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 * Portions Copyright 2011 Martin Matuska
24 * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
25 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
28 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2011 by Delphix. All rights reserved.
32 #include <sys/types.h>
33 #include <sys/param.h>
34 #include <sys/errno.h>
37 #include <sys/modctl.h>
42 #include <sys/cmn_err.h>
44 #include <sys/zfs_ioctl.h>
45 #include <sys/zfs_vfsops.h>
46 #include <sys/zfs_znode.h>
49 #include <sys/spa_impl.h>
51 #include <sys/priv_impl.h>
53 #include <sys/dsl_dir.h>
54 #include <sys/dsl_dataset.h>
55 #include <sys/dsl_prop.h>
56 #include <sys/dsl_deleg.h>
57 #include <sys/dmu_objset.h>
59 #include <sys/sunddi.h>
60 #include <sys/sunldi.h>
61 #include <sys/policy.h>
63 #include <sys/nvpair.h>
64 #include <sys/pathname.h>
65 #include <sys/mount.h>
67 #include <sys/fs/zfs.h>
68 #include <sys/zfs_ctldir.h>
69 #include <sys/zfs_dir.h>
70 #include <sys/zfs_onexit.h>
72 #include <sys/dsl_scan.h>
73 #include <sharefs/share.h>
74 #include <sys/dmu_objset.h>
75 #include <sys/fm/util.h>
77 #include <linux/miscdevice.h>
79 #include "zfs_namecheck.h"
81 #include "zfs_deleg.h"
82 #include "zfs_comutil.h"
84 kmutex_t zfsdev_state_lock;
85 list_t zfsdev_state_list;
87 extern void zfs_init(void);
88 extern void zfs_fini(void);
90 typedef int zfs_ioc_func_t(zfs_cmd_t *);
91 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
97 } zfs_ioc_namecheck_t;
100 POOL_CHECK_NONE = 1 << 0,
101 POOL_CHECK_SUSPENDED = 1 << 1,
102 POOL_CHECK_READONLY = 1 << 2
103 } zfs_ioc_poolcheck_t;
105 typedef struct zfs_ioc_vec {
106 zfs_ioc_func_t *zvec_func;
107 zfs_secpolicy_func_t *zvec_secpolicy;
108 zfs_ioc_namecheck_t zvec_namecheck;
109 boolean_t zvec_his_log;
110 zfs_ioc_poolcheck_t zvec_pool_check;
113 /* This array is indexed by zfs_userquota_prop_t */
114 static const char *userquota_perms[] = {
115 ZFS_DELEG_PERM_USERUSED,
116 ZFS_DELEG_PERM_USERQUOTA,
117 ZFS_DELEG_PERM_GROUPUSED,
118 ZFS_DELEG_PERM_GROUPQUOTA,
121 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
122 static int zfs_check_settable(const char *name, nvpair_t *property,
124 static int zfs_check_clearable(char *dataset, nvlist_t *props,
126 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
128 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **);
131 history_str_free(char *buf)
133 kmem_free(buf, HIS_MAX_RECORD_LEN);
137 history_str_get(zfs_cmd_t *zc)
141 if (zc->zc_history == 0)
144 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP | KM_NODEBUG);
145 if (copyinstr((void *)(uintptr_t)zc->zc_history,
146 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
147 history_str_free(buf);
151 buf[HIS_MAX_RECORD_LEN -1] = '\0';
157 * Check to see if the named dataset is currently defined as bootable
160 zfs_is_bootfs(const char *name)
164 if (dmu_objset_hold(name, FTAG, &os) == 0) {
166 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
167 dmu_objset_rele(os, FTAG);
174 * zfs_earlier_version
176 * Return non-zero if the spa version is less than requested version.
179 zfs_earlier_version(const char *name, int version)
183 if (spa_open(name, &spa, FTAG) == 0) {
184 if (spa_version(spa) < version) {
185 spa_close(spa, FTAG);
188 spa_close(spa, FTAG);
194 * zpl_earlier_version
196 * Return TRUE if the ZPL version is less than requested version.
199 zpl_earlier_version(const char *name, int version)
202 boolean_t rc = B_TRUE;
204 if (dmu_objset_hold(name, FTAG, &os) == 0) {
207 if (dmu_objset_type(os) != DMU_OST_ZFS) {
208 dmu_objset_rele(os, FTAG);
211 /* XXX reading from non-owned objset */
212 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
213 rc = zplversion < version;
214 dmu_objset_rele(os, FTAG);
220 zfs_log_history(zfs_cmd_t *zc)
225 if ((buf = history_str_get(zc)) == NULL)
228 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
229 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
230 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
231 spa_close(spa, FTAG);
233 history_str_free(buf);
237 * Policy for top-level read operations (list pools). Requires no privileges,
238 * and can be used in the local zone, as there is no associated dataset.
242 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
248 * Policy for dataset read operations (list children, get statistics). Requires
249 * no privileges, but must be visible in the local zone.
253 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
255 if (INGLOBALZONE(curproc) ||
256 zone_dataset_visible(zc->zc_name, NULL))
263 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
268 * The dataset must be visible by this zone -- check this first
269 * so they don't see EPERM on something they shouldn't know about.
271 if (!INGLOBALZONE(curproc) &&
272 !zone_dataset_visible(dataset, &writable))
275 if (INGLOBALZONE(curproc)) {
277 * If the fs is zoned, only root can access it from the
280 if (secpolicy_zfs(cr) && zoned)
284 * If we are in a local zone, the 'zoned' property must be set.
289 /* must be writable by this zone */
297 zfs_dozonecheck(const char *dataset, cred_t *cr)
301 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
304 return (zfs_dozonecheck_impl(dataset, zoned, cr));
308 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
312 rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
313 if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
314 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
317 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
319 return (zfs_dozonecheck_impl(dataset, zoned, cr));
323 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
327 error = zfs_dozonecheck(name, cr);
329 error = secpolicy_zfs(cr);
331 error = dsl_deleg_access(name, perm, cr);
337 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
338 const char *perm, cred_t *cr)
342 error = zfs_dozonecheck_ds(name, ds, cr);
344 error = secpolicy_zfs(cr);
346 error = dsl_deleg_access_impl(ds, perm, cr);
352 * Policy for setting the security label property.
354 * Returns 0 for success, non-zero for access and other errors.
357 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
360 char ds_hexsl[MAXNAMELEN];
361 bslabel_t ds_sl, new_sl;
362 boolean_t new_default = FALSE;
364 int needed_priv = -1;
367 /* First get the existing dataset label. */
368 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
369 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
373 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
376 /* The label must be translatable */
377 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
381 * In a non-global zone, disallow attempts to set a label that
382 * doesn't match that of the zone; otherwise no other checks
385 if (!INGLOBALZONE(curproc)) {
386 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
392 * For global-zone datasets (i.e., those whose zoned property is
393 * "off", verify that the specified new label is valid for the
396 if (dsl_prop_get_integer(name,
397 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
400 if (zfs_check_global_label(name, strval) != 0)
405 * If the existing dataset label is nondefault, check if the
406 * dataset is mounted (label cannot be changed while mounted).
407 * Get the zfs_sb_t; if there isn't one, then the dataset isn't
408 * mounted (or isn't a dataset, doesn't exist, ...).
410 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
412 static char *setsl_tag = "setsl_tag";
415 * Try to own the dataset; abort if there is any error,
416 * (e.g., already mounted, in use, or other error).
418 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
423 dmu_objset_disown(os, setsl_tag);
426 needed_priv = PRIV_FILE_DOWNGRADE_SL;
430 if (hexstr_to_label(strval, &new_sl) != 0)
433 if (blstrictdom(&ds_sl, &new_sl))
434 needed_priv = PRIV_FILE_DOWNGRADE_SL;
435 else if (blstrictdom(&new_sl, &ds_sl))
436 needed_priv = PRIV_FILE_UPGRADE_SL;
438 /* dataset currently has a default label */
440 needed_priv = PRIV_FILE_UPGRADE_SL;
444 if (needed_priv != -1)
445 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
449 #endif /* HAVE_MLSLABEL */
453 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
459 * Check permissions for special properties.
466 * Disallow setting of 'zoned' from within a local zone.
468 if (!INGLOBALZONE(curproc))
473 if (!INGLOBALZONE(curproc)) {
475 char setpoint[MAXNAMELEN];
477 * Unprivileged users are allowed to modify the
478 * quota on things *under* (ie. contained by)
479 * the thing they own.
481 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
484 if (!zoned || strlen(dsname) <= strlen(setpoint))
489 case ZFS_PROP_MLSLABEL:
490 if (!is_system_labeled())
493 if (nvpair_value_string(propval, &strval) == 0) {
496 err = zfs_set_slabel_policy(dsname, strval, CRED());
503 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
507 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
511 error = zfs_dozonecheck(zc->zc_name, cr);
516 * permission to set permissions will be evaluated later in
517 * dsl_deleg_can_allow()
523 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
525 return (zfs_secpolicy_write_perms(zc->zc_name,
526 ZFS_DELEG_PERM_ROLLBACK, cr));
530 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
539 * Generate the current snapshot name from the given objsetid, then
540 * use that name for the secpolicy/zone checks.
542 cp = strchr(zc->zc_name, '@');
545 error = spa_open(zc->zc_name, &spa, FTAG);
549 dp = spa_get_dsl(spa);
550 rw_enter(&dp->dp_config_rwlock, RW_READER);
551 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
552 rw_exit(&dp->dp_config_rwlock);
553 spa_close(spa, FTAG);
557 dsl_dataset_name(ds, zc->zc_name);
559 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
560 ZFS_DELEG_PERM_SEND, cr);
561 dsl_dataset_rele(ds, FTAG);
566 #ifdef HAVE_SMB_SHARE
568 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
573 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
574 NO_FOLLOW, NULL, &vp)) != 0)
577 /* Now make sure mntpnt and dataset are ZFS */
579 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
580 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
581 zc->zc_name) != 0)) {
587 return (dsl_deleg_access(zc->zc_name,
588 ZFS_DELEG_PERM_SHARE, cr));
590 #endif /* HAVE_SMB_SHARE */
593 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
595 #ifdef HAVE_SMB_SHARE
596 if (!INGLOBALZONE(curproc))
599 if (secpolicy_nfs(cr) == 0) {
602 return (zfs_secpolicy_deleg_share(zc, cr));
606 #endif /* HAVE_SMB_SHARE */
610 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
612 #ifdef HAVE_SMB_SHARE
613 if (!INGLOBALZONE(curproc))
616 if (secpolicy_smb(cr) == 0) {
619 return (zfs_secpolicy_deleg_share(zc, cr));
623 #endif /* HAVE_SMB_SHARE */
627 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
632 * Remove the @bla or /bla from the end of the name to get the parent.
634 (void) strncpy(parent, datasetname, parentsize);
635 cp = strrchr(parent, '@');
639 cp = strrchr(parent, '/');
649 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
653 if ((error = zfs_secpolicy_write_perms(name,
654 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
657 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
661 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
663 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
667 * Destroying snapshots with delegated permissions requires
668 * descendent mount and destroy permissions.
669 * Reassemble the full filesystem@snap name so dsl_deleg_access()
670 * can do the correct permission check.
672 * Since this routine is used when doing a recursive destroy of snapshots
673 * and destroying snapshots requires descendent permissions, a successfull
674 * check of the top level snapshot applies to snapshots of all descendent
677 * The target snapshot may not exist when doing a recursive destroy.
678 * In this case fallback to permissions of the parent dataset.
681 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, cred_t *cr)
686 dsname = kmem_asprintf("%s@%s", zc->zc_name, zc->zc_value);
688 error = zfs_secpolicy_destroy_perms(dsname, cr);
690 error = zfs_secpolicy_destroy_perms(zc->zc_name, cr);
697 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
699 char parentname[MAXNAMELEN];
702 if ((error = zfs_secpolicy_write_perms(from,
703 ZFS_DELEG_PERM_RENAME, cr)) != 0)
706 if ((error = zfs_secpolicy_write_perms(from,
707 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
710 if ((error = zfs_get_parent(to, parentname,
711 sizeof (parentname))) != 0)
714 if ((error = zfs_secpolicy_write_perms(parentname,
715 ZFS_DELEG_PERM_CREATE, cr)) != 0)
718 if ((error = zfs_secpolicy_write_perms(parentname,
719 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
726 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
728 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
732 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
734 char parentname[MAXNAMELEN];
738 error = zfs_secpolicy_write_perms(zc->zc_name,
739 ZFS_DELEG_PERM_PROMOTE, cr);
743 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
746 dsl_dataset_t *pclone = NULL;
748 dd = clone->os_dsl_dataset->ds_dir;
750 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
751 error = dsl_dataset_hold_obj(dd->dd_pool,
752 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
753 rw_exit(&dd->dd_pool->dp_config_rwlock);
755 dmu_objset_rele(clone, FTAG);
759 error = zfs_secpolicy_write_perms(zc->zc_name,
760 ZFS_DELEG_PERM_MOUNT, cr);
762 dsl_dataset_name(pclone, parentname);
763 dmu_objset_rele(clone, FTAG);
764 dsl_dataset_rele(pclone, FTAG);
766 error = zfs_secpolicy_write_perms(parentname,
767 ZFS_DELEG_PERM_PROMOTE, cr);
773 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
777 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
778 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
781 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
782 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
785 return (zfs_secpolicy_write_perms(zc->zc_name,
786 ZFS_DELEG_PERM_CREATE, cr));
790 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
792 return (zfs_secpolicy_write_perms(name,
793 ZFS_DELEG_PERM_SNAPSHOT, cr));
797 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
800 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
804 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
806 char parentname[MAXNAMELEN];
809 if ((error = zfs_get_parent(zc->zc_name, parentname,
810 sizeof (parentname))) != 0)
813 if (zc->zc_value[0] != '\0') {
814 if ((error = zfs_secpolicy_write_perms(zc->zc_value,
815 ZFS_DELEG_PERM_CLONE, cr)) != 0)
819 if ((error = zfs_secpolicy_write_perms(parentname,
820 ZFS_DELEG_PERM_CREATE, cr)) != 0)
823 error = zfs_secpolicy_write_perms(parentname,
824 ZFS_DELEG_PERM_MOUNT, cr);
830 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
831 * SYS_CONFIG privilege, which is not available in a local zone.
835 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
837 if (secpolicy_sys_config(cr, B_FALSE) != 0)
844 * Policy for object to name lookups.
848 zfs_secpolicy_diff(zfs_cmd_t *zc, cred_t *cr)
852 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
855 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
860 * Policy for fault injection. Requires all privileges.
864 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
866 return (secpolicy_zinject(cr));
870 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
872 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
874 if (prop == ZPROP_INVAL) {
875 if (!zfs_prop_user(zc->zc_value))
877 return (zfs_secpolicy_write_perms(zc->zc_name,
878 ZFS_DELEG_PERM_USERPROP, cr));
880 return (zfs_secpolicy_setprop(zc->zc_name, prop,
886 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
888 int err = zfs_secpolicy_read(zc, cr);
892 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
895 if (zc->zc_value[0] == 0) {
897 * They are asking about a posix uid/gid. If it's
898 * themself, allow it.
900 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
901 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
902 if (zc->zc_guid == crgetuid(cr))
905 if (groupmember(zc->zc_guid, cr))
910 return (zfs_secpolicy_write_perms(zc->zc_name,
911 userquota_perms[zc->zc_objset_type], cr));
915 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
917 int err = zfs_secpolicy_read(zc, cr);
921 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
924 return (zfs_secpolicy_write_perms(zc->zc_name,
925 userquota_perms[zc->zc_objset_type], cr));
929 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
931 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
936 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
938 return (zfs_secpolicy_write_perms(zc->zc_name,
939 ZFS_DELEG_PERM_HOLD, cr));
943 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
945 return (zfs_secpolicy_write_perms(zc->zc_name,
946 ZFS_DELEG_PERM_RELEASE, cr));
950 * Policy for allowing temporary snapshots to be taken or released
953 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, cred_t *cr)
956 * A temporary snapshot is the same as a snapshot,
957 * hold, destroy and release all rolled into one.
958 * Delegated diff alone is sufficient that we allow this.
962 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
963 ZFS_DELEG_PERM_DIFF, cr)) == 0)
966 error = zfs_secpolicy_snapshot(zc, cr);
968 error = zfs_secpolicy_hold(zc, cr);
970 error = zfs_secpolicy_release(zc, cr);
972 error = zfs_secpolicy_destroy(zc, cr);
977 * Returns the nvlist as specified by the user in the zfs_cmd_t.
980 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
984 nvlist_t *list = NULL;
987 * Read in and unpack the user-supplied nvlist.
992 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
994 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
996 kmem_free(packed, size);
1000 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1001 kmem_free(packed, size);
1005 kmem_free(packed, size);
1012 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
1016 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1018 if (size > zc->zc_nvlist_dst_size) {
1019 nvpair_t *more_errors;
1022 if (zc->zc_nvlist_dst_size < 1024)
1025 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
1026 more_errors = nvlist_prev_nvpair(*errors, NULL);
1029 nvpair_t *pair = nvlist_prev_nvpair(*errors,
1031 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
1033 VERIFY(nvlist_size(*errors, &size,
1034 NV_ENCODE_NATIVE) == 0);
1035 } while (size > zc->zc_nvlist_dst_size);
1037 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
1038 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
1039 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1040 ASSERT(size <= zc->zc_nvlist_dst_size);
1047 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1049 char *packed = NULL;
1053 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
1055 if (size > zc->zc_nvlist_dst_size) {
1058 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1059 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
1061 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1062 size, zc->zc_iflags) != 0)
1064 kmem_free(packed, size);
1067 zc->zc_nvlist_dst_size = size;
1072 get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
1077 error = dmu_objset_hold(dsname, FTAG, &os);
1080 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1081 dmu_objset_rele(os, FTAG);
1085 mutex_enter(&os->os_user_ptr_lock);
1086 *zsbp = dmu_objset_get_user(os);
1087 if (*zsbp && (*zsbp)->z_sb) {
1088 atomic_inc(&((*zsbp)->z_sb->s_active));
1092 mutex_exit(&os->os_user_ptr_lock);
1093 dmu_objset_rele(os, FTAG);
1098 * Find a zfs_sb_t for a mounted filesystem, or create our own, in which
1099 * case its z_sb will be NULL, and it will be opened as the owner.
1102 zfs_sb_hold(const char *name, void *tag, zfs_sb_t **zsbp, boolean_t writer)
1106 if (get_zfs_sb(name, zsbp) != 0)
1107 error = zfs_sb_create(name, zsbp);
1109 rrw_enter(&(*zsbp)->z_teardown_lock, (writer) ? RW_WRITER :
1111 if ((*zsbp)->z_unmounted) {
1113 * XXX we could probably try again, since the unmounting
1114 * thread should be just about to disassociate the
1115 * objset from the zfsvfs.
1117 rrw_exit(&(*zsbp)->z_teardown_lock, tag);
1125 zfs_sb_rele(zfs_sb_t *zsb, void *tag)
1127 rrw_exit(&zsb->z_teardown_lock, tag);
1130 deactivate_super(zsb->z_sb);
1132 dmu_objset_disown(zsb->z_os, zsb);
1138 zfs_ioc_pool_create(zfs_cmd_t *zc)
1141 nvlist_t *config, *props = NULL;
1142 nvlist_t *rootprops = NULL;
1143 nvlist_t *zplprops = NULL;
1146 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1147 zc->zc_iflags, &config)))
1150 if (zc->zc_nvlist_src_size != 0 && (error =
1151 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1152 zc->zc_iflags, &props))) {
1153 nvlist_free(config);
1158 nvlist_t *nvl = NULL;
1159 uint64_t version = SPA_VERSION;
1161 (void) nvlist_lookup_uint64(props,
1162 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1163 if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1165 goto pool_props_bad;
1167 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1169 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1171 nvlist_free(config);
1175 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1177 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1178 error = zfs_fill_zplprops_root(version, rootprops,
1181 goto pool_props_bad;
1184 buf = history_str_get(zc);
1186 error = spa_create(zc->zc_name, config, props, buf, zplprops);
1189 * Set the remaining root properties
1191 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1192 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1193 (void) spa_destroy(zc->zc_name);
1196 history_str_free(buf);
1199 nvlist_free(rootprops);
1200 nvlist_free(zplprops);
1201 nvlist_free(config);
1208 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1211 zfs_log_history(zc);
1212 error = spa_destroy(zc->zc_name);
1214 zvol_remove_minors(zc->zc_name);
1219 zfs_ioc_pool_import(zfs_cmd_t *zc)
1221 nvlist_t *config, *props = NULL;
1225 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1226 zc->zc_iflags, &config)) != 0)
1229 if (zc->zc_nvlist_src_size != 0 && (error =
1230 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1231 zc->zc_iflags, &props))) {
1232 nvlist_free(config);
1236 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1237 guid != zc->zc_guid)
1240 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1242 if (zc->zc_nvlist_dst != 0) {
1245 if ((err = put_nvlist(zc, config)) != 0)
1250 zvol_create_minors(zc->zc_name);
1252 nvlist_free(config);
1261 zfs_ioc_pool_export(zfs_cmd_t *zc)
1264 boolean_t force = (boolean_t)zc->zc_cookie;
1265 boolean_t hardforce = (boolean_t)zc->zc_guid;
1267 zfs_log_history(zc);
1268 error = spa_export(zc->zc_name, NULL, force, hardforce);
1270 zvol_remove_minors(zc->zc_name);
1275 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1280 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1283 error = put_nvlist(zc, configs);
1285 nvlist_free(configs);
1291 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1297 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1298 sizeof (zc->zc_value));
1300 if (config != NULL) {
1301 ret = put_nvlist(zc, config);
1302 nvlist_free(config);
1305 * The config may be present even if 'error' is non-zero.
1306 * In this case we return success, and preserve the real errno
1309 zc->zc_cookie = error;
1318 * Try to import the given pool, returning pool stats as appropriate so that
1319 * user land knows which devices are available and overall pool health.
1322 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1324 nvlist_t *tryconfig, *config;
1327 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1328 zc->zc_iflags, &tryconfig)) != 0)
1331 config = spa_tryimport(tryconfig);
1333 nvlist_free(tryconfig);
1338 error = put_nvlist(zc, config);
1339 nvlist_free(config);
1346 * zc_name name of the pool
1347 * zc_cookie scan func (pool_scan_func_t)
1350 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1355 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1358 if (zc->zc_cookie == POOL_SCAN_NONE)
1359 error = spa_scan_stop(spa);
1361 error = spa_scan(spa, zc->zc_cookie);
1363 spa_close(spa, FTAG);
1369 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1374 error = spa_open(zc->zc_name, &spa, FTAG);
1377 spa_close(spa, FTAG);
1383 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1388 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1391 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1392 spa_close(spa, FTAG);
1396 spa_upgrade(spa, zc->zc_cookie);
1397 spa_close(spa, FTAG);
1403 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1410 if ((size = zc->zc_history_len) == 0)
1413 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1416 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1417 spa_close(spa, FTAG);
1421 hist_buf = vmem_alloc(size, KM_SLEEP);
1422 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1423 &zc->zc_history_len, hist_buf)) == 0) {
1424 error = ddi_copyout(hist_buf,
1425 (void *)(uintptr_t)zc->zc_history,
1426 zc->zc_history_len, zc->zc_iflags);
1429 spa_close(spa, FTAG);
1430 vmem_free(hist_buf, size);
1435 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1440 error = spa_open(zc->zc_name, &spa, FTAG);
1442 error = spa_change_guid(spa);
1443 spa_close(spa, FTAG);
1449 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1453 if ((error = dsl_dsobj_to_dsname(zc->zc_name,zc->zc_obj,zc->zc_value)))
1461 * zc_name name of filesystem
1462 * zc_obj object to find
1465 * zc_value name of object
1468 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1473 /* XXX reading from objset not owned */
1474 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1476 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1477 dmu_objset_rele(os, FTAG);
1480 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1481 sizeof (zc->zc_value));
1482 dmu_objset_rele(os, FTAG);
1489 * zc_name name of filesystem
1490 * zc_obj object to find
1493 * zc_stat stats on object
1494 * zc_value path to object
1497 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1502 /* XXX reading from objset not owned */
1503 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1505 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1506 dmu_objset_rele(os, FTAG);
1509 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1510 sizeof (zc->zc_value));
1511 dmu_objset_rele(os, FTAG);
1517 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1521 nvlist_t *config, **l2cache, **spares;
1522 uint_t nl2cache = 0, nspares = 0;
1524 error = spa_open(zc->zc_name, &spa, FTAG);
1528 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1529 zc->zc_iflags, &config);
1530 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1531 &l2cache, &nl2cache);
1533 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1537 * A root pool with concatenated devices is not supported.
1538 * Thus, can not add a device to a root pool.
1540 * Intent log device can not be added to a rootpool because
1541 * during mountroot, zil is replayed, a seperated log device
1542 * can not be accessed during the mountroot time.
1544 * l2cache and spare devices are ok to be added to a rootpool.
1546 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1547 nvlist_free(config);
1548 spa_close(spa, FTAG);
1553 error = spa_vdev_add(spa, config);
1554 nvlist_free(config);
1556 spa_close(spa, FTAG);
1562 * zc_name name of the pool
1563 * zc_nvlist_conf nvlist of devices to remove
1564 * zc_cookie to stop the remove?
1567 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1572 error = spa_open(zc->zc_name, &spa, FTAG);
1575 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1576 spa_close(spa, FTAG);
1581 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1585 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1587 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1589 switch (zc->zc_cookie) {
1590 case VDEV_STATE_ONLINE:
1591 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1594 case VDEV_STATE_OFFLINE:
1595 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1598 case VDEV_STATE_FAULTED:
1599 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1600 zc->zc_obj != VDEV_AUX_EXTERNAL)
1601 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1603 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1606 case VDEV_STATE_DEGRADED:
1607 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1608 zc->zc_obj != VDEV_AUX_EXTERNAL)
1609 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1611 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1617 zc->zc_cookie = newstate;
1618 spa_close(spa, FTAG);
1623 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1626 int replacing = zc->zc_cookie;
1630 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1633 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1634 zc->zc_iflags, &config)) == 0) {
1635 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1636 nvlist_free(config);
1639 spa_close(spa, FTAG);
1644 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1649 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1652 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1654 spa_close(spa, FTAG);
1659 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1662 nvlist_t *config, *props = NULL;
1664 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1666 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1670 zc->zc_iflags, &config))) {
1671 spa_close(spa, FTAG);
1675 if (zc->zc_nvlist_src_size != 0 && (error =
1676 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1677 zc->zc_iflags, &props))) {
1678 spa_close(spa, FTAG);
1679 nvlist_free(config);
1683 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1685 spa_close(spa, FTAG);
1687 nvlist_free(config);
1694 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1697 char *path = zc->zc_value;
1698 uint64_t guid = zc->zc_guid;
1701 error = spa_open(zc->zc_name, &spa, FTAG);
1705 error = spa_vdev_setpath(spa, guid, path);
1706 spa_close(spa, FTAG);
1711 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1714 char *fru = zc->zc_value;
1715 uint64_t guid = zc->zc_guid;
1718 error = spa_open(zc->zc_name, &spa, FTAG);
1722 error = spa_vdev_setfru(spa, guid, fru);
1723 spa_close(spa, FTAG);
1728 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1733 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1735 if (zc->zc_nvlist_dst != 0 &&
1736 (error = dsl_prop_get_all(os, &nv)) == 0) {
1737 dmu_objset_stats(os, nv);
1739 * NB: zvol_get_stats() will read the objset contents,
1740 * which we aren't supposed to do with a
1741 * DS_MODE_USER hold, because it could be
1742 * inconsistent. So this is a bit of a workaround...
1743 * XXX reading with out owning
1745 if (!zc->zc_objset_stats.dds_inconsistent) {
1746 if (dmu_objset_type(os) == DMU_OST_ZVOL)
1747 error = zvol_get_stats(os, nv);
1750 error = put_nvlist(zc, nv);
1759 * zc_name name of filesystem
1760 * zc_nvlist_dst_size size of buffer for property nvlist
1763 * zc_objset_stats stats
1764 * zc_nvlist_dst property nvlist
1765 * zc_nvlist_dst_size size of property nvlist
1768 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1770 objset_t *os = NULL;
1773 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1776 error = zfs_ioc_objset_stats_impl(zc, os);
1778 dmu_objset_rele(os, FTAG);
1785 * zc_name name of filesystem
1786 * zc_nvlist_dst_size size of buffer for property nvlist
1789 * zc_nvlist_dst received property nvlist
1790 * zc_nvlist_dst_size size of received property nvlist
1792 * Gets received properties (distinct from local properties on or after
1793 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1794 * local property values.
1797 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1799 objset_t *os = NULL;
1803 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1807 * Without this check, we would return local property values if the
1808 * caller has not already received properties on or after
1809 * SPA_VERSION_RECVD_PROPS.
1811 if (!dsl_prop_get_hasrecvd(os)) {
1812 dmu_objset_rele(os, FTAG);
1816 if (zc->zc_nvlist_dst != 0 &&
1817 (error = dsl_prop_get_received(os, &nv)) == 0) {
1818 error = put_nvlist(zc, nv);
1822 dmu_objset_rele(os, FTAG);
1827 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1833 * zfs_get_zplprop() will either find a value or give us
1834 * the default value (if there is one).
1836 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1838 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1844 * zc_name name of filesystem
1845 * zc_nvlist_dst_size size of buffer for zpl property nvlist
1848 * zc_nvlist_dst zpl property nvlist
1849 * zc_nvlist_dst_size size of zpl property nvlist
1852 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1857 /* XXX reading without owning */
1858 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1861 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1864 * NB: nvl_add_zplprop() will read the objset contents,
1865 * which we aren't supposed to do with a DS_MODE_USER
1866 * hold, because it could be inconsistent.
1868 if (zc->zc_nvlist_dst != 0 &&
1869 !zc->zc_objset_stats.dds_inconsistent &&
1870 dmu_objset_type(os) == DMU_OST_ZFS) {
1873 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1874 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1875 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1876 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1877 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1878 err = put_nvlist(zc, nv);
1883 dmu_objset_rele(os, FTAG);
1888 dataset_name_hidden(const char *name)
1891 * Skip over datasets that are not visible in this zone,
1892 * internal datasets (which have a $ in their name), and
1893 * temporary datasets (which have a % in their name).
1895 if (strchr(name, '$') != NULL)
1897 if (strchr(name, '%') != NULL)
1899 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1906 * zc_name name of filesystem
1907 * zc_cookie zap cursor
1908 * zc_nvlist_dst_size size of buffer for property nvlist
1911 * zc_name name of next filesystem
1912 * zc_cookie zap cursor
1913 * zc_objset_stats stats
1914 * zc_nvlist_dst property nvlist
1915 * zc_nvlist_dst_size size of property nvlist
1918 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1923 size_t orig_len = strlen(zc->zc_name);
1926 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
1927 if (error == ENOENT)
1932 p = strrchr(zc->zc_name, '/');
1933 if (p == NULL || p[1] != '\0')
1934 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1935 p = zc->zc_name + strlen(zc->zc_name);
1938 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
1939 * but is not declared void because its called by dmu_objset_find().
1941 if (zc->zc_cookie == 0) {
1942 uint64_t cookie = 0;
1943 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1945 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) {
1946 if (!dataset_name_hidden(zc->zc_name))
1947 (void) dmu_objset_prefetch(zc->zc_name, NULL);
1952 error = dmu_dir_list_next(os,
1953 sizeof (zc->zc_name) - (p - zc->zc_name), p,
1954 NULL, &zc->zc_cookie);
1955 if (error == ENOENT)
1957 } while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1958 !(zc->zc_iflags & FKIOCTL));
1959 dmu_objset_rele(os, FTAG);
1962 * If it's an internal dataset (ie. with a '$' in its name),
1963 * don't try to get stats for it, otherwise we'll return ENOENT.
1965 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1966 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1967 if (error == ENOENT) {
1968 /* We lost a race with destroy, get the next one. */
1969 zc->zc_name[orig_len] = '\0';
1978 * zc_name name of filesystem
1979 * zc_cookie zap cursor
1980 * zc_nvlist_dst_size size of buffer for property nvlist
1983 * zc_name name of next snapshot
1984 * zc_objset_stats stats
1985 * zc_nvlist_dst property nvlist
1986 * zc_nvlist_dst_size size of property nvlist
1989 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1995 if (zc->zc_cookie == 0 && !zc->zc_simple)
1996 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1997 NULL, DS_FIND_SNAPSHOTS);
1999 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2001 return (error == ENOENT ? ESRCH : error);
2004 * A dataset name of maximum length cannot have any snapshots,
2005 * so exit immediately.
2007 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2008 dmu_objset_rele(os, FTAG);
2012 error = dmu_snapshot_list_next(os,
2013 sizeof (zc->zc_name) - strlen(zc->zc_name),
2014 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2017 if (error == 0 && !zc->zc_simple) {
2019 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2022 * Since we probably don't have a hold on this snapshot,
2023 * it's possible that the objsetid could have been destroyed
2024 * and reused for a new objset. It's OK if this happens during
2025 * a zfs send operation, since the new createtxg will be
2026 * beyond the range we're interested in.
2028 rw_enter(&dp->dp_config_rwlock, RW_READER);
2029 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2030 rw_exit(&dp->dp_config_rwlock);
2032 if (error == ENOENT) {
2033 /* Racing with destroy, get the next one. */
2034 *strchr(zc->zc_name, '@') = '\0';
2035 dmu_objset_rele(os, FTAG);
2041 error = dmu_objset_from_ds(ds, &ossnap);
2043 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2044 dsl_dataset_rele(ds, FTAG);
2046 } else if (error == ENOENT) {
2050 dmu_objset_rele(os, FTAG);
2051 /* if we failed, undo the @ that we tacked on to zc_name */
2053 *strchr(zc->zc_name, '@') = '\0';
2058 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2060 const char *propname = nvpair_name(pair);
2062 unsigned int vallen;
2065 zfs_userquota_prop_t type;
2071 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2073 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2074 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2080 * A correctly constructed propname is encoded as
2081 * userquota@<rid>-<domain>.
2083 if ((dash = strchr(propname, '-')) == NULL ||
2084 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2093 err = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
2095 err = zfs_set_userquota(zsb, type, domain, rid, quota);
2096 zfs_sb_rele(zsb, FTAG);
2103 * If the named property is one that has a special function to set its value,
2104 * return 0 on success and a positive error code on failure; otherwise if it is
2105 * not one of the special properties handled by this function, return -1.
2107 * XXX: It would be better for callers of the property interface if we handled
2108 * these special cases in dsl_prop.c (in the dsl layer).
2111 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2114 const char *propname = nvpair_name(pair);
2115 zfs_prop_t prop = zfs_name_to_prop(propname);
2119 if (prop == ZPROP_INVAL) {
2120 if (zfs_prop_userquota(propname))
2121 return (zfs_prop_set_userquota(dsname, pair));
2125 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2127 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2128 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2132 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2135 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2138 case ZFS_PROP_QUOTA:
2139 err = dsl_dir_set_quota(dsname, source, intval);
2141 case ZFS_PROP_REFQUOTA:
2142 err = dsl_dataset_set_quota(dsname, source, intval);
2144 case ZFS_PROP_RESERVATION:
2145 err = dsl_dir_set_reservation(dsname, source, intval);
2147 case ZFS_PROP_REFRESERVATION:
2148 err = dsl_dataset_set_reservation(dsname, source, intval);
2150 case ZFS_PROP_VOLSIZE:
2151 err = zvol_set_volsize(dsname, intval);
2153 case ZFS_PROP_VERSION:
2157 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2160 err = zfs_set_version(zsb, intval);
2161 zfs_sb_rele(zsb, FTAG);
2163 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2166 zc = kmem_zalloc(sizeof (zfs_cmd_t),
2167 KM_SLEEP | KM_NODEBUG);
2168 (void) strcpy(zc->zc_name, dsname);
2169 (void) zfs_ioc_userspace_upgrade(zc);
2170 kmem_free(zc, sizeof (zfs_cmd_t));
2183 * This function is best effort. If it fails to set any of the given properties,
2184 * it continues to set as many as it can and returns the first error
2185 * encountered. If the caller provides a non-NULL errlist, it also gives the
2186 * complete list of names of all the properties it failed to set along with the
2187 * corresponding error numbers. The caller is responsible for freeing the
2190 * If every property is set successfully, zero is returned and the list pointed
2191 * at by errlist is NULL.
2194 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2202 nvlist_t *genericnvl;
2206 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2207 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2208 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2212 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2213 const char *propname = nvpair_name(pair);
2214 zfs_prop_t prop = zfs_name_to_prop(propname);
2217 /* decode the property value */
2219 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2221 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2222 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2227 /* Validate value type */
2228 if (err == 0 && prop == ZPROP_INVAL) {
2229 if (zfs_prop_user(propname)) {
2230 if (nvpair_type(propval) != DATA_TYPE_STRING)
2232 } else if (zfs_prop_userquota(propname)) {
2233 if (nvpair_type(propval) !=
2234 DATA_TYPE_UINT64_ARRAY)
2237 } else if (err == 0) {
2238 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2239 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2241 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2244 VERIFY(nvpair_value_uint64(propval,
2247 switch (zfs_prop_get_type(prop)) {
2248 case PROP_TYPE_NUMBER:
2250 case PROP_TYPE_STRING:
2253 case PROP_TYPE_INDEX:
2254 if (zfs_prop_index_to_string(prop,
2255 intval, &unused) != 0)
2260 "unknown property type");
2267 /* Validate permissions */
2269 err = zfs_check_settable(dsname, pair, CRED());
2272 err = zfs_prop_set_special(dsname, source, pair);
2275 * For better performance we build up a list of
2276 * properties to set in a single transaction.
2278 err = nvlist_add_nvpair(genericnvl, pair);
2279 } else if (err != 0 && nvl != retrynvl) {
2281 * This may be a spurious error caused by
2282 * receiving quota and reservation out of order.
2283 * Try again in a second pass.
2285 err = nvlist_add_nvpair(retrynvl, pair);
2290 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2293 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2298 if (!nvlist_empty(genericnvl) &&
2299 dsl_props_set(dsname, source, genericnvl) != 0) {
2301 * If this fails, we still want to set as many properties as we
2302 * can, so try setting them individually.
2305 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2306 const char *propname = nvpair_name(pair);
2310 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2312 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2313 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2317 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2318 VERIFY(nvpair_value_string(propval,
2320 err = dsl_prop_set(dsname, propname, source, 1,
2321 strlen(strval) + 1, strval);
2323 VERIFY(nvpair_value_uint64(propval,
2325 err = dsl_prop_set(dsname, propname, source, 8,
2330 VERIFY(nvlist_add_int32(errors, propname,
2335 nvlist_free(genericnvl);
2336 nvlist_free(retrynvl);
2338 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2339 nvlist_free(errors);
2342 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2345 if (errlist == NULL)
2346 nvlist_free(errors);
2354 * Check that all the properties are valid user properties.
2357 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2359 nvpair_t *pair = NULL;
2362 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2363 const char *propname = nvpair_name(pair);
2366 if (!zfs_prop_user(propname) ||
2367 nvpair_type(pair) != DATA_TYPE_STRING)
2370 if ((error = zfs_secpolicy_write_perms(fsname,
2371 ZFS_DELEG_PERM_USERPROP, CRED())))
2374 if (strlen(propname) >= ZAP_MAXNAMELEN)
2375 return (ENAMETOOLONG);
2377 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2378 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2385 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2389 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2392 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2393 if (nvlist_exists(skipped, nvpair_name(pair)))
2396 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2401 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2405 nvlist_t *cleared_props = NULL;
2406 props_skip(props, skipped, &cleared_props);
2407 if (!nvlist_empty(cleared_props)) {
2409 * Acts on local properties until the dataset has received
2410 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2412 zprop_source_t flags = (ZPROP_SRC_NONE |
2413 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2414 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2416 nvlist_free(cleared_props);
2422 * zc_name name of filesystem
2423 * zc_value name of property to set
2424 * zc_nvlist_src{_size} nvlist of properties to apply
2425 * zc_cookie received properties flag
2428 * zc_nvlist_dst{_size} error for each unapplied received property
2431 zfs_ioc_set_prop(zfs_cmd_t *zc)
2434 boolean_t received = zc->zc_cookie;
2435 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2437 nvlist_t *errors = NULL;
2440 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2441 zc->zc_iflags, &nvl)) != 0)
2445 nvlist_t *origprops;
2448 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2449 if (dsl_prop_get_received(os, &origprops) == 0) {
2450 (void) clear_received_props(os,
2451 zc->zc_name, origprops, nvl);
2452 nvlist_free(origprops);
2455 dsl_prop_set_hasrecvd(os);
2456 dmu_objset_rele(os, FTAG);
2460 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2462 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2463 (void) put_nvlist(zc, errors);
2466 nvlist_free(errors);
2473 * zc_name name of filesystem
2474 * zc_value name of property to inherit
2475 * zc_cookie revert to received value if TRUE
2480 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2482 const char *propname = zc->zc_value;
2483 zfs_prop_t prop = zfs_name_to_prop(propname);
2484 boolean_t received = zc->zc_cookie;
2485 zprop_source_t source = (received
2486 ? ZPROP_SRC_NONE /* revert to received value, if any */
2487 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2496 * zfs_prop_set_special() expects properties in the form of an
2497 * nvpair with type info.
2499 if (prop == ZPROP_INVAL) {
2500 if (!zfs_prop_user(propname))
2503 type = PROP_TYPE_STRING;
2504 } else if (prop == ZFS_PROP_VOLSIZE ||
2505 prop == ZFS_PROP_VERSION) {
2508 type = zfs_prop_get_type(prop);
2511 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2514 case PROP_TYPE_STRING:
2515 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2517 case PROP_TYPE_NUMBER:
2518 case PROP_TYPE_INDEX:
2519 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2526 pair = nvlist_next_nvpair(dummy, NULL);
2527 err = zfs_prop_set_special(zc->zc_name, source, pair);
2530 return (err); /* special property already handled */
2533 * Only check this in the non-received case. We want to allow
2534 * 'inherit -S' to revert non-inheritable properties like quota
2535 * and reservation to the received or default values even though
2536 * they are not considered inheritable.
2538 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2542 /* the property name has been validated by zfs_secpolicy_inherit() */
2543 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2547 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2554 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2555 zc->zc_iflags, &props)))
2559 * If the only property is the configfile, then just do a spa_lookup()
2560 * to handle the faulted case.
2562 pair = nvlist_next_nvpair(props, NULL);
2563 if (pair != NULL && strcmp(nvpair_name(pair),
2564 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2565 nvlist_next_nvpair(props, pair) == NULL) {
2566 mutex_enter(&spa_namespace_lock);
2567 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2568 spa_configfile_set(spa, props, B_FALSE);
2569 spa_config_sync(spa, B_FALSE, B_TRUE);
2571 mutex_exit(&spa_namespace_lock);
2578 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2583 error = spa_prop_set(spa, props);
2586 spa_close(spa, FTAG);
2592 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2596 nvlist_t *nvp = NULL;
2598 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2600 * If the pool is faulted, there may be properties we can still
2601 * get (such as altroot and cachefile), so attempt to get them
2604 mutex_enter(&spa_namespace_lock);
2605 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2606 error = spa_prop_get(spa, &nvp);
2607 mutex_exit(&spa_namespace_lock);
2609 error = spa_prop_get(spa, &nvp);
2610 spa_close(spa, FTAG);
2613 if (error == 0 && zc->zc_nvlist_dst != 0)
2614 error = put_nvlist(zc, nvp);
2624 * zc_name name of volume
2629 zfs_ioc_create_minor(zfs_cmd_t *zc)
2631 return (zvol_create_minor(zc->zc_name));
2636 * zc_name name of volume
2641 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2643 return (zvol_remove_minor(zc->zc_name));
2648 * zc_name name of filesystem
2649 * zc_nvlist_src{_size} nvlist of delegated permissions
2650 * zc_perm_action allow/unallow flag
2655 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2658 nvlist_t *fsaclnv = NULL;
2660 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2661 zc->zc_iflags, &fsaclnv)) != 0)
2665 * Verify nvlist is constructed correctly
2667 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2668 nvlist_free(fsaclnv);
2673 * If we don't have PRIV_SYS_MOUNT, then validate
2674 * that user is allowed to hand out each permission in
2678 error = secpolicy_zfs(CRED());
2680 if (zc->zc_perm_action == B_FALSE) {
2681 error = dsl_deleg_can_allow(zc->zc_name,
2684 error = dsl_deleg_can_unallow(zc->zc_name,
2690 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2692 nvlist_free(fsaclnv);
2698 * zc_name name of filesystem
2701 * zc_nvlist_src{_size} nvlist of delegated permissions
2704 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2709 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2710 error = put_nvlist(zc, nvp);
2719 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2721 zfs_creat_t *zct = arg;
2723 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2726 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2730 * createprops list of properties requested by creator
2731 * default_zplver zpl version to use if unspecified in createprops
2732 * fuids_ok fuids allowed in this version of the spa?
2733 * os parent objset pointer (NULL if root fs)
2736 * zplprops values for the zplprops we attach to the master node object
2737 * is_ci true if requested file system will be purely case-insensitive
2739 * Determine the settings for utf8only, normalization and
2740 * casesensitivity. Specific values may have been requested by the
2741 * creator and/or we can inherit values from the parent dataset. If
2742 * the file system is of too early a vintage, a creator can not
2743 * request settings for these properties, even if the requested
2744 * setting is the default value. We don't actually want to create dsl
2745 * properties for these, so remove them from the source nvlist after
2749 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2750 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2751 nvlist_t *zplprops, boolean_t *is_ci)
2753 uint64_t sense = ZFS_PROP_UNDEFINED;
2754 uint64_t norm = ZFS_PROP_UNDEFINED;
2755 uint64_t u8 = ZFS_PROP_UNDEFINED;
2758 ASSERT(zplprops != NULL);
2761 * Pull out creator prop choices, if any.
2764 (void) nvlist_lookup_uint64(createprops,
2765 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2766 (void) nvlist_lookup_uint64(createprops,
2767 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2768 (void) nvlist_remove_all(createprops,
2769 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2770 (void) nvlist_lookup_uint64(createprops,
2771 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2772 (void) nvlist_remove_all(createprops,
2773 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2774 (void) nvlist_lookup_uint64(createprops,
2775 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2776 (void) nvlist_remove_all(createprops,
2777 zfs_prop_to_name(ZFS_PROP_CASE));
2781 * If the zpl version requested is whacky or the file system
2782 * or pool is version is too "young" to support normalization
2783 * and the creator tried to set a value for one of the props,
2786 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2787 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2788 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2789 (zplver < ZPL_VERSION_NORMALIZATION &&
2790 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2791 sense != ZFS_PROP_UNDEFINED)))
2795 * Put the version in the zplprops
2797 VERIFY(nvlist_add_uint64(zplprops,
2798 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2800 if (norm == ZFS_PROP_UNDEFINED &&
2801 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
2803 VERIFY(nvlist_add_uint64(zplprops,
2804 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2807 * If we're normalizing, names must always be valid UTF-8 strings.
2811 if (u8 == ZFS_PROP_UNDEFINED &&
2812 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
2814 VERIFY(nvlist_add_uint64(zplprops,
2815 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2817 if (sense == ZFS_PROP_UNDEFINED &&
2818 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
2820 VERIFY(nvlist_add_uint64(zplprops,
2821 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2824 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2830 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2831 nvlist_t *zplprops, boolean_t *is_ci)
2833 boolean_t fuids_ok, sa_ok;
2834 uint64_t zplver = ZPL_VERSION;
2835 objset_t *os = NULL;
2836 char parentname[MAXNAMELEN];
2842 (void) strlcpy(parentname, dataset, sizeof (parentname));
2843 cp = strrchr(parentname, '/');
2847 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2850 spa_vers = spa_version(spa);
2851 spa_close(spa, FTAG);
2853 zplver = zfs_zpl_version_map(spa_vers);
2854 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2855 sa_ok = (zplver >= ZPL_VERSION_SA);
2858 * Open parent object set so we can inherit zplprop values.
2860 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2863 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2865 dmu_objset_rele(os, FTAG);
2870 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2871 nvlist_t *zplprops, boolean_t *is_ci)
2875 uint64_t zplver = ZPL_VERSION;
2878 zplver = zfs_zpl_version_map(spa_vers);
2879 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2880 sa_ok = (zplver >= ZPL_VERSION_SA);
2882 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2883 createprops, zplprops, is_ci);
2889 * zc_objset_type type of objset to create (fs vs zvol)
2890 * zc_name name of new objset
2891 * zc_value name of snapshot to clone from (may be empty)
2892 * zc_nvlist_src{_size} nvlist of properties to apply
2897 zfs_ioc_create(zfs_cmd_t *zc)
2902 nvlist_t *nvprops = NULL;
2903 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2904 dmu_objset_type_t type = zc->zc_objset_type;
2909 cbfunc = zfs_create_cb;
2913 cbfunc = zvol_create_cb;
2920 if (strchr(zc->zc_name, '@') ||
2921 strchr(zc->zc_name, '%'))
2924 if (zc->zc_nvlist_src != 0 &&
2925 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2926 zc->zc_iflags, &nvprops)) != 0)
2929 zct.zct_zplprops = NULL;
2930 zct.zct_props = nvprops;
2932 if (zc->zc_value[0] != '\0') {
2934 * We're creating a clone of an existing snapshot.
2936 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2937 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2938 nvlist_free(nvprops);
2942 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2944 nvlist_free(nvprops);
2948 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2949 dmu_objset_rele(clone, FTAG);
2951 nvlist_free(nvprops);
2955 boolean_t is_insensitive = B_FALSE;
2957 if (cbfunc == NULL) {
2958 nvlist_free(nvprops);
2962 if (type == DMU_OST_ZVOL) {
2963 uint64_t volsize, volblocksize;
2965 if (nvprops == NULL ||
2966 nvlist_lookup_uint64(nvprops,
2967 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2969 nvlist_free(nvprops);
2973 if ((error = nvlist_lookup_uint64(nvprops,
2974 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2975 &volblocksize)) != 0 && error != ENOENT) {
2976 nvlist_free(nvprops);
2981 volblocksize = zfs_prop_default_numeric(
2982 ZFS_PROP_VOLBLOCKSIZE);
2984 if ((error = zvol_check_volblocksize(
2985 volblocksize)) != 0 ||
2986 (error = zvol_check_volsize(volsize,
2987 volblocksize)) != 0) {
2988 nvlist_free(nvprops);
2991 } else if (type == DMU_OST_ZFS) {
2995 * We have to have normalization and
2996 * case-folding flags correct when we do the
2997 * file system creation, so go figure them out
3000 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3001 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3002 error = zfs_fill_zplprops(zc->zc_name, nvprops,
3003 zct.zct_zplprops, &is_insensitive);
3005 nvlist_free(nvprops);
3006 nvlist_free(zct.zct_zplprops);
3010 error = dmu_objset_create(zc->zc_name, type,
3011 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3012 nvlist_free(zct.zct_zplprops);
3016 * It would be nice to do this atomically.
3019 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3022 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3024 nvlist_free(nvprops);
3030 * zc_name name of filesystem
3031 * zc_value short name of snapshot
3032 * zc_cookie recursive flag
3033 * zc_nvlist_src[_size] property list
3036 * zc_value short snapname (i.e. part after the '@')
3039 zfs_ioc_snapshot(zfs_cmd_t *zc)
3041 nvlist_t *nvprops = NULL;
3043 boolean_t recursive = zc->zc_cookie;
3045 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3048 if (zc->zc_nvlist_src != 0 &&
3049 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3050 zc->zc_iflags, &nvprops)) != 0)
3053 error = zfs_check_userprops(zc->zc_name, nvprops);
3057 if (!nvlist_empty(nvprops) &&
3058 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3063 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3064 nvprops, recursive, B_FALSE, -1);
3067 nvlist_free(nvprops);
3073 * name dataset name, or when 'arg == NULL' the full snapshot name
3074 * arg short snapshot name (i.e. part after the '@')
3077 zfs_unmount_snap(const char *name, void *arg)
3079 zfs_sb_t *zsb = NULL;
3087 dsname = strdup(name);
3088 snapname = strdup(arg);
3090 ptr = strchr(name, '@');
3092 dsname = strdup(name);
3093 dsname[ptr - name] = '\0';
3094 snapname = strdup(ptr + 1);
3100 fullname = kmem_asprintf("%s@%s", dsname, snapname);
3102 error = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
3104 error = zfsctl_unmount_snapshot(zsb, fullname, MNT_FORCE);
3105 zfs_sb_rele(zsb, FTAG);
3107 /* Allow ENOENT for consistency with upstream */
3108 if (error == ENOENT)
3121 * zc_name name of filesystem
3122 * zc_value short name of snapshot
3123 * zc_defer_destroy mark for deferred destroy
3128 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
3132 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3134 err = dmu_objset_find(zc->zc_name,
3135 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
3138 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
3139 zc->zc_defer_destroy));
3144 * zc_name name of dataset to destroy
3145 * zc_objset_type type of objset
3146 * zc_defer_destroy mark for deferred destroy
3151 zfs_ioc_destroy(zfs_cmd_t *zc)
3154 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3155 err = zfs_unmount_snap(zc->zc_name, NULL);
3160 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3161 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3162 (void) zvol_remove_minor(zc->zc_name);
3168 * zc_name name of dataset to rollback (to most recent snapshot)
3173 zfs_ioc_rollback(zfs_cmd_t *zc)
3175 dsl_dataset_t *ds, *clone;
3180 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3184 /* must not be a snapshot */
3185 if (dsl_dataset_is_snapshot(ds)) {
3186 dsl_dataset_rele(ds, FTAG);
3190 /* must have a most recent snapshot */
3191 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3192 dsl_dataset_rele(ds, FTAG);
3197 * Create clone of most recent snapshot.
3199 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3200 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3204 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3211 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
3212 error = zfs_suspend_fs(zsb);
3216 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3217 error = dsl_dataset_clone_swap(clone, ds,
3219 dsl_dataset_disown(ds, FTAG);
3224 resume_err = zfs_resume_fs(zsb, zc->zc_name);
3225 error = error ? error : resume_err;
3227 deactivate_super(zsb->z_sb);
3229 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3230 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3231 dsl_dataset_disown(ds, FTAG);
3239 * Destroy clone (which also closes it).
3241 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3244 strfree(clone_name);
3246 dsl_dataset_rele(ds, FTAG);
3252 * zc_name old name of dataset
3253 * zc_value new name of dataset
3254 * zc_cookie recursive flag (only valid for snapshots)
3259 zfs_ioc_rename(zfs_cmd_t *zc)
3261 boolean_t recursive = zc->zc_cookie & 1;
3264 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3265 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3266 strchr(zc->zc_value, '%'))
3270 * Unmount snapshot unless we're doing a recursive rename,
3271 * in which case the dataset code figures out which snapshots
3274 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3275 zc->zc_objset_type == DMU_OST_ZFS) {
3276 err = zfs_unmount_snap(zc->zc_name, NULL);
3281 err = dmu_objset_rename(zc->zc_name, zc->zc_value, recursive);
3282 if ((err == 0) && (zc->zc_objset_type == DMU_OST_ZVOL)) {
3283 (void) zvol_remove_minor(zc->zc_name);
3284 (void) zvol_create_minor(zc->zc_value);
3291 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3293 const char *propname = nvpair_name(pair);
3294 boolean_t issnap = (strchr(dsname, '@') != NULL);
3295 zfs_prop_t prop = zfs_name_to_prop(propname);
3299 if (prop == ZPROP_INVAL) {
3300 if (zfs_prop_user(propname)) {
3301 if ((err = zfs_secpolicy_write_perms(dsname,
3302 ZFS_DELEG_PERM_USERPROP, cr)))
3307 if (!issnap && zfs_prop_userquota(propname)) {
3308 const char *perm = NULL;
3309 const char *uq_prefix =
3310 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3311 const char *gq_prefix =
3312 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3314 if (strncmp(propname, uq_prefix,
3315 strlen(uq_prefix)) == 0) {
3316 perm = ZFS_DELEG_PERM_USERQUOTA;
3317 } else if (strncmp(propname, gq_prefix,
3318 strlen(gq_prefix)) == 0) {
3319 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3321 /* USERUSED and GROUPUSED are read-only */
3325 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3336 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3338 * dsl_prop_get_all_impl() returns properties in this
3342 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3343 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3348 * Check that this value is valid for this pool version
3351 case ZFS_PROP_COMPRESSION:
3353 * If the user specified gzip compression, make sure
3354 * the SPA supports it. We ignore any errors here since
3355 * we'll catch them later.
3357 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3358 nvpair_value_uint64(pair, &intval) == 0) {
3359 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3360 intval <= ZIO_COMPRESS_GZIP_9 &&
3361 zfs_earlier_version(dsname,
3362 SPA_VERSION_GZIP_COMPRESSION)) {
3366 if (intval == ZIO_COMPRESS_ZLE &&
3367 zfs_earlier_version(dsname,
3368 SPA_VERSION_ZLE_COMPRESSION))
3372 * If this is a bootable dataset then
3373 * verify that the compression algorithm
3374 * is supported for booting. We must return
3375 * something other than ENOTSUP since it
3376 * implies a downrev pool version.
3378 if (zfs_is_bootfs(dsname) &&
3379 !BOOTFS_COMPRESS_VALID(intval)) {
3385 case ZFS_PROP_COPIES:
3386 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3390 case ZFS_PROP_DEDUP:
3391 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3395 case ZFS_PROP_SHARESMB:
3396 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3400 case ZFS_PROP_ACLINHERIT:
3401 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3402 nvpair_value_uint64(pair, &intval) == 0) {
3403 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3404 zfs_earlier_version(dsname,
3405 SPA_VERSION_PASSTHROUGH_X))
3413 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3417 * Removes properties from the given props list that fail permission checks
3418 * needed to clear them and to restore them in case of a receive error. For each
3419 * property, make sure we have both set and inherit permissions.
3421 * Returns the first error encountered if any permission checks fail. If the
3422 * caller provides a non-NULL errlist, it also gives the complete list of names
3423 * of all the properties that failed a permission check along with the
3424 * corresponding error numbers. The caller is responsible for freeing the
3427 * If every property checks out successfully, zero is returned and the list
3428 * pointed at by errlist is NULL.
3431 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3434 nvpair_t *pair, *next_pair;
3441 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3443 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3444 (void) strcpy(zc->zc_name, dataset);
3445 pair = nvlist_next_nvpair(props, NULL);
3446 while (pair != NULL) {
3447 next_pair = nvlist_next_nvpair(props, pair);
3449 (void) strcpy(zc->zc_value, nvpair_name(pair));
3450 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3451 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3452 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3453 VERIFY(nvlist_add_int32(errors,
3454 zc->zc_value, err) == 0);
3458 kmem_free(zc, sizeof (zfs_cmd_t));
3460 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3461 nvlist_free(errors);
3464 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3467 if (errlist == NULL)
3468 nvlist_free(errors);
3476 propval_equals(nvpair_t *p1, nvpair_t *p2)
3478 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3479 /* dsl_prop_get_all_impl() format */
3481 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3482 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3486 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3488 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3489 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3493 if (nvpair_type(p1) != nvpair_type(p2))
3496 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3497 char *valstr1, *valstr2;
3499 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3500 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3501 return (strcmp(valstr1, valstr2) == 0);
3503 uint64_t intval1, intval2;
3505 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3506 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3507 return (intval1 == intval2);
3512 * Remove properties from props if they are not going to change (as determined
3513 * by comparison with origprops). Remove them from origprops as well, since we
3514 * do not need to clear or restore properties that won't change.
3517 props_reduce(nvlist_t *props, nvlist_t *origprops)
3519 nvpair_t *pair, *next_pair;
3521 if (origprops == NULL)
3522 return; /* all props need to be received */
3524 pair = nvlist_next_nvpair(props, NULL);
3525 while (pair != NULL) {
3526 const char *propname = nvpair_name(pair);
3529 next_pair = nvlist_next_nvpair(props, pair);
3531 if ((nvlist_lookup_nvpair(origprops, propname,
3532 &match) != 0) || !propval_equals(pair, match))
3533 goto next; /* need to set received value */
3535 /* don't clear the existing received value */
3536 (void) nvlist_remove_nvpair(origprops, match);
3537 /* don't bother receiving the property */
3538 (void) nvlist_remove_nvpair(props, pair);
3545 static boolean_t zfs_ioc_recv_inject_err;
3550 * zc_name name of containing filesystem
3551 * zc_nvlist_src{_size} nvlist of properties to apply
3552 * zc_value name of snapshot to create
3553 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3554 * zc_cookie file descriptor to recv from
3555 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3556 * zc_guid force flag
3557 * zc_cleanup_fd cleanup-on-exit file descriptor
3558 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3561 * zc_cookie number of bytes read
3562 * zc_nvlist_dst{_size} error for each unapplied received property
3563 * zc_obj zprop_errflags_t
3564 * zc_action_handle handle for this guid/ds mapping
3567 zfs_ioc_recv(zfs_cmd_t *zc)
3571 dmu_recv_cookie_t drc;
3572 boolean_t force = (boolean_t)zc->zc_guid;
3575 int props_error = 0;
3578 nvlist_t *props = NULL; /* sent properties */
3579 nvlist_t *origprops = NULL; /* existing properties */
3580 objset_t *origin = NULL;
3582 char tofs[ZFS_MAXNAMELEN];
3583 boolean_t first_recvd_props = B_FALSE;
3585 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3586 strchr(zc->zc_value, '@') == NULL ||
3587 strchr(zc->zc_value, '%'))
3590 (void) strcpy(tofs, zc->zc_value);
3591 tosnap = strchr(tofs, '@');
3594 if (zc->zc_nvlist_src != 0 &&
3595 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3596 zc->zc_iflags, &props)) != 0)
3606 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3608 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3609 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3610 !dsl_prop_get_hasrecvd(os)) {
3611 first_recvd_props = B_TRUE;
3615 * If new received properties are supplied, they are to
3616 * completely replace the existing received properties, so stash
3617 * away the existing ones.
3619 if (dsl_prop_get_received(os, &origprops) == 0) {
3620 nvlist_t *errlist = NULL;
3622 * Don't bother writing a property if its value won't
3623 * change (and avoid the unnecessary security checks).
3625 * The first receive after SPA_VERSION_RECVD_PROPS is a
3626 * special case where we blow away all local properties
3629 if (!first_recvd_props)
3630 props_reduce(props, origprops);
3631 if (zfs_check_clearable(tofs, origprops,
3633 (void) nvlist_merge(errors, errlist, 0);
3634 nvlist_free(errlist);
3637 dmu_objset_rele(os, FTAG);
3640 if (zc->zc_string[0]) {
3641 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3646 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3647 &zc->zc_begin_record, force, origin, &drc);
3649 dmu_objset_rele(origin, FTAG);
3654 * Set properties before we receive the stream so that they are applied
3655 * to the new data. Note that we must call dmu_recv_stream() if
3656 * dmu_recv_begin() succeeds.
3661 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3662 if (drc.drc_newfs) {
3663 if (spa_version(os->os_spa) >=
3664 SPA_VERSION_RECVD_PROPS)
3665 first_recvd_props = B_TRUE;
3666 } else if (origprops != NULL) {
3667 if (clear_received_props(os, tofs, origprops,
3668 first_recvd_props ? NULL : props) != 0)
3669 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3671 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3673 dsl_prop_set_hasrecvd(os);
3674 } else if (!drc.drc_newfs) {
3675 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3678 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3680 (void) nvlist_merge(errors, errlist, 0);
3681 nvlist_free(errlist);
3684 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3686 * Caller made zc->zc_nvlist_dst less than the minimum expected
3687 * size or supplied an invalid address.
3689 props_error = EINVAL;
3693 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3694 &zc->zc_action_handle);
3697 zfs_sb_t *zsb = NULL;
3699 if (get_zfs_sb(tofs, &zsb) == 0) {
3703 error = zfs_suspend_fs(zsb);
3705 * If the suspend fails, then the recv_end will
3706 * likely also fail, and clean up after itself.
3708 end_err = dmu_recv_end(&drc);
3710 error = zfs_resume_fs(zsb, tofs);
3711 error = error ? error : end_err;
3712 deactivate_super(zsb->z_sb);
3714 error = dmu_recv_end(&drc);
3718 zc->zc_cookie = off - fp->f_offset;
3719 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3723 if (zfs_ioc_recv_inject_err) {
3724 zfs_ioc_recv_inject_err = B_FALSE;
3729 * On error, restore the original props.
3731 if (error && props) {
3732 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3733 if (clear_received_props(os, tofs, props, NULL) != 0) {
3735 * We failed to clear the received properties.
3736 * Since we may have left a $recvd value on the
3737 * system, we can't clear the $hasrecvd flag.
3739 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3740 } else if (first_recvd_props) {
3741 dsl_prop_unset_hasrecvd(os);
3743 dmu_objset_rele(os, FTAG);
3744 } else if (!drc.drc_newfs) {
3745 /* We failed to clear the received properties. */
3746 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3749 if (origprops == NULL && !drc.drc_newfs) {
3750 /* We failed to stash the original properties. */
3751 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3755 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3756 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3757 * explictly if we're restoring local properties cleared in the
3758 * first new-style receive.
3760 if (origprops != NULL &&
3761 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3762 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3763 origprops, NULL) != 0) {
3765 * We stashed the original properties but failed to
3768 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3773 nvlist_free(origprops);
3774 nvlist_free(errors);
3778 error = props_error;
3785 * zc_name name of snapshot to send
3786 * zc_cookie file descriptor to send stream to
3787 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3788 * zc_sendobj objsetid of snapshot to send
3789 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3794 zfs_ioc_send(zfs_cmd_t *zc)
3796 objset_t *fromsnap = NULL;
3802 dsl_dataset_t *dsfrom = NULL;
3806 error = spa_open(zc->zc_name, &spa, FTAG);
3810 dp = spa_get_dsl(spa);
3811 rw_enter(&dp->dp_config_rwlock, RW_READER);
3812 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3813 rw_exit(&dp->dp_config_rwlock);
3815 spa_close(spa, FTAG);
3819 error = dmu_objset_from_ds(ds, &tosnap);
3821 dsl_dataset_rele(ds, FTAG);
3822 spa_close(spa, FTAG);
3826 if (zc->zc_fromobj != 0) {
3827 rw_enter(&dp->dp_config_rwlock, RW_READER);
3828 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3829 rw_exit(&dp->dp_config_rwlock);
3830 spa_close(spa, FTAG);
3832 dsl_dataset_rele(ds, FTAG);
3835 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3837 dsl_dataset_rele(dsfrom, FTAG);
3838 dsl_dataset_rele(ds, FTAG);
3842 spa_close(spa, FTAG);
3845 fp = getf(zc->zc_cookie);
3847 dsl_dataset_rele(ds, FTAG);
3849 dsl_dataset_rele(dsfrom, FTAG);
3854 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3856 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3858 releasef(zc->zc_cookie);
3860 dsl_dataset_rele(dsfrom, FTAG);
3861 dsl_dataset_rele(ds, FTAG);
3866 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3870 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3871 &zc->zc_inject_record);
3874 zc->zc_guid = (uint64_t)id;
3880 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3882 return (zio_clear_fault((int)zc->zc_guid));
3886 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3888 int id = (int)zc->zc_guid;
3891 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3892 &zc->zc_inject_record);
3900 zfs_ioc_error_log(zfs_cmd_t *zc)
3904 size_t count = (size_t)zc->zc_nvlist_dst_size;
3906 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3909 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3912 zc->zc_nvlist_dst_size = count;
3914 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3916 spa_close(spa, FTAG);
3922 zfs_ioc_clear(zfs_cmd_t *zc)
3929 * On zpool clear we also fix up missing slogs
3931 mutex_enter(&spa_namespace_lock);
3932 spa = spa_lookup(zc->zc_name);
3934 mutex_exit(&spa_namespace_lock);
3937 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3938 /* we need to let spa_open/spa_load clear the chains */
3939 spa_set_log_state(spa, SPA_LOG_CLEAR);
3941 spa->spa_last_open_failed = 0;
3942 mutex_exit(&spa_namespace_lock);
3944 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3945 error = spa_open(zc->zc_name, &spa, FTAG);
3948 nvlist_t *config = NULL;
3950 if (zc->zc_nvlist_src == 0)
3953 if ((error = get_nvlist(zc->zc_nvlist_src,
3954 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3955 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3957 if (config != NULL) {
3960 if ((err = put_nvlist(zc, config)) != 0)
3962 nvlist_free(config);
3964 nvlist_free(policy);
3971 spa_vdev_state_enter(spa, SCL_NONE);
3973 if (zc->zc_guid == 0) {
3976 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3978 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
3979 spa_close(spa, FTAG);
3984 vdev_clear(spa, vd);
3986 (void) spa_vdev_state_exit(spa, NULL, 0);
3989 * Resume any suspended I/Os.
3991 if (zio_resume(spa) != 0)
3994 spa_close(spa, FTAG);
4001 * zc_name name of filesystem
4002 * zc_value name of origin snapshot
4005 * zc_string name of conflicting snapshot, if there is one
4008 zfs_ioc_promote(zfs_cmd_t *zc)
4013 * We don't need to unmount *all* the origin fs's snapshots, but
4016 cp = strchr(zc->zc_value, '@');
4019 (void) dmu_objset_find(zc->zc_value,
4020 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4021 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4025 * Retrieve a single {user|group}{used|quota}@... property.
4028 * zc_name name of filesystem
4029 * zc_objset_type zfs_userquota_prop_t
4030 * zc_value domain name (eg. "S-1-234-567-89")
4031 * zc_guid RID/UID/GID
4034 * zc_cookie property value
4037 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4042 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4045 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4049 error = zfs_userspace_one(zsb,
4050 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4051 zfs_sb_rele(zsb, FTAG);
4058 * zc_name name of filesystem
4059 * zc_cookie zap cursor
4060 * zc_objset_type zfs_userquota_prop_t
4061 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4064 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4065 * zc_cookie zap cursor
4068 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4071 int bufsize = zc->zc_nvlist_dst_size;
4078 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4082 buf = vmem_alloc(bufsize, KM_SLEEP);
4084 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4085 buf, &zc->zc_nvlist_dst_size);
4088 error = xcopyout(buf,
4089 (void *)(uintptr_t)zc->zc_nvlist_dst,
4090 zc->zc_nvlist_dst_size);
4092 vmem_free(buf, bufsize);
4093 zfs_sb_rele(zsb, FTAG);
4100 * zc_name name of filesystem
4106 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4112 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4113 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4115 * If userused is not enabled, it may be because the
4116 * objset needs to be closed & reopened (to grow the
4117 * objset_phys_t). Suspend/resume the fs will do that.
4119 error = zfs_suspend_fs(zsb);
4121 error = zfs_resume_fs(zsb, zc->zc_name);
4124 error = dmu_objset_userspace_upgrade(zsb->z_os);
4125 deactivate_super(zsb->z_sb);
4127 /* XXX kind of reading contents without owning */
4128 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4132 error = dmu_objset_userspace_upgrade(os);
4133 dmu_objset_rele(os, FTAG);
4140 zfs_ioc_share(zfs_cmd_t *zc)
4145 ace_t full_access[] = {
4146 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4151 * zc_name name of containing filesystem
4152 * zc_obj object # beyond which we want next in-use object #
4155 * zc_obj next in-use object #
4158 zfs_ioc_next_obj(zfs_cmd_t *zc)
4160 objset_t *os = NULL;
4163 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4167 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4168 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4170 dmu_objset_rele(os, FTAG);
4176 * zc_name name of filesystem
4177 * zc_value prefix name for snapshot
4178 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4183 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4188 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4189 (u_longlong_t)ddi_get_lbolt64());
4191 if (strlen(snap_name) >= MAXNAMELEN) {
4196 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4197 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4203 (void) strcpy(zc->zc_value, snap_name);
4210 * zc_name name of "to" snapshot
4211 * zc_value name of "from" snapshot
4212 * zc_cookie file descriptor to write diff data on
4215 * dmu_diff_record_t's to the file descriptor
4218 zfs_ioc_diff(zfs_cmd_t *zc)
4226 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4230 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4232 dmu_objset_rele(tosnap, FTAG);
4236 fp = getf(zc->zc_cookie);
4238 dmu_objset_rele(fromsnap, FTAG);
4239 dmu_objset_rele(tosnap, FTAG);
4245 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4247 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4249 releasef(zc->zc_cookie);
4251 dmu_objset_rele(fromsnap, FTAG);
4252 dmu_objset_rele(tosnap, FTAG);
4257 * Remove all ACL files in shares dir
4259 #ifdef HAVE_SMB_SHARE
4261 zfs_smb_acl_purge(znode_t *dzp)
4264 zap_attribute_t zap;
4265 zfs_sb_t *zsb = ZTOZSB(dzp);
4268 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4269 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4270 zap_cursor_advance(&zc)) {
4271 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4275 zap_cursor_fini(&zc);
4278 #endif /* HAVE_SMB_SHARE */
4281 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4283 #ifdef HAVE_SMB_SHARE
4286 vnode_t *resourcevp = NULL;
4295 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4296 NO_FOLLOW, NULL, &vp)) != 0)
4299 /* Now make sure mntpnt and dataset are ZFS */
4301 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4302 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4303 zc->zc_name) != 0)) {
4313 * Create share dir if its missing.
4315 mutex_enter(&zsb->z_lock);
4316 if (zsb->z_shares_dir == 0) {
4319 tx = dmu_tx_create(zsb->z_os);
4320 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4322 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4323 error = dmu_tx_assign(tx, TXG_WAIT);
4327 error = zfs_create_share_dir(zsb, tx);
4331 mutex_exit(&zsb->z_lock);
4337 mutex_exit(&zsb->z_lock);
4339 ASSERT(zsb->z_shares_dir);
4340 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4346 switch (zc->zc_cookie) {
4347 case ZFS_SMB_ACL_ADD:
4348 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4349 vattr.va_mode = S_IFREG|0777;
4353 vsec.vsa_mask = VSA_ACE;
4354 vsec.vsa_aclentp = &full_access;
4355 vsec.vsa_aclentsz = sizeof (full_access);
4356 vsec.vsa_aclcnt = 1;
4358 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4359 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4361 VN_RELE(resourcevp);
4364 case ZFS_SMB_ACL_REMOVE:
4365 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4369 case ZFS_SMB_ACL_RENAME:
4370 if ((error = get_nvlist(zc->zc_nvlist_src,
4371 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4376 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4377 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4380 VN_RELE(ZTOV(sharedir));
4382 nvlist_free(nvlist);
4385 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4387 nvlist_free(nvlist);
4390 case ZFS_SMB_ACL_PURGE:
4391 error = zfs_smb_acl_purge(sharedir);
4400 VN_RELE(ZTOV(sharedir));
4407 #endif /* HAVE_SMB_SHARE */
4412 * zc_name name of filesystem
4413 * zc_value short name of snap
4414 * zc_string user-supplied tag for this hold
4415 * zc_cookie recursive flag
4416 * zc_temphold set if hold is temporary
4417 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4418 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4419 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4424 zfs_ioc_hold(zfs_cmd_t *zc)
4426 boolean_t recursive = zc->zc_cookie;
4433 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4436 if (zc->zc_sendobj == 0) {
4437 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4438 zc->zc_string, recursive, zc->zc_temphold,
4439 zc->zc_cleanup_fd));
4445 error = spa_open(zc->zc_name, &spa, FTAG);
4449 dp = spa_get_dsl(spa);
4450 rw_enter(&dp->dp_config_rwlock, RW_READER);
4451 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4452 rw_exit(&dp->dp_config_rwlock);
4453 spa_close(spa, FTAG);
4458 * Until we have a hold on this snapshot, it's possible that
4459 * zc_sendobj could've been destroyed and reused as part
4460 * of a later txg. Make sure we're looking at the right object.
4462 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4463 dsl_dataset_rele(ds, FTAG);
4467 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4468 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4470 dsl_dataset_rele(ds, FTAG);
4475 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4479 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4482 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4484 dsl_dataset_rele(ds, FTAG);
4491 * zc_name name of dataset from which we're releasing a user hold
4492 * zc_value short name of snap
4493 * zc_string user-supplied tag for this hold
4494 * zc_cookie recursive flag
4499 zfs_ioc_release(zfs_cmd_t *zc)
4501 boolean_t recursive = zc->zc_cookie;
4503 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4506 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4507 zc->zc_string, recursive));
4512 * zc_name name of filesystem
4515 * zc_nvlist_src{_size} nvlist of snapshot holds
4518 zfs_ioc_get_holds(zfs_cmd_t *zc)
4523 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4524 error = put_nvlist(zc, nvp);
4533 * zc_guid flags (ZEVENT_NONBLOCK)
4536 * zc_nvlist_dst next nvlist event
4537 * zc_cookie dropped events since last get
4538 * zc_cleanup_fd cleanup-on-exit file descriptor
4541 zfs_ioc_events_next(zfs_cmd_t *zc)
4544 nvlist_t *event = NULL;
4546 uint64_t dropped = 0;
4549 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4554 error = zfs_zevent_next(ze, &event,
4555 &zc->zc_nvlist_dst_size, &dropped);
4556 if (event != NULL) {
4557 zc->zc_cookie = dropped;
4558 error = put_nvlist(zc, event);
4562 if (zc->zc_guid & ZEVENT_NONBLOCK)
4565 if ((error == 0) || (error != ENOENT))
4568 error = zfs_zevent_wait(ze);
4573 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4580 * zc_cookie cleared events count
4583 zfs_ioc_events_clear(zfs_cmd_t *zc)
4587 zfs_zevent_drain_all(&count);
4588 zc->zc_cookie = count;
4594 * pool create, destroy, and export don't log the history as part of
4595 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4596 * do the logging of those commands.
4598 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4599 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4601 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4603 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4605 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4607 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4609 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4611 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4613 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4614 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4615 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4616 POOL_CHECK_READONLY },
4617 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4618 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4619 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4621 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4622 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4623 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4624 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4625 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4626 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4627 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4628 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4629 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4630 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4631 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4632 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4633 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4634 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4635 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4636 POOL_CHECK_SUSPENDED },
4637 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4639 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4640 POOL_CHECK_SUSPENDED },
4641 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4642 POOL_CHECK_SUSPENDED },
4643 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4644 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4645 { zfs_ioc_create_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4647 { zfs_ioc_remove_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4649 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4650 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4651 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4652 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4653 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4654 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4655 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4656 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4657 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4658 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4659 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE,
4661 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4663 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4665 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4667 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4669 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4671 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4672 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4673 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4674 B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4675 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4676 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4677 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4679 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4680 POOL_CHECK_SUSPENDED },
4681 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4682 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4683 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4685 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4686 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4687 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4689 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4691 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4692 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4693 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4695 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4696 B_FALSE, POOL_CHECK_NONE },
4697 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
4698 B_FALSE, POOL_CHECK_NONE },
4699 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4700 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4701 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
4702 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4703 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4704 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4705 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4706 POOL_CHECK_SUSPENDED },
4707 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4709 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4710 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4711 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4713 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4715 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
4716 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4717 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4718 POOL_CHECK_SUSPENDED },
4719 { zfs_ioc_pool_reguid, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4720 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4721 { zfs_ioc_events_next, zfs_secpolicy_config, NO_NAME, B_FALSE,
4723 { zfs_ioc_events_clear, zfs_secpolicy_config, NO_NAME, B_FALSE,
4728 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
4729 zfs_ioc_poolcheck_t check)
4734 ASSERT(type == POOL_NAME || type == DATASET_NAME);
4736 if (check & POOL_CHECK_NONE)
4739 error = spa_open(name, &spa, FTAG);
4741 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
4743 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
4745 spa_close(spa, FTAG);
4751 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
4755 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4757 for (zs = list_head(&zfsdev_state_list); zs != NULL;
4758 zs = list_next(&zfsdev_state_list, zs)) {
4759 if (zs->zs_minor == minor) {
4761 case ZST_ONEXIT: return (zs->zs_onexit);
4762 case ZST_ZEVENT: return (zs->zs_zevent);
4763 case ZST_ALL: return (zs);
4772 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
4776 mutex_enter(&zfsdev_state_lock);
4777 ptr = zfsdev_get_state_impl(minor, which);
4778 mutex_exit(&zfsdev_state_lock);
4784 zfsdev_getminor(struct file *filp)
4786 ASSERT(filp != NULL);
4787 ASSERT(filp->private_data != NULL);
4789 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
4793 * Find a free minor number. The zfsdev_state_list is expected to
4794 * be short since it is only a list of currently open file handles.
4797 zfsdev_minor_alloc(void)
4799 static minor_t last_minor = 0;
4802 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4804 for (m = last_minor + 1; m != last_minor; m++) {
4805 if (m > ZFSDEV_MAX_MINOR)
4807 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
4817 zfsdev_state_init(struct file *filp)
4822 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4824 minor = zfsdev_minor_alloc();
4828 zs = kmem_zalloc( sizeof(zfsdev_state_t), KM_SLEEP);
4833 zs->zs_minor = minor;
4834 filp->private_data = zs;
4836 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
4837 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
4839 list_insert_tail(&zfsdev_state_list, zs);
4845 zfsdev_state_destroy(struct file *filp)
4849 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4850 ASSERT(filp->private_data != NULL);
4852 zs = filp->private_data;
4853 zfs_onexit_destroy(zs->zs_onexit);
4854 zfs_zevent_destroy(zs->zs_zevent);
4856 list_remove(&zfsdev_state_list, zs);
4857 kmem_free(zs, sizeof(zfsdev_state_t));
4863 zfsdev_open(struct inode *ino, struct file *filp)
4867 mutex_enter(&zfsdev_state_lock);
4868 error = zfsdev_state_init(filp);
4869 mutex_exit(&zfsdev_state_lock);
4875 zfsdev_release(struct inode *ino, struct file *filp)
4879 mutex_enter(&zfsdev_state_lock);
4880 error = zfsdev_state_destroy(filp);
4881 mutex_exit(&zfsdev_state_lock);
4887 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4891 int error, rc, flag = 0;
4893 vec = cmd - ZFS_IOC;
4894 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4897 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
4899 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4903 if ((error == 0) && !(flag & FKIOCTL))
4904 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, CRED());
4907 * Ensure that all pool/dataset names are valid before we pass down to
4911 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4912 zc->zc_iflags = flag & FKIOCTL;
4913 switch (zfs_ioc_vec[vec].zvec_namecheck) {
4915 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4917 error = pool_status_check(zc->zc_name,
4918 zfs_ioc_vec[vec].zvec_namecheck,
4919 zfs_ioc_vec[vec].zvec_pool_check);
4923 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4925 error = pool_status_check(zc->zc_name,
4926 zfs_ioc_vec[vec].zvec_namecheck,
4927 zfs_ioc_vec[vec].zvec_pool_check);
4936 error = zfs_ioc_vec[vec].zvec_func(zc);
4938 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4942 if (zfs_ioc_vec[vec].zvec_his_log)
4943 zfs_log_history(zc);
4946 kmem_free(zc, sizeof (zfs_cmd_t));
4950 #ifdef CONFIG_COMPAT
4952 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4954 return zfsdev_ioctl(filp, cmd, arg);
4957 #define zfsdev_compat_ioctl NULL
4960 static const struct file_operations zfsdev_fops = {
4961 .open = zfsdev_open,
4962 .release = zfsdev_release,
4963 .unlocked_ioctl = zfsdev_ioctl,
4964 .compat_ioctl = zfsdev_compat_ioctl,
4965 .owner = THIS_MODULE,
4968 static struct miscdevice zfs_misc = {
4969 .minor = MISC_DYNAMIC_MINOR,
4971 .fops = &zfsdev_fops,
4979 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
4980 list_create(&zfsdev_state_list, sizeof (zfsdev_state_t),
4981 offsetof(zfsdev_state_t, zs_next));
4983 error = misc_register(&zfs_misc);
4985 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
4997 error = misc_deregister(&zfs_misc);
4999 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5001 mutex_destroy(&zfsdev_state_lock);
5002 list_destroy(&zfsdev_state_list);
5005 uint_t zfs_fsyncer_key;
5006 extern uint_t rrw_tsd_key;
5009 #define ZFS_DEBUG_STR " (DEBUG mode)"
5011 #define ZFS_DEBUG_STR ""
5019 spa_init(FREAD | FWRITE);
5022 if ((error = zvol_init()) != 0)
5025 if ((error = zfs_attach()) != 0)
5028 tsd_create(&zfs_fsyncer_key, NULL);
5029 tsd_create(&rrw_tsd_key, NULL);
5031 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
5032 "ZFS pool version %s, ZFS filesystem version %s\n",
5033 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
5034 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5043 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
5044 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
5045 ZFS_DEBUG_STR, error);
5058 tsd_destroy(&zfs_fsyncer_key);
5059 tsd_destroy(&rrw_tsd_key);
5061 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
5062 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
5068 spl_module_init(_init);
5069 spl_module_exit(_fini);
5071 MODULE_DESCRIPTION("ZFS");
5072 MODULE_AUTHOR(ZFS_META_AUTHOR);
5073 MODULE_LICENSE(ZFS_META_LICENSE);
5074 #endif /* HAVE_SPL */