to let it know to
569 if (bonustype == DMU_OT_ZNODE) {
570 znode_phys_t *znp = data;
571 *userp = znp->zp_uid;
572 *groupp = znp->zp_gid;
575 sa_hdr_phys_t *sap = data;
576 sa_hdr_phys_t sa = *sap;
577 boolean_t swap = B_FALSE;
579 ASSERT(bonustype == DMU_OT_SA);
581 if (sa.sa_magic == 0) {
583 * This should only happen for newly created
584 * files that haven't had the znode data filled
591 if (sa.sa_magic == BSWAP_32(SA_MAGIC)) {
592 sa.sa_magic = SA_MAGIC;
593 sa.sa_layout_info = BSWAP_16(sa.sa_layout_info);
596 VERIFY3U(sa.sa_magic, ==, SA_MAGIC);
599 hdrsize = sa_hdrsize(&sa);
600 VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t));
601 *userp = *((uint64_t *)((uintptr_t)data + hdrsize +
603 *groupp = *((uint64_t *)((uintptr_t)data + hdrsize +
606 *userp = BSWAP_64(*userp);
607 *groupp = BSWAP_64(*groupp);
614 fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr,
615 char *domainbuf, int buflen, uid_t *ridp)
620 fuid = strtonum(fuidstr, NULL);
622 domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid));
624 (void) strlcpy(domainbuf, domain, buflen);
627 *ridp = FUID_RID(fuid);
631 zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type)
634 case ZFS_PROP_USERUSED:
635 return (DMU_USERUSED_OBJECT);
636 case ZFS_PROP_GROUPUSED:
637 return (DMU_GROUPUSED_OBJECT);
638 case ZFS_PROP_USERQUOTA:
639 return (zfsvfs->z_userquota_obj);
640 case ZFS_PROP_GROUPQUOTA:
641 return (zfsvfs->z_groupquota_obj);
647 zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
648 uint64_t *cookiep, void *vbuf, uint64_t *bufsizep)
653 zfs_useracct_t *buf = vbuf;
656 if (!dmu_objset_userspace_present(zfsvfs->z_os))
659 obj = zfs_userquota_prop_to_obj(zfsvfs, type);
665 for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep);
666 (error = zap_cursor_retrieve(&zc, &za)) == 0;
667 zap_cursor_advance(&zc)) {
668 if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) >
672 fuidstr_to_sid(zfsvfs, za.za_name,
673 buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid);
675 buf->zu_space = za.za_first_integer;
681 ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep);
682 *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf;
683 *cookiep = zap_cursor_serialize(&zc);
684 zap_cursor_fini(&zc);
689 * buf must be big enough (eg, 32 bytes)
692 id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid,
693 char *buf, boolean_t addok)
698 if (domain && domain[0]) {
699 domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok);
703 fuid = FUID_ENCODE(domainid, rid);
704 (void) sprintf(buf, "%llx", (longlong_t)fuid);
709 zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
710 const char *domain, uint64_t rid, uint64_t *valp)
718 if (!dmu_objset_userspace_present(zfsvfs->z_os))
721 obj = zfs_userquota_prop_to_obj(zfsvfs, type);
725 err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE);
729 err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp);
736 zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
737 const char *domain, uint64_t rid, uint64_t quota)
743 boolean_t fuid_dirtied;
745 if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
748 if (zfsvfs->z_version < ZPL_VERSION_USERSPACE)
751 objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj :
752 &zfsvfs->z_groupquota_obj;
754 err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE);
757 fuid_dirtied = zfsvfs->z_fuid_dirty;
759 tx = dmu_tx_create(zfsvfs->z_os);
760 dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
762 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
763 zfs_userquota_prop_prefixes[type]);
766 zfs_fuid_txhold(zfsvfs, tx);
767 err = dmu_tx_assign(tx, TXG_WAIT);
773 mutex_enter(&zfsvfs->z_lock);
775 *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA,
777 VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
778 zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
780 mutex_exit(&zfsvfs->z_lock);
783 err = zap_remove(zfsvfs->z_os, *objp, buf, tx);
787 err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx);
791 zfs_fuid_sync(zfsvfs, tx);
797 zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid)
800 uint64_t used, quota, usedobj, quotaobj;
803 usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
804 quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;
806 if (quotaobj == 0 || zfsvfs->z_replay)
809 (void) sprintf(buf, "%llx", (longlong_t)fuid);
810 err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a);
814 err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used);
817 return (used >= quota);
821 zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup)
826 quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;
828 fuid = isgroup ? zp->z_gid : zp->z_uid;
830 if (quotaobj == 0 || zfsvfs->z_replay)
833 return (zfs_fuid_overquota(zfsvfs, isgroup, fuid));
837 zfsvfs_create(const char *osname, zfsvfs_t **zfvp)
845 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
848 * We claim to always be readonly so we can open snapshots;
849 * other ZPL code will prevent us from writing to snapshots.
851 error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os);
853 kmem_free(zfsvfs, sizeof (zfsvfs_t));
858 * Initialize the zfs-specific filesystem structure.
859 * Should probably make this a kmem cache, shuffle fields,
860 * and just bzero up to z_hold_mtx[].
862 zfsvfs->z_vfs = NULL;
863 zfsvfs->z_parent = zfsvfs;
864 zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
865 zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
868 error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
871 } else if (zfsvfs->z_version >
872 zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
873 (void) printf("Can't mount a version %lld file system "
874 "on a version %lld pool\n. Pool must be upgraded to mount "
875 "this file system.", (u_longlong_t)zfsvfs->z_version,
876 (u_longlong_t)spa_version(dmu_objset_spa(os)));
880 if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
882 zfsvfs->z_norm = (int)zval;
884 if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
886 zfsvfs->z_utf8 = (zval != 0);
888 if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
890 zfsvfs->z_case = (uint_t)zval;
893 * Fold case on file systems that are always or sometimes case
896 if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
897 zfsvfs->z_case == ZFS_CASE_MIXED)
898 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
900 zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
901 zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os);
903 if (zfsvfs->z_use_sa) {
904 /* should either have both of these objects or none */
905 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
911 * Pre SA versions file systems should never touch
912 * either the attribute registration or layout objects.
917 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
918 &zfsvfs->z_attr_table);
922 if (zfsvfs->z_version >= ZPL_VERSION_SA)
923 sa_register_update_callback(os, zfs_sa_upgrade);
925 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
929 ASSERT(zfsvfs->z_root != 0);
931 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
932 &zfsvfs->z_unlinkedobj);
936 error = zap_lookup(os, MASTER_NODE_OBJ,
937 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
938 8, 1, &zfsvfs->z_userquota_obj);
939 if (error && error != ENOENT)
942 error = zap_lookup(os, MASTER_NODE_OBJ,
943 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
944 8, 1, &zfsvfs->z_groupquota_obj);
945 if (error && error != ENOENT)
948 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
949 &zfsvfs->z_fuid_obj);
950 if (error && error != ENOENT)
953 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
954 &zfsvfs->z_shares_dir);
955 if (error && error != ENOENT)
958 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
959 mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL);
960 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
961 offsetof(znode_t, z_link_node));
962 rrw_init(&zfsvfs->z_teardown_lock);
963 rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
964 rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
965 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
966 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
972 dmu_objset_disown(os, zfsvfs);
974 kmem_free(zfsvfs, sizeof (zfsvfs_t));
979 zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
983 error = zfs_register_callbacks(zfsvfs->z_vfs);
988 * Set the objset user_ptr to track its zfsvfs.
990 mutex_enter(&zfsvfs->z_os->os_user_ptr_lock);
991 dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
992 mutex_exit(&zfsvfs->z_os->os_user_ptr_lock);
994 zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
997 * If we are not mounting (ie: online recv), then we don't
998 * have to worry about replaying the log as we blocked all
999 * operations out since we closed the ZIL.
1005 * During replay we remove the read only flag to
1006 * allow replays to succeed.
1008 readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY;
1010 zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
1012 zfs_unlinked_drain(zfsvfs);
1015 * Parse and replay the intent log.
1017 * Because of ziltest, this must be done after
1018 * zfs_unlinked_drain(). (Further note: ziltest
1019 * doesn't use readonly mounts, where
1020 * zfs_unlinked_drain() isn't called.) This is because
1021 * ziltest causes spa_sync() to think it's committed,
1022 * but actually it is not, so the intent log contains
1023 * many txg's worth of changes.
1025 * In particular, if object N is in the unlinked set in
1026 * the last txg to actually sync, then it could be
1027 * actually freed in a later txg and then reallocated
1028 * in a yet later txg. This would write a "create
1029 * object N" record to the intent log. Normally, this
1030 * would be fine because the spa_sync() would have
1031 * written out the fact that object N is free, before
1032 * we could write the "create object N" intent log
1035 * But when we are in ziltest mode, we advance the "open
1036 * txg" without actually spa_sync()-ing the changes to
1037 * disk. So we would see that object N is still
1038 * allocated and in the unlinked set, and there is an
1039 * intent log record saying to allocate it.
1041 if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) {
1042 if (zil_replay_disable) {
1043 zil_destroy(zfsvfs->z_log, B_FALSE);
1045 zfsvfs->z_replay = B_TRUE;
1046 zil_replay(zfsvfs->z_os, zfsvfs,
1048 zfsvfs->z_replay = B_FALSE;
1051 zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */
1057 extern krwlock_t zfsvfs_lock; /* in zfs_znode.c */
1060 zfsvfs_free(zfsvfs_t *zfsvfs)
1065 * This is a barrier to prevent the filesystem from going away in
1066 * zfs_znode_move() until we can safely ensure that the filesystem is
1067 * not unmounted. We consider the filesystem valid before the barrier
1068 * and invalid after the barrier.
1070 rw_enter(&zfsvfs_lock, RW_READER);
1071 rw_exit(&zfsvfs_lock);
1073 zfs_fuid_destroy(zfsvfs);
1075 mutex_destroy(&zfsvfs->z_znodes_lock);
1076 mutex_destroy(&zfsvfs->z_lock);
1077 list_destroy(&zfsvfs->z_all_znodes);
1078 rrw_destroy(&zfsvfs->z_teardown_lock);
1079 rw_destroy(&zfsvfs->z_teardown_inactive_lock);
1080 rw_destroy(&zfsvfs->z_fuid_lock);
1081 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1082 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1083 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1087 zfs_set_fuid_feature(zfsvfs_t *zfsvfs)
1089 zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
1090 if (zfsvfs->z_vfs) {
1091 if (zfsvfs->z_use_fuids) {
1092 vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR);
1093 vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS);
1094 vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS);
1095 vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE);
1096 vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER);
1097 vfs_set_feature(zfsvfs->z_vfs, VFSFT_REPARSE);
1099 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_XVATTR);
1100 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS);
1101 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS);
1102 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE);
1103 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER);
1104 vfs_clear_feature(zfsvfs->z_vfs, VFSFT_REPARSE);
1107 zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os);
1111 zfs_domount(vfs_t *vfsp, char *osname)
1113 uint64_t recordsize, fsid_guid;
1121 error = zfsvfs_create(osname, &zfsvfs);
1124 zfsvfs->z_vfs = vfsp;
1126 if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize,
1129 zfsvfs->z_vfs->vfs_bsize = SPA_MINBLOCKSIZE;
1130 zfsvfs->z_vfs->mnt_stat.f_iosize = recordsize;
1132 vfsp->vfs_data = zfsvfs;
1133 vfsp->mnt_flag |= MNT_LOCAL;
1134 vfsp->mnt_kern_flag |= MNTK_MPSAFE;
1135 vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED;
1136 vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES;
1137 vfsp->mnt_kern_flag |= MNTK_EXTENDED_SHARED;
1140 * The fsid is 64 bits, composed of an 8-bit fs type, which
1141 * separates our fsid from any other filesystem types, and a
1142 * 56-bit objset unique ID. The objset unique ID is unique to
1143 * all objsets open on this system, provided by unique_create().
1144 * The 8-bit fs type must be put in the low bits of fsid[1]
1145 * because that's where other Solaris filesystems put it.
1147 fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os);
1148 ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
1149 vfsp->vfs_fsid.val[0] = fsid_guid;
1150 vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
1151 vfsp->mnt_vfc->vfc_typenum & 0xFF;
1154 * Set features for file system.
1156 zfs_set_fuid_feature(zfsvfs);
1157 if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
1158 vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
1159 vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
1160 vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE);
1161 } else if (zfsvfs->z_case == ZFS_CASE_MIXED) {
1162 vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
1163 vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
1165 vfs_set_feature(vfsp, VFSFT_ZEROCOPY_SUPPORTED);
1167 if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
1170 atime_changed_cb(zfsvfs, B_FALSE);
1171 readonly_changed_cb(zfsvfs, B_TRUE);
1172 if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))
1174 xattr_changed_cb(zfsvfs, pval);
1175 zfsvfs->z_issnap = B_TRUE;
1176 zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED;
1178 mutex_enter(&zfsvfs->z_os->os_user_ptr_lock);
1179 dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
1180 mutex_exit(&zfsvfs->z_os->os_user_ptr_lock);
1182 error = zfsvfs_setup(zfsvfs, B_TRUE);
1185 vfs_mountedfrom(vfsp, osname);
1186 /* Grab extra reference. */
1187 VERIFY(VFS_ROOT(vfsp, LK_EXCLUSIVE, &vp) == 0);
1190 if (!zfsvfs->z_issnap)
1191 zfsctl_create(zfsvfs);
1194 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1195 zfsvfs_free(zfsvfs);
1197 atomic_add_32(&zfs_active_fs_count, 1);
1204 zfs_unregister_callbacks(zfsvfs_t *zfsvfs)
1206 objset_t *os = zfsvfs->z_os;
1207 struct dsl_dataset *ds;
1210 * Unregister properties.
1212 if (!dmu_objset_is_snapshot(os)) {
1213 ds = dmu_objset_ds(os);
1214 VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
1217 VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
1220 VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
1223 VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
1226 VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
1229 VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
1232 VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
1235 VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb,
1238 VERIFY(dsl_prop_unregister(ds, "aclinherit",
1239 acl_inherit_changed_cb, zfsvfs) == 0);
1241 VERIFY(dsl_prop_unregister(ds, "vscan",
1242 vscan_changed_cb, zfsvfs) == 0);
1248 * Convert a decimal digit string to a uint64_t integer.
1251 str_to_uint64(char *str, uint64_t *objnum)
1256 if (*str < '0' || *str > '9')
1259 num = num*10 + *str++ - '0';
1267 * The boot path passed from the boot loader is in the form of
1268 * "rootpool-name/root-filesystem-object-number'. Convert this
1269 * string to a dataset name: "rootpool-name/root-filesystem-name".
1272 zfs_parse_bootfs(char *bpath, char *outpath)
1278 if (*bpath == 0 || *bpath == '/')
1281 (void) strcpy(outpath, bpath);
1283 slashp = strchr(bpath, '/');
1285 /* if no '/', just return the pool name */
1286 if (slashp == NULL) {
1290 /* if not a number, just return the root dataset name */
1291 if (str_to_uint64(slashp+1, &objnum)) {
1296 error = dsl_dsobj_to_dsname(bpath, objnum, outpath);
1303 * zfs_check_global_label:
1304 * Check that the hex label string is appropriate for the dataset
1305 * being mounted into the global_zone proper.
1307 * Return an error if the hex label string is not default or
1308 * admin_low/admin_high. For admin_low labels, the corresponding
1309 * dataset must be readonly.
1312 zfs_check_global_label(const char *dsname, const char *hexsl)
1314 if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0)
1316 if (strcasecmp(hexsl, ADMIN_HIGH) == 0)
1318 if (strcasecmp(hexsl, ADMIN_LOW) == 0) {
1319 /* must be readonly */
1322 if (dsl_prop_get_integer(dsname,
1323 zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL))
1325 return (rdonly ? 0 : EACCES);
1331 * zfs_mount_label_policy:
1332 * Determine whether the mount is allowed according to MAC check.
1333 * by comparing (where appropriate) label of the dataset against
1334 * the label of the zone being mounted into. If the dataset has
1335 * no label, create one.
1338 * 0 : access allowed
1339 * >0 : error code, such as EACCES
1342 zfs_mount_label_policy(vfs_t *vfsp, char *osname)
1345 zone_t *mntzone = NULL;
1346 ts_label_t *mnt_tsl;
1349 char ds_hexsl[MAXNAMELEN];
1351 retv = EACCES; /* assume the worst */
1354 * Start by getting the dataset label if it exists.
1356 error = dsl_prop_get(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
1357 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
1362 * If labeling is NOT enabled, then disallow the mount of datasets
1363 * which have a non-default label already. No other label checks
1366 if (!is_system_labeled()) {
1367 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0)
1373 * Get the label of the mountpoint. If mounting into the global
1374 * zone (i.e. mountpoint is not within an active zone and the
1375 * zoned property is off), the label must be default or
1376 * admin_low/admin_high only; no other checks are needed.
1378 mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE);
1379 if (mntzone->zone_id == GLOBAL_ZONEID) {
1384 if (dsl_prop_get_integer(osname,
1385 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
1388 return (zfs_check_global_label(osname, ds_hexsl));
1391 * This is the case of a zone dataset being mounted
1392 * initially, before the zone has been fully created;
1393 * allow this mount into global zone.
1398 mnt_tsl = mntzone->zone_slabel;
1399 ASSERT(mnt_tsl != NULL);
1400 label_hold(mnt_tsl);
1401 mnt_sl = label2bslabel(mnt_tsl);
1403 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) {
1405 * The dataset doesn't have a real label, so fabricate one.
1409 if (l_to_str_internal(mnt_sl, &str) == 0 &&
1410 dsl_prop_set(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
1411 ZPROP_SRC_LOCAL, 1, strlen(str) + 1, str) == 0)
1414 kmem_free(str, strlen(str) + 1);
1415 } else if (hexstr_to_label(ds_hexsl, &ds_sl) == 0) {
1417 * Now compare labels to complete the MAC check. If the
1418 * labels are equal then allow access. If the mountpoint
1419 * label dominates the dataset label, allow readonly access.
1420 * Otherwise, access is denied.
1422 if (blequal(mnt_sl, &ds_sl))
1424 else if (bldominates(mnt_sl, &ds_sl)) {
1425 vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0);
1430 label_rele(mnt_tsl);
1434 #endif /* SECLABEL */
1436 #ifdef OPENSOLARIS_MOUNTROOT
1438 zfs_mountroot(vfs_t *vfsp, enum whymountroot why)
1441 static int zfsrootdone = 0;
1442 zfsvfs_t *zfsvfs = NULL;
1451 * The filesystem that we mount as root is defined in the
1452 * boot property "zfs-bootfs" with a format of
1453 * "poolname/root-dataset-objnum".
1455 if (why == ROOT_INIT) {
1459 * the process of doing a spa_load will require the
1460 * clock to be set before we could (for example) do
1461 * something better by looking at the timestamp on
1462 * an uberblock, so just set it to -1.
1466 if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) {
1467 cmn_err(CE_NOTE, "spa_get_bootfs: can not get "
1471 zfs_devid = spa_get_bootprop("diskdevid");
1472 error = spa_import_rootpool(rootfs.bo_name, zfs_devid);
1474 spa_free_bootprop(zfs_devid);
1476 spa_free_bootprop(zfs_bootfs);
1477 cmn_err(CE_NOTE, "spa_import_rootpool: error %d",
1481 if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) {
1482 spa_free_bootprop(zfs_bootfs);
1483 cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d",
1488 spa_free_bootprop(zfs_bootfs);
1490 if (error = vfs_lock(vfsp))
1493 if (error = zfs_domount(vfsp, rootfs.bo_name)) {
1494 cmn_err(CE_NOTE, "zfs_domount: error %d", error);
1498 zfsvfs = (zfsvfs_t *)vfsp->vfs_data;
1500 if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) {
1501 cmn_err(CE_NOTE, "zfs_zget: error %d", error);
1506 mutex_enter(&vp->v_lock);
1507 vp->v_flag |= VROOT;
1508 mutex_exit(&vp->v_lock);
1512 * Leave rootvp held. The root file system is never unmounted.
1515 vfs_add((struct vnode *)0, vfsp,
1516 (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0);
1520 } else if (why == ROOT_REMOUNT) {
1521 readonly_changed_cb(vfsp->vfs_data, B_FALSE);
1522 vfsp->vfs_flag |= VFS_REMOUNT;
1524 /* refresh mount options */
1525 zfs_unregister_callbacks(vfsp->vfs_data);
1526 return (zfs_register_callbacks(vfsp));
1528 } else if (why == ROOT_UNMOUNT) {
1529 zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data);
1530 (void) zfs_sync(vfsp, 0, 0);
1535 * if "why" is equal to anything else other than ROOT_INIT,
1536 * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it.
1540 #endif /* OPENSOLARIS_MOUNTROOT */
1543 getpoolname(const char *osname, char *poolname)
1547 p = strchr(osname, '/');
1549 if (strlen(osname) >= MAXNAMELEN)
1550 return (ENAMETOOLONG);
1551 (void) strcpy(poolname, osname);
1553 if (p - osname >= MAXNAMELEN)
1554 return (ENAMETOOLONG);
1555 (void) strncpy(poolname, osname, p - osname);
1556 poolname[p - osname] = '\0';
1563 zfs_mount(vfs_t *vfsp)
1565 kthread_t *td = curthread;
1566 vnode_t *mvp = vfsp->mnt_vnodecovered;
1567 cred_t *cr = td->td_ucred;
1572 if (vfs_getopt(vfsp->mnt_optnew, "from", (void **)&osname, NULL))
1576 * If full-owner-access is enabled and delegated administration is
1577 * turned on, we must set nosuid.
1579 if (zfs_super_owner &&
1580 dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != ECANCELED) {
1581 secpolicy_fs_mount_clearopts(cr, vfsp);
1585 * Check for mount privilege?
1587 * If we don't have privilege then see if
1588 * we have local permission to allow it
1590 error = secpolicy_fs_mount(cr, mvp, vfsp);
1592 if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != 0)
1595 if (!(vfsp->vfs_flag & MS_REMOUNT)) {
1599 * Make sure user is the owner of the mount point
1600 * or has sufficient privileges.
1603 vattr.va_mask = AT_UID;
1605 vn_lock(mvp, LK_SHARED | LK_RETRY);
1606 if (VOP_GETATTR(mvp, &vattr, cr)) {
1611 if (secpolicy_vnode_owner(mvp, cr, vattr.va_uid) != 0 &&
1612 VOP_ACCESS(mvp, VWRITE, cr, td) != 0) {
1619 secpolicy_fs_mount_clearopts(cr, vfsp);
1623 * Refuse to mount a filesystem if we are in a local zone and the
1624 * dataset is not visible.
1626 if (!INGLOBALZONE(curthread) &&
1627 (!zone_dataset_visible(osname, &canwrite) || !canwrite)) {
1633 error = zfs_mount_label_policy(vfsp, osname);
1638 vfsp->vfs_flag |= MNT_NFS4ACLS;
1641 * When doing a remount, we simply refresh our temporary properties
1642 * according to those options set in the current VFS options.
1644 if (vfsp->vfs_flag & MS_REMOUNT) {
1645 /* refresh mount options */
1646 zfs_unregister_callbacks(vfsp->vfs_data);
1647 error = zfs_register_callbacks(vfsp);
1651 /* Initial root mount: try hard to import the requested root pool. */
1652 if ((vfsp->vfs_flag & MNT_ROOTFS) != 0 &&
1653 (vfsp->vfs_flag & MNT_UPDATE) == 0) {
1654 char pname[MAXNAMELEN];
1656 error = getpoolname(osname, pname);
1658 error = spa_import_rootpool(pname);
1663 error = zfs_domount(vfsp, osname);
1668 * Add an extra VFS_HOLD on our parent vfs so that it can't
1669 * disappear due to a forced unmount.
1671 if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap)
1672 VFS_HOLD(mvp->v_vfsp);
1680 zfs_statfs(vfs_t *vfsp, struct statfs *statp)
1682 zfsvfs_t *zfsvfs = vfsp->vfs_data;
1683 uint64_t refdbytes, availbytes, usedobjs, availobjs;
1685 statp->f_version = STATFS_VERSION;
1689 dmu_objset_space(zfsvfs->z_os,
1690 &refdbytes, &availbytes, &usedobjs, &availobjs);
1693 * The underlying storage pool actually uses multiple block sizes.
1694 * We report the fragsize as the smallest block size we support,
1695 * and we report our blocksize as the filesystem's maximum blocksize.
1697 statp->f_bsize = SPA_MINBLOCKSIZE;
1698 statp->f_iosize = zfsvfs->z_vfs->mnt_stat.f_iosize;
1701 * The following report "total" blocks of various kinds in the
1702 * file system, but reported in terms of f_frsize - the
1706 statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT;
1707 statp->f_bfree = availbytes / statp->f_bsize;
1708 statp->f_bavail = statp->f_bfree; /* no root reservation */
1711 * statvfs() should really be called statufs(), because it assumes
1712 * static metadata. ZFS doesn't preallocate files, so the best
1713 * we can do is report the max that could possibly fit in f_files,
1714 * and that minus the number actually used in f_ffree.
1715 * For f_ffree, report the smaller of the number of object available
1716 * and the number of blocks (each object will take at least a block).
1718 statp->f_ffree = MIN(availobjs, statp->f_bfree);
1719 statp->f_files = statp->f_ffree + usedobjs;
1722 * We're a zfs filesystem.
1724 (void) strlcpy(statp->f_fstypename, "zfs", sizeof(statp->f_fstypename));
1726 strlcpy(statp->f_mntfromname, vfsp->mnt_stat.f_mntfromname,
1727 sizeof(statp->f_mntfromname));
1728 strlcpy(statp->f_mntonname, vfsp->mnt_stat.f_mntonname,
1729 sizeof(statp->f_mntonname));
1731 statp->f_namemax = ZFS_MAXNAMELEN;
1738 zfs_vnode_lock(vnode_t *vp, int flags)
1744 error = vn_lock(vp, flags);
1749 zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp)
1751 zfsvfs_t *zfsvfs = vfsp->vfs_data;
1755 ZFS_ENTER_NOERROR(zfsvfs);
1757 error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
1759 *vpp = ZTOV(rootzp);
1764 error = zfs_vnode_lock(*vpp, flags);
1766 (*vpp)->v_vflag |= VV_ROOT;
1775 * Teardown the zfsvfs::z_os.
1777 * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
1778 * and 'z_teardown_inactive_lock' held.
1781 zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
1785 rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG);
1789 * We purge the parent filesystem's vfsp as the parent
1790 * filesystem and all of its snapshots have their vnode's
1791 * v_vfsp set to the parent's filesystem's vfsp. Note,
1792 * 'z_parent' is self referential for non-snapshots.
1794 (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
1795 #ifdef FREEBSD_NAMECACHE
1796 cache_purgevfs(zfsvfs->z_parent->z_vfs);
1801 * Close the zil. NB: Can't close the zil while zfs_inactive
1802 * threads are blocked as zil_close can call zfs_inactive.
1804 if (zfsvfs->z_log) {
1805 zil_close(zfsvfs->z_log);
1806 zfsvfs->z_log = NULL;
1809 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER);
1812 * If we are not unmounting (ie: online recv) and someone already
1813 * unmounted this file system while we were doing the switcheroo,
1814 * or a reopen of z_os failed then just bail out now.
1816 if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) {
1817 rw_exit(&zfsvfs->z_teardown_inactive_lock);
1818 rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
1823 * At this point there are no vops active, and any new vops will
1824 * fail with EIO since we have z_teardown_lock for writer (only
1825 * relavent for forced unmount).
1827 * Release all holds on dbufs.
1829 mutex_enter(&zfsvfs->z_znodes_lock);
1830 for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL;
1831 zp = list_next(&zfsvfs->z_all_znodes, zp))
1833 ASSERT(ZTOV(zp)->v_count >= 0);
1834 zfs_znode_dmu_fini(zp);
1836 mutex_exit(&zfsvfs->z_znodes_lock);
1839 * If we are unmounting, set the unmounted flag and let new vops
1840 * unblock. zfs_inactive will have the unmounted behavior, and all
1841 * other vops will fail with EIO.
1844 zfsvfs->z_unmounted = B_TRUE;
1845 rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
1846 rw_exit(&zfsvfs->z_teardown_inactive_lock);
1850 * Some znodes might not be fully reclaimed, wait for them.
1852 mutex_enter(&zfsvfs->z_znodes_lock);
1853 while (list_head(&zfsvfs->z_all_znodes) != NULL) {
1854 msleep(zfsvfs, &zfsvfs->z_znodes_lock, 0,
1857 mutex_exit(&zfsvfs->z_znodes_lock);
1862 * z_os will be NULL if there was an error in attempting to reopen
1863 * zfsvfs, so just return as the properties had already been
1864 * unregistered and cached data had been evicted before.
1866 if (zfsvfs->z_os == NULL)
1870 * Unregister properties.
1872 zfs_unregister_callbacks(zfsvfs);
1877 if (dsl_dataset_is_dirty(dmu_objset_ds(zfsvfs->z_os)) &&
1878 !(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY))
1879 txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
1880 (void) dmu_objset_evict_dbufs(zfsvfs->z_os);
1887 zfs_umount(vfs_t *vfsp, int fflag)
1889 kthread_t *td = curthread;
1890 zfsvfs_t *zfsvfs = vfsp->vfs_data;
1892 cred_t *cr = td->td_ucred;
1895 ret = secpolicy_fs_unmount(cr, vfsp);
1897 if (dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource),
1898 ZFS_DELEG_PERM_MOUNT, cr))
1903 * We purge the parent filesystem's vfsp as the parent filesystem
1904 * and all of its snapshots have their vnode's v_vfsp set to the
1905 * parent's filesystem's vfsp. Note, 'z_parent' is self
1906 * referential for non-snapshots.
1908 (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
1911 * Unmount any snapshots mounted under .zfs before unmounting the
1914 if (zfsvfs->z_ctldir != NULL) {
1915 if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0)
1917 ret = vflush(vfsp, 0, 0, td);
1918 ASSERT(ret == EBUSY);
1919 if (!(fflag & MS_FORCE)) {
1920 if (zfsvfs->z_ctldir->v_count > 1)
1922 ASSERT(zfsvfs->z_ctldir->v_count == 1);
1924 zfsctl_destroy(zfsvfs);
1925 ASSERT(zfsvfs->z_ctldir == NULL);
1928 if (fflag & MS_FORCE) {
1930 * Mark file system as unmounted before calling
1931 * vflush(FORCECLOSE). This way we ensure no future vnops
1932 * will be called and risk operating on DOOMED vnodes.
1934 rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG);
1935 zfsvfs->z_unmounted = B_TRUE;
1936 rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
1940 * Flush all the files.
1942 ret = vflush(vfsp, 1, (fflag & MS_FORCE) ? FORCECLOSE : 0, td);
1944 if (!zfsvfs->z_issnap) {
1945 zfsctl_create(zfsvfs);
1946 ASSERT(zfsvfs->z_ctldir != NULL);
1951 if (!(fflag & MS_FORCE)) {
1953 * Check the number of active vnodes in the file system.
1954 * Our count is maintained in the vfs structure, but the
1955 * number is off by 1 to indicate a hold on the vfs
1958 * The '.zfs' directory maintains a reference of its
1959 * own, and any active references underneath are
1960 * reflected in the vnode count.
1962 if (zfsvfs->z_ctldir == NULL) {
1963 if (vfsp->vfs_count > 1)
1966 if (vfsp->vfs_count > 2 ||
1967 zfsvfs->z_ctldir->v_count > 1)
1972 VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0);
1976 * z_os will be NULL if there was an error in
1977 * attempting to reopen zfsvfs.
1981 * Unset the objset user_ptr.
1983 mutex_enter(&os->os_user_ptr_lock);
1984 dmu_objset_set_user(os, NULL);
1985 mutex_exit(&os->os_user_ptr_lock);
1988 * Finally release the objset
1990 dmu_objset_disown(os, zfsvfs);
1994 * We can now safely destroy the '.zfs' directory node.
1996 if (zfsvfs->z_ctldir != NULL)
1997 zfsctl_destroy(zfsvfs);
1998 if (zfsvfs->z_issnap) {
1999 vnode_t *svp = vfsp->mnt_vnodecovered;
2001 if (svp->v_count >= 2)
2010 zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp)
2012 zfsvfs_t *zfsvfs = vfsp->vfs_data;
2017 * zfs_zget() can't operate on virtual entries like .zfs/ or
2018 * .zfs/snapshot/ directories, that's why we return EOPNOTSUPP.
2019 * This will make NFS to switch to LOOKUP instead of using VGET.
2021 if (ino == ZFSCTL_INO_ROOT || ino == ZFSCTL_INO_SNAPDIR ||
2022 (zfsvfs->z_shares_dir != 0 && ino == zfsvfs->z_shares_dir))
2023 return (EOPNOTSUPP);
2026 err = zfs_zget(zfsvfs, ino, &zp);
2027 if (err == 0 && zp->z_unlinked) {
2035 err = zfs_vnode_lock(*vpp, flags);
2042 zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp,
2043 struct ucred **credanonp, int *numsecflavors, int **secflavors)
2045 zfsvfs_t *zfsvfs = vfsp->vfs_data;
2048 * If this is regular file system vfsp is the same as
2049 * zfsvfs->z_parent->z_vfs, but if it is snapshot,
2050 * zfsvfs->z_parent->z_vfs represents parent file system
2051 * which we have to use here, because only this file system
2052 * has mnt_export configured.
2054 return (vfs_stdcheckexp(zfsvfs->z_parent->z_vfs, nam, extflagsp,
2055 credanonp, numsecflavors, secflavors));
2058 CTASSERT(SHORT_FID_LEN <= sizeof(struct fid));
2059 CTASSERT(LONG_FID_LEN <= sizeof(struct fid));
2062 zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, vnode_t **vpp)
2064 zfsvfs_t *zfsvfs = vfsp->vfs_data;
2066 uint64_t object = 0;
2067 uint64_t fid_gen = 0;
2077 * On FreeBSD we can get snapshot's mount point or its parent file
2078 * system mount point depending if snapshot is already mounted or not.
2080 if (zfsvfs->z_parent == zfsvfs && fidp->fid_len == LONG_FID_LEN) {
2081 zfid_long_t *zlfid = (zfid_long_t *)fidp;
2082 uint64_t objsetid = 0;
2083 uint64_t setgen = 0;
2085 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
2086 objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i);
2088 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
2089 setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);
2093 err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs);
2099 if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) {
2100 zfid_short_t *zfid = (zfid_short_t *)fidp;
2102 for (i = 0; i < sizeof (zfid->zf_object); i++)
2103 object |= ((uint64_t)zfid->zf_object[i]) << (8 * i);
2105 for (i = 0; i < sizeof (zfid->zf_gen); i++)
2106 fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
2113 * A zero fid_gen means we are in .zfs or the .zfs/snapshot
2114 * directory tree. If the object == zfsvfs->z_shares_dir, then
2115 * we are in the .zfs/shares directory tree.
2117 if ((fid_gen == 0 &&
2118 (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) ||
2119 (zfsvfs->z_shares_dir != 0 && object == zfsvfs->z_shares_dir)) {
2120 *vpp = zfsvfs->z_ctldir;
2121 ASSERT(*vpp != NULL);
2122 if (object == ZFSCTL_INO_SNAPDIR) {
2123 VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL,
2124 0, NULL, NULL, NULL, NULL, NULL) == 0);
2125 } else if (object == zfsvfs->z_shares_dir) {
2126 VERIFY(zfsctl_root_lookup(*vpp, "shares", vpp, NULL,
2127 0, NULL, NULL, NULL, NULL, NULL) == 0);
2132 err = zfs_vnode_lock(*vpp, LK_EXCLUSIVE | LK_RETRY);
2138 gen_mask = -1ULL >> (64 - 8 * i);
2140 dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask);
2141 if (err = zfs_zget(zfsvfs, object, &zp)) {
2145 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen,
2147 zp_gen = zp_gen & gen_mask;
2150 if (zp->z_unlinked || zp_gen != fid_gen) {
2151 dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen);
2159 err = zfs_vnode_lock(*vpp, LK_EXCLUSIVE | LK_RETRY);
2161 vnode_create_vobject(*vpp, zp->z_size, curthread);
2168 * Block out VOPs and close zfsvfs_t::z_os
2170 * Note, if successful, then we return with the 'z_teardown_lock' and
2171 * 'z_teardown_inactive_lock' write held.
2174 zfs_suspend_fs(zfsvfs_t *zfsvfs)
2178 if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0)
2180 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
2186 * Reopen zfsvfs_t::z_os and release VOPs.
2189 zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname)
2193 ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock));
2194 ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock));
2196 err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zfsvfs,
2199 zfsvfs->z_os = NULL;
2202 uint64_t sa_obj = 0;
2205 * Make sure version hasn't changed
2208 err = zfs_get_zplprop(zfsvfs->z_os, ZFS_PROP_VERSION,
2209 &zfsvfs->z_version);
2214 err = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
2215 ZFS_SA_ATTRS, 8, 1, &sa_obj);
2217 if (err && zfsvfs->z_version >= ZPL_VERSION_SA)
2220 if ((err = sa_setup(zfsvfs->z_os, sa_obj,
2221 zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table)) != 0)
2224 if (zfsvfs->z_version >= ZPL_VERSION_SA)
2225 sa_register_update_callback(zfsvfs->z_os,
2228 VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0);
2230 zfs_set_fuid_feature(zfsvfs);
2233 * Attempt to re-establish all the active znodes with
2234 * their dbufs. If a zfs_rezget() fails, then we'll let
2235 * any potential callers discover that via ZFS_ENTER_VERIFY_VP
2236 * when they try to use their znode.
2238 mutex_enter(&zfsvfs->z_znodes_lock);
2239 for (zp = list_head(&zfsvfs->z_all_znodes); zp;
2240 zp = list_next(&zfsvfs->z_all_znodes, zp)) {
2241 (void) zfs_rezget(zp);
2243 mutex_exit(&zfsvfs->z_znodes_lock);
2247 /* release the VOPs */
2248 rw_exit(&zfsvfs->z_teardown_inactive_lock);
2249 rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
2253 * Since we couldn't reopen zfsvfs::z_os, or
2254 * setup the sa framework force unmount this file system.
2256 if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0)
2257 (void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread);
2263 zfs_freevfs(vfs_t *vfsp)
2265 zfsvfs_t *zfsvfs = vfsp->vfs_data;
2269 * If this is a snapshot, we have an extra VFS_HOLD on our parent
2270 * from zfs_mount(). Release it here. If we came through
2271 * zfs_mountroot() instead, we didn't grab an extra hold, so
2272 * skip the VFS_RELE for rootvfs.
2274 if (zfsvfs->z_issnap && (vfsp != rootvfs))
2275 VFS_RELE(zfsvfs->z_parent->z_vfs);
2278 zfsvfs_free(zfsvfs);
2280 atomic_add_32(&zfs_active_fs_count, -1);
2284 static int desiredvnodes_backup;
2288 zfs_vnodes_adjust(void)
2291 int newdesiredvnodes;
2293 desiredvnodes_backup = desiredvnodes;
2296 * We calculate newdesiredvnodes the same way it is done in
2297 * vntblinit(). If it is equal to desiredvnodes, it means that
2298 * it wasn't tuned by the administrator and we can tune it down.
2300 newdesiredvnodes = min(maxproc + cnt.v_page_count / 4, 2 *
2301 vm_kmem_size / (5 * (sizeof(struct vm_object) +
2302 sizeof(struct vnode))));
2303 if (newdesiredvnodes == desiredvnodes)
2304 desiredvnodes = (3 * newdesiredvnodes) / 4;
2309 zfs_vnodes_adjust_back(void)
2313 desiredvnodes = desiredvnodes_backup;
2321 printf("ZFS filesystem version: " ZPL_VERSION_STRING "\n");
2324 * Initialize .zfs directory structures
2329 * Initialize znode cache, vnode ops, etc...
2334 * Reduce number of vnodes. Originally number of vnodes is calculated
2335 * with UFS inode in mind. We reduce it here, because it's too big for
2338 zfs_vnodes_adjust();
2340 dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb);
2348 zfs_vnodes_adjust_back();
2354 return (zfs_active_fs_count != 0);
2358 zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers)
2361 objset_t *os = zfsvfs->z_os;
2364 if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
2367 if (newvers < zfsvfs->z_version)
2370 if (zfs_spa_version_map(newvers) >
2371 spa_version(dmu_objset_spa(zfsvfs->z_os)))
2374 tx = dmu_tx_create(os);
2375 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
2376 if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) {
2377 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
2379 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
2381 error = dmu_tx_assign(tx, TXG_WAIT);
2387 error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
2388 8, 1, &newvers, tx);
2395 if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) {
2398 ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=,
2400 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
2401 DMU_OT_NONE, 0, tx);
2403 error = zap_add(os, MASTER_NODE_OBJ,
2404 ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
2407 VERIFY(0 == sa_set_sa_object(os, sa_obj));
2408 sa_register_update_callback(os, zfs_sa_upgrade);
2411 spa_history_log_internal(LOG_DS_UPGRADE,
2412 dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu",
2413 zfsvfs->z_version, newvers, dmu_objset_id(os));
2417 zfsvfs->z_version = newvers;
2419 zfs_set_fuid_feature(zfsvfs);
2425 * Read a property stored within the master node.
2428 zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value)
2434 * Look up the file system's value for the property. For the
2435 * version property, we look up a slightly different string.
2437 if (prop == ZFS_PROP_VERSION)
2438 pname = ZPL_VERSION_STR;
2440 pname = zfs_prop_to_name(prop);
2443 error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value);
2445 if (error == ENOENT) {
2446 /* No value set, use the default value */
2448 case ZFS_PROP_VERSION:
2449 *value = ZPL_VERSION;
2451 case ZFS_PROP_NORMALIZE:
2452 case ZFS_PROP_UTF8ONLY:
2456 *value = ZFS_CASE_SENSITIVE;
2468 zfsvfs_update_fromname(const char *oldname, const char *newname)
2470 char tmpbuf[MAXPATHLEN];
2475 oldlen = strlen(oldname);
2477 mtx_lock(&mountlist_mtx);
2478 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
2479 fromname = mp->mnt_stat.f_mntfromname;
2480 if (strcmp(fromname, oldname) == 0) {
2481 (void)strlcpy(fromname, newname,
2482 sizeof(mp->mnt_stat.f_mntfromname));
2485 if (strncmp(fromname, oldname, oldlen) == 0 &&
2486 (fromname[oldlen] == '/' || fromname[oldlen] == '@')) {
2487 (void)snprintf(tmpbuf, sizeof(tmpbuf), "%s%s",
2488 newname, fromname + oldlen);
2489 (void)strlcpy(fromname, tmpbuf,
2490 sizeof(mp->mnt_stat.f_mntfromname));
2494 mtx_unlock(&mountlist_mtx);