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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Portions Copyright 2011 Martin Matuska
25 * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
28 * Copyright (c) 2012 by Delphix. All rights reserved.
29 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
33 #include <sys/types.h>
34 #include <sys/param.h>
35 #include <sys/errno.h>
38 #include <sys/modctl.h>
43 #include <sys/cmn_err.h>
45 #include <sys/zfs_ioctl.h>
46 #include <sys/zfs_vfsops.h>
47 #include <sys/zfs_znode.h>
50 #include <sys/spa_impl.h>
52 #include <sys/priv_impl.h>
54 #include <sys/dsl_dir.h>
55 #include <sys/dsl_dataset.h>
56 #include <sys/dsl_prop.h>
57 #include <sys/dsl_deleg.h>
58 #include <sys/dmu_objset.h>
59 #include <sys/dmu_impl.h>
61 #include <sys/sunddi.h>
62 #include <sys/sunldi.h>
63 #include <sys/policy.h>
65 #include <sys/nvpair.h>
66 #include <sys/pathname.h>
67 #include <sys/mount.h>
69 #include <sys/fs/zfs.h>
70 #include <sys/zfs_ctldir.h>
71 #include <sys/zfs_dir.h>
72 #include <sys/zfs_onexit.h>
74 #include <sys/dsl_scan.h>
75 #include <sharefs/share.h>
76 #include <sys/dmu_objset.h>
77 #include <sys/fm/util.h>
79 #include <sys/zfeature.h>
81 #include <linux/miscdevice.h>
83 #include "zfs_namecheck.h"
85 #include "zfs_deleg.h"
86 #include "zfs_comutil.h"
88 kmutex_t zfsdev_state_lock;
89 list_t zfsdev_state_list;
91 extern void zfs_init(void);
92 extern void zfs_fini(void);
94 typedef int zfs_ioc_func_t(zfs_cmd_t *);
95 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
101 } zfs_ioc_namecheck_t;
104 POOL_CHECK_NONE = 1 << 0,
105 POOL_CHECK_SUSPENDED = 1 << 1,
106 POOL_CHECK_READONLY = 1 << 2
107 } zfs_ioc_poolcheck_t;
109 typedef struct zfs_ioc_vec {
110 zfs_ioc_func_t *zvec_func;
111 zfs_secpolicy_func_t *zvec_secpolicy;
112 zfs_ioc_namecheck_t zvec_namecheck;
113 boolean_t zvec_his_log;
114 zfs_ioc_poolcheck_t zvec_pool_check;
117 /* This array is indexed by zfs_userquota_prop_t */
118 static const char *userquota_perms[] = {
119 ZFS_DELEG_PERM_USERUSED,
120 ZFS_DELEG_PERM_USERQUOTA,
121 ZFS_DELEG_PERM_GROUPUSED,
122 ZFS_DELEG_PERM_GROUPQUOTA,
125 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
126 static int zfs_check_settable(const char *name, nvpair_t *property,
128 static int zfs_check_clearable(char *dataset, nvlist_t *props,
130 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
132 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **);
134 static int zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature);
135 static int zfs_prop_activate_feature_check(void *arg1, void *arg2,
137 static void zfs_prop_activate_feature_sync(void *arg1, void *arg2,
141 history_str_free(char *buf)
143 kmem_free(buf, HIS_MAX_RECORD_LEN);
147 history_str_get(zfs_cmd_t *zc)
151 if (zc->zc_history == 0)
154 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP | KM_NODEBUG);
155 if (copyinstr((void *)(uintptr_t)zc->zc_history,
156 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
157 history_str_free(buf);
161 buf[HIS_MAX_RECORD_LEN -1] = '\0';
167 * Check to see if the named dataset is currently defined as bootable
170 zfs_is_bootfs(const char *name)
174 if (dmu_objset_hold(name, FTAG, &os) == 0) {
176 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
177 dmu_objset_rele(os, FTAG);
184 * zfs_earlier_version
186 * Return non-zero if the spa version is less than requested version.
189 zfs_earlier_version(const char *name, int version)
193 if (spa_open(name, &spa, FTAG) == 0) {
194 if (spa_version(spa) < version) {
195 spa_close(spa, FTAG);
198 spa_close(spa, FTAG);
204 * zpl_earlier_version
206 * Return TRUE if the ZPL version is less than requested version.
209 zpl_earlier_version(const char *name, int version)
212 boolean_t rc = B_TRUE;
214 if (dmu_objset_hold(name, FTAG, &os) == 0) {
217 if (dmu_objset_type(os) != DMU_OST_ZFS) {
218 dmu_objset_rele(os, FTAG);
221 /* XXX reading from non-owned objset */
222 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
223 rc = zplversion < version;
224 dmu_objset_rele(os, FTAG);
230 zfs_log_history(zfs_cmd_t *zc)
235 if ((buf = history_str_get(zc)) == NULL)
238 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
239 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
240 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
241 spa_close(spa, FTAG);
243 history_str_free(buf);
247 * Policy for top-level read operations (list pools). Requires no privileges,
248 * and can be used in the local zone, as there is no associated dataset.
252 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
258 * Policy for dataset read operations (list children, get statistics). Requires
259 * no privileges, but must be visible in the local zone.
263 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
265 if (INGLOBALZONE(curproc) ||
266 zone_dataset_visible(zc->zc_name, NULL))
273 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
278 * The dataset must be visible by this zone -- check this first
279 * so they don't see EPERM on something they shouldn't know about.
281 if (!INGLOBALZONE(curproc) &&
282 !zone_dataset_visible(dataset, &writable))
285 if (INGLOBALZONE(curproc)) {
287 * If the fs is zoned, only root can access it from the
290 if (secpolicy_zfs(cr) && zoned)
294 * If we are in a local zone, the 'zoned' property must be set.
299 /* must be writable by this zone */
307 zfs_dozonecheck(const char *dataset, cred_t *cr)
311 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
314 return (zfs_dozonecheck_impl(dataset, zoned, cr));
318 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
322 rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
323 if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
324 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
327 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
329 return (zfs_dozonecheck_impl(dataset, zoned, cr));
333 * If name ends in a '@', then require recursive permissions.
336 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
339 boolean_t descendent = B_FALSE;
343 at = strchr(name, '@');
344 if (at != NULL && at[1] == '\0') {
349 error = dsl_dataset_hold(name, FTAG, &ds);
355 error = zfs_dozonecheck_ds(name, ds, cr);
357 error = secpolicy_zfs(cr);
359 error = dsl_deleg_access_impl(ds, descendent, perm, cr);
362 dsl_dataset_rele(ds, FTAG);
367 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
368 const char *perm, cred_t *cr)
372 error = zfs_dozonecheck_ds(name, ds, cr);
374 error = secpolicy_zfs(cr);
376 error = dsl_deleg_access_impl(ds, B_FALSE, perm, cr);
382 * Policy for setting the security label property.
384 * Returns 0 for success, non-zero for access and other errors.
387 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
390 char ds_hexsl[MAXNAMELEN];
391 bslabel_t ds_sl, new_sl;
392 boolean_t new_default = FALSE;
394 int needed_priv = -1;
397 /* First get the existing dataset label. */
398 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
399 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
403 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
406 /* The label must be translatable */
407 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
411 * In a non-global zone, disallow attempts to set a label that
412 * doesn't match that of the zone; otherwise no other checks
415 if (!INGLOBALZONE(curproc)) {
416 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
422 * For global-zone datasets (i.e., those whose zoned property is
423 * "off", verify that the specified new label is valid for the
426 if (dsl_prop_get_integer(name,
427 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
430 if (zfs_check_global_label(name, strval) != 0)
435 * If the existing dataset label is nondefault, check if the
436 * dataset is mounted (label cannot be changed while mounted).
437 * Get the zfs_sb_t; if there isn't one, then the dataset isn't
438 * mounted (or isn't a dataset, doesn't exist, ...).
440 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
442 static char *setsl_tag = "setsl_tag";
445 * Try to own the dataset; abort if there is any error,
446 * (e.g., already mounted, in use, or other error).
448 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
453 dmu_objset_disown(os, setsl_tag);
456 needed_priv = PRIV_FILE_DOWNGRADE_SL;
460 if (hexstr_to_label(strval, &new_sl) != 0)
463 if (blstrictdom(&ds_sl, &new_sl))
464 needed_priv = PRIV_FILE_DOWNGRADE_SL;
465 else if (blstrictdom(&new_sl, &ds_sl))
466 needed_priv = PRIV_FILE_UPGRADE_SL;
468 /* dataset currently has a default label */
470 needed_priv = PRIV_FILE_UPGRADE_SL;
474 if (needed_priv != -1)
475 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
479 #endif /* HAVE_MLSLABEL */
483 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
489 * Check permissions for special properties.
496 * Disallow setting of 'zoned' from within a local zone.
498 if (!INGLOBALZONE(curproc))
503 if (!INGLOBALZONE(curproc)) {
505 char setpoint[MAXNAMELEN];
507 * Unprivileged users are allowed to modify the
508 * quota on things *under* (ie. contained by)
509 * the thing they own.
511 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
514 if (!zoned || strlen(dsname) <= strlen(setpoint))
519 case ZFS_PROP_MLSLABEL:
520 if (!is_system_labeled())
523 if (nvpair_value_string(propval, &strval) == 0) {
526 err = zfs_set_slabel_policy(dsname, strval, CRED());
533 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
537 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
541 error = zfs_dozonecheck(zc->zc_name, cr);
546 * permission to set permissions will be evaluated later in
547 * dsl_deleg_can_allow()
553 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
555 return (zfs_secpolicy_write_perms(zc->zc_name,
556 ZFS_DELEG_PERM_ROLLBACK, cr));
560 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
569 * Generate the current snapshot name from the given objsetid, then
570 * use that name for the secpolicy/zone checks.
572 cp = strchr(zc->zc_name, '@');
575 error = spa_open(zc->zc_name, &spa, FTAG);
579 dp = spa_get_dsl(spa);
580 rw_enter(&dp->dp_config_rwlock, RW_READER);
581 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
582 rw_exit(&dp->dp_config_rwlock);
583 spa_close(spa, FTAG);
587 dsl_dataset_name(ds, zc->zc_name);
589 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
590 ZFS_DELEG_PERM_SEND, cr);
591 dsl_dataset_rele(ds, FTAG);
596 #ifdef HAVE_SMB_SHARE
598 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
603 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
604 NO_FOLLOW, NULL, &vp)) != 0)
607 /* Now make sure mntpnt and dataset are ZFS */
609 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
610 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
611 zc->zc_name) != 0)) {
617 return (dsl_deleg_access(zc->zc_name,
618 ZFS_DELEG_PERM_SHARE, cr));
620 #endif /* HAVE_SMB_SHARE */
623 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
625 #ifdef HAVE_SMB_SHARE
626 if (!INGLOBALZONE(curproc))
629 if (secpolicy_nfs(cr) == 0) {
632 return (zfs_secpolicy_deleg_share(zc, cr));
636 #endif /* HAVE_SMB_SHARE */
640 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
642 #ifdef HAVE_SMB_SHARE
643 if (!INGLOBALZONE(curproc))
646 if (secpolicy_smb(cr) == 0) {
649 return (zfs_secpolicy_deleg_share(zc, cr));
653 #endif /* HAVE_SMB_SHARE */
657 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
662 * Remove the @bla or /bla from the end of the name to get the parent.
664 (void) strncpy(parent, datasetname, parentsize);
665 cp = strrchr(parent, '@');
669 cp = strrchr(parent, '/');
679 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
683 if ((error = zfs_secpolicy_write_perms(name,
684 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
687 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
691 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
693 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
697 * Destroying snapshots with delegated permissions requires
698 * descendent mount and destroy permissions.
701 zfs_secpolicy_destroy_recursive(zfs_cmd_t *zc, cred_t *cr)
706 dsname = kmem_asprintf("%s@", zc->zc_name);
708 error = zfs_secpolicy_destroy_perms(dsname, cr);
710 error = zfs_secpolicy_destroy_perms(zc->zc_name, cr);
717 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
719 char parentname[MAXNAMELEN];
722 if ((error = zfs_secpolicy_write_perms(from,
723 ZFS_DELEG_PERM_RENAME, cr)) != 0)
726 if ((error = zfs_secpolicy_write_perms(from,
727 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
730 if ((error = zfs_get_parent(to, parentname,
731 sizeof (parentname))) != 0)
734 if ((error = zfs_secpolicy_write_perms(parentname,
735 ZFS_DELEG_PERM_CREATE, cr)) != 0)
738 if ((error = zfs_secpolicy_write_perms(parentname,
739 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
746 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
748 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
752 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
754 char parentname[MAXNAMELEN];
758 error = zfs_secpolicy_write_perms(zc->zc_name,
759 ZFS_DELEG_PERM_PROMOTE, cr);
763 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
766 dsl_dataset_t *pclone = NULL;
768 dd = clone->os_dsl_dataset->ds_dir;
770 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
771 error = dsl_dataset_hold_obj(dd->dd_pool,
772 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
773 rw_exit(&dd->dd_pool->dp_config_rwlock);
775 dmu_objset_rele(clone, FTAG);
779 error = zfs_secpolicy_write_perms(zc->zc_name,
780 ZFS_DELEG_PERM_MOUNT, cr);
782 dsl_dataset_name(pclone, parentname);
783 dmu_objset_rele(clone, FTAG);
784 dsl_dataset_rele(pclone, FTAG);
786 error = zfs_secpolicy_write_perms(parentname,
787 ZFS_DELEG_PERM_PROMOTE, cr);
793 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
797 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
798 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
801 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
802 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
805 return (zfs_secpolicy_write_perms(zc->zc_name,
806 ZFS_DELEG_PERM_CREATE, cr));
810 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
812 return (zfs_secpolicy_write_perms(name,
813 ZFS_DELEG_PERM_SNAPSHOT, cr));
817 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
820 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
824 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
826 char parentname[MAXNAMELEN];
829 if ((error = zfs_get_parent(zc->zc_name, parentname,
830 sizeof (parentname))) != 0)
833 if (zc->zc_value[0] != '\0') {
834 if ((error = zfs_secpolicy_write_perms(zc->zc_value,
835 ZFS_DELEG_PERM_CLONE, cr)) != 0)
839 if ((error = zfs_secpolicy_write_perms(parentname,
840 ZFS_DELEG_PERM_CREATE, cr)) != 0)
843 error = zfs_secpolicy_write_perms(parentname,
844 ZFS_DELEG_PERM_MOUNT, cr);
850 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
851 * SYS_CONFIG privilege, which is not available in a local zone.
855 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
857 if (secpolicy_sys_config(cr, B_FALSE) != 0)
864 * Policy for object to name lookups.
868 zfs_secpolicy_diff(zfs_cmd_t *zc, cred_t *cr)
872 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
875 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
880 * Policy for fault injection. Requires all privileges.
884 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
886 return (secpolicy_zinject(cr));
890 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
892 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
894 if (prop == ZPROP_INVAL) {
895 if (!zfs_prop_user(zc->zc_value))
897 return (zfs_secpolicy_write_perms(zc->zc_name,
898 ZFS_DELEG_PERM_USERPROP, cr));
900 return (zfs_secpolicy_setprop(zc->zc_name, prop,
906 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
908 int err = zfs_secpolicy_read(zc, cr);
912 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
915 if (zc->zc_value[0] == 0) {
917 * They are asking about a posix uid/gid. If it's
918 * themself, allow it.
920 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
921 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
922 if (zc->zc_guid == crgetuid(cr))
925 if (groupmember(zc->zc_guid, cr))
930 return (zfs_secpolicy_write_perms(zc->zc_name,
931 userquota_perms[zc->zc_objset_type], cr));
935 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
937 int err = zfs_secpolicy_read(zc, cr);
941 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
944 return (zfs_secpolicy_write_perms(zc->zc_name,
945 userquota_perms[zc->zc_objset_type], cr));
949 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
951 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
956 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
958 return (zfs_secpolicy_write_perms(zc->zc_name,
959 ZFS_DELEG_PERM_HOLD, cr));
963 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
965 return (zfs_secpolicy_write_perms(zc->zc_name,
966 ZFS_DELEG_PERM_RELEASE, cr));
970 * Policy for allowing temporary snapshots to be taken or released
973 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, cred_t *cr)
976 * A temporary snapshot is the same as a snapshot,
977 * hold, destroy and release all rolled into one.
978 * Delegated diff alone is sufficient that we allow this.
982 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
983 ZFS_DELEG_PERM_DIFF, cr)) == 0)
986 error = zfs_secpolicy_snapshot(zc, cr);
988 error = zfs_secpolicy_hold(zc, cr);
990 error = zfs_secpolicy_release(zc, cr);
992 error = zfs_secpolicy_destroy(zc, cr);
997 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1000 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1004 nvlist_t *list = NULL;
1007 * Read in and unpack the user-supplied nvlist.
1012 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1014 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1016 kmem_free(packed, size);
1020 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1021 kmem_free(packed, size);
1025 kmem_free(packed, size);
1032 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
1036 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1038 if (size > zc->zc_nvlist_dst_size) {
1039 nvpair_t *more_errors;
1042 if (zc->zc_nvlist_dst_size < 1024)
1045 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
1046 more_errors = nvlist_prev_nvpair(*errors, NULL);
1049 nvpair_t *pair = nvlist_prev_nvpair(*errors,
1051 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
1053 VERIFY(nvlist_size(*errors, &size,
1054 NV_ENCODE_NATIVE) == 0);
1055 } while (size > zc->zc_nvlist_dst_size);
1057 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
1058 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
1059 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1060 ASSERT(size <= zc->zc_nvlist_dst_size);
1067 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1069 char *packed = NULL;
1073 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
1075 if (size > zc->zc_nvlist_dst_size) {
1078 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1079 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
1081 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1082 size, zc->zc_iflags) != 0)
1084 kmem_free(packed, size);
1087 zc->zc_nvlist_dst_size = size;
1092 get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
1097 error = dmu_objset_hold(dsname, FTAG, &os);
1100 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1101 dmu_objset_rele(os, FTAG);
1105 mutex_enter(&os->os_user_ptr_lock);
1106 *zsbp = dmu_objset_get_user(os);
1107 if (*zsbp && (*zsbp)->z_sb) {
1108 atomic_inc(&((*zsbp)->z_sb->s_active));
1112 mutex_exit(&os->os_user_ptr_lock);
1113 dmu_objset_rele(os, FTAG);
1118 * Find a zfs_sb_t for a mounted filesystem, or create our own, in which
1119 * case its z_sb will be NULL, and it will be opened as the owner.
1120 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1121 * which prevents all inode ops from running.
1124 zfs_sb_hold(const char *name, void *tag, zfs_sb_t **zsbp, boolean_t writer)
1128 if (get_zfs_sb(name, zsbp) != 0)
1129 error = zfs_sb_create(name, zsbp);
1131 rrw_enter(&(*zsbp)->z_teardown_lock, (writer) ? RW_WRITER :
1133 if ((*zsbp)->z_unmounted) {
1135 * XXX we could probably try again, since the unmounting
1136 * thread should be just about to disassociate the
1137 * objset from the zfsvfs.
1139 rrw_exit(&(*zsbp)->z_teardown_lock, tag);
1147 zfs_sb_rele(zfs_sb_t *zsb, void *tag)
1149 rrw_exit(&zsb->z_teardown_lock, tag);
1152 deactivate_super(zsb->z_sb);
1154 dmu_objset_disown(zsb->z_os, zsb);
1160 zfs_ioc_pool_create(zfs_cmd_t *zc)
1163 nvlist_t *config, *props = NULL;
1164 nvlist_t *rootprops = NULL;
1165 nvlist_t *zplprops = NULL;
1168 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1169 zc->zc_iflags, &config)))
1172 if (zc->zc_nvlist_src_size != 0 && (error =
1173 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1174 zc->zc_iflags, &props))) {
1175 nvlist_free(config);
1180 nvlist_t *nvl = NULL;
1181 uint64_t version = SPA_VERSION;
1183 (void) nvlist_lookup_uint64(props,
1184 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1185 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1187 goto pool_props_bad;
1189 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1191 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1193 nvlist_free(config);
1197 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1199 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1200 error = zfs_fill_zplprops_root(version, rootprops,
1203 goto pool_props_bad;
1206 buf = history_str_get(zc);
1208 error = spa_create(zc->zc_name, config, props, buf, zplprops);
1211 * Set the remaining root properties
1213 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1214 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1215 (void) spa_destroy(zc->zc_name);
1218 history_str_free(buf);
1221 nvlist_free(rootprops);
1222 nvlist_free(zplprops);
1223 nvlist_free(config);
1230 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1233 zfs_log_history(zc);
1234 error = spa_destroy(zc->zc_name);
1236 zvol_remove_minors(zc->zc_name);
1241 zfs_ioc_pool_import(zfs_cmd_t *zc)
1243 nvlist_t *config, *props = NULL;
1247 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1248 zc->zc_iflags, &config)) != 0)
1251 if (zc->zc_nvlist_src_size != 0 && (error =
1252 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1253 zc->zc_iflags, &props))) {
1254 nvlist_free(config);
1258 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1259 guid != zc->zc_guid)
1262 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1264 if (zc->zc_nvlist_dst != 0) {
1267 if ((err = put_nvlist(zc, config)) != 0)
1272 zvol_create_minors(zc->zc_name);
1274 nvlist_free(config);
1283 zfs_ioc_pool_export(zfs_cmd_t *zc)
1286 boolean_t force = (boolean_t)zc->zc_cookie;
1287 boolean_t hardforce = (boolean_t)zc->zc_guid;
1289 zfs_log_history(zc);
1290 error = spa_export(zc->zc_name, NULL, force, hardforce);
1292 zvol_remove_minors(zc->zc_name);
1297 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1302 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1305 error = put_nvlist(zc, configs);
1307 nvlist_free(configs);
1314 * zc_name name of the pool
1317 * zc_cookie real errno
1318 * zc_nvlist_dst config nvlist
1319 * zc_nvlist_dst_size size of config nvlist
1322 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1328 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1329 sizeof (zc->zc_value));
1331 if (config != NULL) {
1332 ret = put_nvlist(zc, config);
1333 nvlist_free(config);
1336 * The config may be present even if 'error' is non-zero.
1337 * In this case we return success, and preserve the real errno
1340 zc->zc_cookie = error;
1349 * Try to import the given pool, returning pool stats as appropriate so that
1350 * user land knows which devices are available and overall pool health.
1353 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1355 nvlist_t *tryconfig, *config;
1358 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1359 zc->zc_iflags, &tryconfig)) != 0)
1362 config = spa_tryimport(tryconfig);
1364 nvlist_free(tryconfig);
1369 error = put_nvlist(zc, config);
1370 nvlist_free(config);
1377 * zc_name name of the pool
1378 * zc_cookie scan func (pool_scan_func_t)
1381 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1386 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1389 if (zc->zc_cookie == POOL_SCAN_NONE)
1390 error = spa_scan_stop(spa);
1392 error = spa_scan(spa, zc->zc_cookie);
1394 spa_close(spa, FTAG);
1400 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1405 error = spa_open(zc->zc_name, &spa, FTAG);
1408 spa_close(spa, FTAG);
1414 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1419 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1422 if (zc->zc_cookie < spa_version(spa) ||
1423 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1424 spa_close(spa, FTAG);
1428 spa_upgrade(spa, zc->zc_cookie);
1429 spa_close(spa, FTAG);
1435 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1442 if ((size = zc->zc_history_len) == 0)
1445 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1448 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1449 spa_close(spa, FTAG);
1453 hist_buf = vmem_alloc(size, KM_SLEEP);
1454 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1455 &zc->zc_history_len, hist_buf)) == 0) {
1456 error = ddi_copyout(hist_buf,
1457 (void *)(uintptr_t)zc->zc_history,
1458 zc->zc_history_len, zc->zc_iflags);
1461 spa_close(spa, FTAG);
1462 vmem_free(hist_buf, size);
1467 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1472 error = spa_open(zc->zc_name, &spa, FTAG);
1474 error = spa_change_guid(spa);
1475 spa_close(spa, FTAG);
1481 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1485 if ((error = dsl_dsobj_to_dsname(zc->zc_name,zc->zc_obj,zc->zc_value)))
1493 * zc_name name of filesystem
1494 * zc_obj object to find
1497 * zc_value name of object
1500 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1505 /* XXX reading from objset not owned */
1506 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1508 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1509 dmu_objset_rele(os, FTAG);
1512 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1513 sizeof (zc->zc_value));
1514 dmu_objset_rele(os, FTAG);
1521 * zc_name name of filesystem
1522 * zc_obj object to find
1525 * zc_stat stats on object
1526 * zc_value path to object
1529 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1534 /* XXX reading from objset not owned */
1535 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1537 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1538 dmu_objset_rele(os, FTAG);
1541 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1542 sizeof (zc->zc_value));
1543 dmu_objset_rele(os, FTAG);
1549 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1553 nvlist_t *config, **l2cache, **spares;
1554 uint_t nl2cache = 0, nspares = 0;
1556 error = spa_open(zc->zc_name, &spa, FTAG);
1560 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1561 zc->zc_iflags, &config);
1562 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1563 &l2cache, &nl2cache);
1565 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1569 * A root pool with concatenated devices is not supported.
1570 * Thus, can not add a device to a root pool.
1572 * Intent log device can not be added to a rootpool because
1573 * during mountroot, zil is replayed, a seperated log device
1574 * can not be accessed during the mountroot time.
1576 * l2cache and spare devices are ok to be added to a rootpool.
1578 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1579 nvlist_free(config);
1580 spa_close(spa, FTAG);
1585 error = spa_vdev_add(spa, config);
1586 nvlist_free(config);
1588 spa_close(spa, FTAG);
1594 * zc_name name of the pool
1595 * zc_nvlist_conf nvlist of devices to remove
1596 * zc_cookie to stop the remove?
1599 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1604 error = spa_open(zc->zc_name, &spa, FTAG);
1607 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1608 spa_close(spa, FTAG);
1613 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1617 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1619 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1621 switch (zc->zc_cookie) {
1622 case VDEV_STATE_ONLINE:
1623 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1626 case VDEV_STATE_OFFLINE:
1627 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1630 case VDEV_STATE_FAULTED:
1631 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1632 zc->zc_obj != VDEV_AUX_EXTERNAL)
1633 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1635 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1638 case VDEV_STATE_DEGRADED:
1639 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1640 zc->zc_obj != VDEV_AUX_EXTERNAL)
1641 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1643 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1649 zc->zc_cookie = newstate;
1650 spa_close(spa, FTAG);
1655 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1658 int replacing = zc->zc_cookie;
1662 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1665 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1666 zc->zc_iflags, &config)) == 0) {
1667 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1668 nvlist_free(config);
1671 spa_close(spa, FTAG);
1676 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1681 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1684 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1686 spa_close(spa, FTAG);
1691 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1694 nvlist_t *config, *props = NULL;
1696 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1698 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1701 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1702 zc->zc_iflags, &config))) {
1703 spa_close(spa, FTAG);
1707 if (zc->zc_nvlist_src_size != 0 && (error =
1708 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1709 zc->zc_iflags, &props))) {
1710 spa_close(spa, FTAG);
1711 nvlist_free(config);
1715 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1717 spa_close(spa, FTAG);
1719 nvlist_free(config);
1726 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1729 char *path = zc->zc_value;
1730 uint64_t guid = zc->zc_guid;
1733 error = spa_open(zc->zc_name, &spa, FTAG);
1737 error = spa_vdev_setpath(spa, guid, path);
1738 spa_close(spa, FTAG);
1743 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1746 char *fru = zc->zc_value;
1747 uint64_t guid = zc->zc_guid;
1750 error = spa_open(zc->zc_name, &spa, FTAG);
1754 error = spa_vdev_setfru(spa, guid, fru);
1755 spa_close(spa, FTAG);
1760 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1765 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1767 if (zc->zc_nvlist_dst != 0 &&
1768 (error = dsl_prop_get_all(os, &nv)) == 0) {
1769 dmu_objset_stats(os, nv);
1771 * NB: zvol_get_stats() will read the objset contents,
1772 * which we aren't supposed to do with a
1773 * DS_MODE_USER hold, because it could be
1774 * inconsistent. So this is a bit of a workaround...
1775 * XXX reading with out owning
1777 if (!zc->zc_objset_stats.dds_inconsistent &&
1778 dmu_objset_type(os) == DMU_OST_ZVOL) {
1779 error = zvol_get_stats(os, nv);
1782 VERIFY3S(error, ==, 0);
1785 error = put_nvlist(zc, nv);
1794 * zc_name name of filesystem
1795 * zc_nvlist_dst_size size of buffer for property nvlist
1798 * zc_objset_stats stats
1799 * zc_nvlist_dst property nvlist
1800 * zc_nvlist_dst_size size of property nvlist
1803 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1805 objset_t *os = NULL;
1808 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1811 error = zfs_ioc_objset_stats_impl(zc, os);
1813 dmu_objset_rele(os, FTAG);
1820 * zc_name name of filesystem
1821 * zc_nvlist_dst_size size of buffer for property nvlist
1824 * zc_nvlist_dst received property nvlist
1825 * zc_nvlist_dst_size size of received property nvlist
1827 * Gets received properties (distinct from local properties on or after
1828 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1829 * local property values.
1832 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1834 objset_t *os = NULL;
1838 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1842 * Without this check, we would return local property values if the
1843 * caller has not already received properties on or after
1844 * SPA_VERSION_RECVD_PROPS.
1846 if (!dsl_prop_get_hasrecvd(os)) {
1847 dmu_objset_rele(os, FTAG);
1851 if (zc->zc_nvlist_dst != 0 &&
1852 (error = dsl_prop_get_received(os, &nv)) == 0) {
1853 error = put_nvlist(zc, nv);
1857 dmu_objset_rele(os, FTAG);
1862 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1868 * zfs_get_zplprop() will either find a value or give us
1869 * the default value (if there is one).
1871 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1873 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1879 * zc_name name of filesystem
1880 * zc_nvlist_dst_size size of buffer for zpl property nvlist
1883 * zc_nvlist_dst zpl property nvlist
1884 * zc_nvlist_dst_size size of zpl property nvlist
1887 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1892 /* XXX reading without owning */
1893 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1896 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1899 * NB: nvl_add_zplprop() will read the objset contents,
1900 * which we aren't supposed to do with a DS_MODE_USER
1901 * hold, because it could be inconsistent.
1903 if (zc->zc_nvlist_dst != 0 &&
1904 !zc->zc_objset_stats.dds_inconsistent &&
1905 dmu_objset_type(os) == DMU_OST_ZFS) {
1908 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1909 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1910 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1911 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1912 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1913 err = put_nvlist(zc, nv);
1918 dmu_objset_rele(os, FTAG);
1923 dataset_name_hidden(const char *name)
1926 * Skip over datasets that are not visible in this zone,
1927 * internal datasets (which have a $ in their name), and
1928 * temporary datasets (which have a % in their name).
1930 if (strchr(name, '$') != NULL)
1932 if (strchr(name, '%') != NULL)
1934 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1941 * zc_name name of filesystem
1942 * zc_cookie zap cursor
1943 * zc_nvlist_dst_size size of buffer for property nvlist
1946 * zc_name name of next filesystem
1947 * zc_cookie zap cursor
1948 * zc_objset_stats stats
1949 * zc_nvlist_dst property nvlist
1950 * zc_nvlist_dst_size size of property nvlist
1953 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1958 size_t orig_len = strlen(zc->zc_name);
1961 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
1962 if (error == ENOENT)
1967 p = strrchr(zc->zc_name, '/');
1968 if (p == NULL || p[1] != '\0')
1969 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1970 p = zc->zc_name + strlen(zc->zc_name);
1973 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
1974 * but is not declared void because its called by dmu_objset_find().
1976 if (zc->zc_cookie == 0) {
1977 uint64_t cookie = 0;
1978 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1980 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) {
1981 if (!dataset_name_hidden(zc->zc_name))
1982 (void) dmu_objset_prefetch(zc->zc_name, NULL);
1987 error = dmu_dir_list_next(os,
1988 sizeof (zc->zc_name) - (p - zc->zc_name), p,
1989 NULL, &zc->zc_cookie);
1990 if (error == ENOENT)
1992 } while (error == 0 && dataset_name_hidden(zc->zc_name));
1993 dmu_objset_rele(os, FTAG);
1996 * If it's an internal dataset (ie. with a '$' in its name),
1997 * don't try to get stats for it, otherwise we'll return ENOENT.
1999 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2000 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2001 if (error == ENOENT) {
2002 /* We lost a race with destroy, get the next one. */
2003 zc->zc_name[orig_len] = '\0';
2012 * zc_name name of filesystem
2013 * zc_cookie zap cursor
2014 * zc_nvlist_dst_size size of buffer for property nvlist
2017 * zc_name name of next snapshot
2018 * zc_objset_stats stats
2019 * zc_nvlist_dst property nvlist
2020 * zc_nvlist_dst_size size of property nvlist
2023 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2029 if (zc->zc_cookie == 0 && !zc->zc_simple)
2030 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
2031 NULL, DS_FIND_SNAPSHOTS);
2033 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2035 return (error == ENOENT ? ESRCH : error);
2038 * A dataset name of maximum length cannot have any snapshots,
2039 * so exit immediately.
2041 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2042 dmu_objset_rele(os, FTAG);
2046 error = dmu_snapshot_list_next(os,
2047 sizeof (zc->zc_name) - strlen(zc->zc_name),
2048 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2051 if (error == 0 && !zc->zc_simple) {
2053 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2056 * Since we probably don't have a hold on this snapshot,
2057 * it's possible that the objsetid could have been destroyed
2058 * and reused for a new objset. It's OK if this happens during
2059 * a zfs send operation, since the new createtxg will be
2060 * beyond the range we're interested in.
2062 rw_enter(&dp->dp_config_rwlock, RW_READER);
2063 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2064 rw_exit(&dp->dp_config_rwlock);
2066 if (error == ENOENT) {
2067 /* Racing with destroy, get the next one. */
2068 *strchr(zc->zc_name, '@') = '\0';
2069 dmu_objset_rele(os, FTAG);
2075 error = dmu_objset_from_ds(ds, &ossnap);
2077 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2078 dsl_dataset_rele(ds, FTAG);
2080 } else if (error == ENOENT) {
2084 dmu_objset_rele(os, FTAG);
2085 /* if we failed, undo the @ that we tacked on to zc_name */
2087 *strchr(zc->zc_name, '@') = '\0';
2092 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2094 const char *propname = nvpair_name(pair);
2096 unsigned int vallen;
2099 zfs_userquota_prop_t type;
2105 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2107 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2108 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2114 * A correctly constructed propname is encoded as
2115 * userquota@<rid>-<domain>.
2117 if ((dash = strchr(propname, '-')) == NULL ||
2118 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2127 err = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
2129 err = zfs_set_userquota(zsb, type, domain, rid, quota);
2130 zfs_sb_rele(zsb, FTAG);
2137 * If the named property is one that has a special function to set its value,
2138 * return 0 on success and a positive error code on failure; otherwise if it is
2139 * not one of the special properties handled by this function, return -1.
2141 * XXX: It would be better for callers of the property interface if we handled
2142 * these special cases in dsl_prop.c (in the dsl layer).
2145 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2148 const char *propname = nvpair_name(pair);
2149 zfs_prop_t prop = zfs_name_to_prop(propname);
2153 if (prop == ZPROP_INVAL) {
2154 if (zfs_prop_userquota(propname))
2155 return (zfs_prop_set_userquota(dsname, pair));
2159 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2161 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2162 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2166 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2169 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2172 case ZFS_PROP_QUOTA:
2173 err = dsl_dir_set_quota(dsname, source, intval);
2175 case ZFS_PROP_REFQUOTA:
2176 err = dsl_dataset_set_quota(dsname, source, intval);
2178 case ZFS_PROP_RESERVATION:
2179 err = dsl_dir_set_reservation(dsname, source, intval);
2181 case ZFS_PROP_REFRESERVATION:
2182 err = dsl_dataset_set_reservation(dsname, source, intval);
2184 case ZFS_PROP_VOLSIZE:
2185 err = zvol_set_volsize(dsname, intval);
2187 case ZFS_PROP_VERSION:
2191 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2194 err = zfs_set_version(zsb, intval);
2195 zfs_sb_rele(zsb, FTAG);
2197 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2200 zc = kmem_zalloc(sizeof (zfs_cmd_t),
2201 KM_SLEEP | KM_NODEBUG);
2202 (void) strcpy(zc->zc_name, dsname);
2203 (void) zfs_ioc_userspace_upgrade(zc);
2204 kmem_free(zc, sizeof (zfs_cmd_t));
2208 case ZFS_PROP_COMPRESSION:
2210 if (intval == ZIO_COMPRESS_LZ4) {
2211 zfeature_info_t *feature =
2212 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2216 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2219 dp = spa->spa_dsl_pool;
2222 * Setting the LZ4 compression algorithm activates
2225 if (!spa_feature_is_active(spa, feature)) {
2226 if ((err = zfs_prop_activate_feature(dp,
2228 spa_close(spa, FTAG);
2233 spa_close(spa, FTAG);
2236 * We still want the default set action to be performed in the
2237 * caller, we only performed zfeature settings here.
2251 * This function is best effort. If it fails to set any of the given properties,
2252 * it continues to set as many as it can and returns the first error
2253 * encountered. If the caller provides a non-NULL errlist, it also gives the
2254 * complete list of names of all the properties it failed to set along with the
2255 * corresponding error numbers. The caller is responsible for freeing the
2258 * If every property is set successfully, zero is returned and the list pointed
2259 * at by errlist is NULL.
2262 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2270 nvlist_t *genericnvl;
2274 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2275 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2276 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2280 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2281 const char *propname = nvpair_name(pair);
2282 zfs_prop_t prop = zfs_name_to_prop(propname);
2285 /* decode the property value */
2287 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2289 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2290 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2295 /* Validate value type */
2296 if (err == 0 && prop == ZPROP_INVAL) {
2297 if (zfs_prop_user(propname)) {
2298 if (nvpair_type(propval) != DATA_TYPE_STRING)
2300 } else if (zfs_prop_userquota(propname)) {
2301 if (nvpair_type(propval) !=
2302 DATA_TYPE_UINT64_ARRAY)
2307 } else if (err == 0) {
2308 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2309 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2311 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2314 VERIFY(nvpair_value_uint64(propval,
2317 switch (zfs_prop_get_type(prop)) {
2318 case PROP_TYPE_NUMBER:
2320 case PROP_TYPE_STRING:
2323 case PROP_TYPE_INDEX:
2324 if (zfs_prop_index_to_string(prop,
2325 intval, &unused) != 0)
2330 "unknown property type");
2337 /* Validate permissions */
2339 err = zfs_check_settable(dsname, pair, CRED());
2342 err = zfs_prop_set_special(dsname, source, pair);
2345 * For better performance we build up a list of
2346 * properties to set in a single transaction.
2348 err = nvlist_add_nvpair(genericnvl, pair);
2349 } else if (err != 0 && nvl != retrynvl) {
2351 * This may be a spurious error caused by
2352 * receiving quota and reservation out of order.
2353 * Try again in a second pass.
2355 err = nvlist_add_nvpair(retrynvl, pair);
2360 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2363 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2368 if (!nvlist_empty(genericnvl) &&
2369 dsl_props_set(dsname, source, genericnvl) != 0) {
2371 * If this fails, we still want to set as many properties as we
2372 * can, so try setting them individually.
2375 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2376 const char *propname = nvpair_name(pair);
2380 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2382 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2383 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2387 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2388 VERIFY(nvpair_value_string(propval,
2390 err = dsl_prop_set(dsname, propname, source, 1,
2391 strlen(strval) + 1, strval);
2393 VERIFY(nvpair_value_uint64(propval,
2395 err = dsl_prop_set(dsname, propname, source, 8,
2400 VERIFY(nvlist_add_int32(errors, propname,
2405 nvlist_free(genericnvl);
2406 nvlist_free(retrynvl);
2408 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2409 nvlist_free(errors);
2412 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2415 if (errlist == NULL)
2416 nvlist_free(errors);
2424 * Check that all the properties are valid user properties.
2427 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2429 nvpair_t *pair = NULL;
2432 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2433 const char *propname = nvpair_name(pair);
2436 if (!zfs_prop_user(propname) ||
2437 nvpair_type(pair) != DATA_TYPE_STRING)
2440 if ((error = zfs_secpolicy_write_perms(fsname,
2441 ZFS_DELEG_PERM_USERPROP, CRED())))
2444 if (strlen(propname) >= ZAP_MAXNAMELEN)
2445 return (ENAMETOOLONG);
2447 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2448 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2455 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2459 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2462 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2463 if (nvlist_exists(skipped, nvpair_name(pair)))
2466 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2471 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2475 nvlist_t *cleared_props = NULL;
2476 props_skip(props, skipped, &cleared_props);
2477 if (!nvlist_empty(cleared_props)) {
2479 * Acts on local properties until the dataset has received
2480 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2482 zprop_source_t flags = (ZPROP_SRC_NONE |
2483 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2484 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2486 nvlist_free(cleared_props);
2492 * zc_name name of filesystem
2493 * zc_value name of property to set
2494 * zc_nvlist_src{_size} nvlist of properties to apply
2495 * zc_cookie received properties flag
2498 * zc_nvlist_dst{_size} error for each unapplied received property
2501 zfs_ioc_set_prop(zfs_cmd_t *zc)
2504 boolean_t received = zc->zc_cookie;
2505 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2507 nvlist_t *errors = NULL;
2510 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2511 zc->zc_iflags, &nvl)) != 0)
2515 nvlist_t *origprops;
2518 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2519 if (dsl_prop_get_received(os, &origprops) == 0) {
2520 (void) clear_received_props(os,
2521 zc->zc_name, origprops, nvl);
2522 nvlist_free(origprops);
2525 dsl_prop_set_hasrecvd(os);
2526 dmu_objset_rele(os, FTAG);
2530 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2532 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2533 (void) put_nvlist(zc, errors);
2536 nvlist_free(errors);
2543 * zc_name name of filesystem
2544 * zc_value name of property to inherit
2545 * zc_cookie revert to received value if TRUE
2550 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2552 const char *propname = zc->zc_value;
2553 zfs_prop_t prop = zfs_name_to_prop(propname);
2554 boolean_t received = zc->zc_cookie;
2555 zprop_source_t source = (received
2556 ? ZPROP_SRC_NONE /* revert to received value, if any */
2557 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2566 * zfs_prop_set_special() expects properties in the form of an
2567 * nvpair with type info.
2569 if (prop == ZPROP_INVAL) {
2570 if (!zfs_prop_user(propname))
2573 type = PROP_TYPE_STRING;
2574 } else if (prop == ZFS_PROP_VOLSIZE ||
2575 prop == ZFS_PROP_VERSION) {
2578 type = zfs_prop_get_type(prop);
2581 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2584 case PROP_TYPE_STRING:
2585 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2587 case PROP_TYPE_NUMBER:
2588 case PROP_TYPE_INDEX:
2589 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2596 pair = nvlist_next_nvpair(dummy, NULL);
2597 err = zfs_prop_set_special(zc->zc_name, source, pair);
2600 return (err); /* special property already handled */
2603 * Only check this in the non-received case. We want to allow
2604 * 'inherit -S' to revert non-inheritable properties like quota
2605 * and reservation to the received or default values even though
2606 * they are not considered inheritable.
2608 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2612 /* the property name has been validated by zfs_secpolicy_inherit() */
2613 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2617 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2624 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2625 zc->zc_iflags, &props)))
2629 * If the only property is the configfile, then just do a spa_lookup()
2630 * to handle the faulted case.
2632 pair = nvlist_next_nvpair(props, NULL);
2633 if (pair != NULL && strcmp(nvpair_name(pair),
2634 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2635 nvlist_next_nvpair(props, pair) == NULL) {
2636 mutex_enter(&spa_namespace_lock);
2637 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2638 spa_configfile_set(spa, props, B_FALSE);
2639 spa_config_sync(spa, B_FALSE, B_TRUE);
2641 mutex_exit(&spa_namespace_lock);
2648 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2653 error = spa_prop_set(spa, props);
2656 spa_close(spa, FTAG);
2662 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2666 nvlist_t *nvp = NULL;
2668 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2670 * If the pool is faulted, there may be properties we can still
2671 * get (such as altroot and cachefile), so attempt to get them
2674 mutex_enter(&spa_namespace_lock);
2675 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2676 error = spa_prop_get(spa, &nvp);
2677 mutex_exit(&spa_namespace_lock);
2679 error = spa_prop_get(spa, &nvp);
2680 spa_close(spa, FTAG);
2683 if (error == 0 && zc->zc_nvlist_dst != 0)
2684 error = put_nvlist(zc, nvp);
2694 * zc_name name of volume
2699 zfs_ioc_create_minor(zfs_cmd_t *zc)
2701 return (zvol_create_minor(zc->zc_name));
2706 * zc_name name of volume
2711 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2713 return (zvol_remove_minor(zc->zc_name));
2718 * zc_name name of filesystem
2719 * zc_nvlist_src{_size} nvlist of delegated permissions
2720 * zc_perm_action allow/unallow flag
2725 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2728 nvlist_t *fsaclnv = NULL;
2730 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2731 zc->zc_iflags, &fsaclnv)) != 0)
2735 * Verify nvlist is constructed correctly
2737 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2738 nvlist_free(fsaclnv);
2743 * If we don't have PRIV_SYS_MOUNT, then validate
2744 * that user is allowed to hand out each permission in
2748 error = secpolicy_zfs(CRED());
2750 if (zc->zc_perm_action == B_FALSE) {
2751 error = dsl_deleg_can_allow(zc->zc_name,
2754 error = dsl_deleg_can_unallow(zc->zc_name,
2760 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2762 nvlist_free(fsaclnv);
2768 * zc_name name of filesystem
2771 * zc_nvlist_src{_size} nvlist of delegated permissions
2774 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2779 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2780 error = put_nvlist(zc, nvp);
2789 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2791 zfs_creat_t *zct = arg;
2793 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2796 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2800 * createprops list of properties requested by creator
2801 * default_zplver zpl version to use if unspecified in createprops
2802 * fuids_ok fuids allowed in this version of the spa?
2803 * os parent objset pointer (NULL if root fs)
2806 * zplprops values for the zplprops we attach to the master node object
2807 * is_ci true if requested file system will be purely case-insensitive
2809 * Determine the settings for utf8only, normalization and
2810 * casesensitivity. Specific values may have been requested by the
2811 * creator and/or we can inherit values from the parent dataset. If
2812 * the file system is of too early a vintage, a creator can not
2813 * request settings for these properties, even if the requested
2814 * setting is the default value. We don't actually want to create dsl
2815 * properties for these, so remove them from the source nvlist after
2819 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2820 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2821 nvlist_t *zplprops, boolean_t *is_ci)
2823 uint64_t sense = ZFS_PROP_UNDEFINED;
2824 uint64_t norm = ZFS_PROP_UNDEFINED;
2825 uint64_t u8 = ZFS_PROP_UNDEFINED;
2828 ASSERT(zplprops != NULL);
2831 * Pull out creator prop choices, if any.
2834 (void) nvlist_lookup_uint64(createprops,
2835 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2836 (void) nvlist_lookup_uint64(createprops,
2837 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2838 (void) nvlist_remove_all(createprops,
2839 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2840 (void) nvlist_lookup_uint64(createprops,
2841 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2842 (void) nvlist_remove_all(createprops,
2843 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2844 (void) nvlist_lookup_uint64(createprops,
2845 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2846 (void) nvlist_remove_all(createprops,
2847 zfs_prop_to_name(ZFS_PROP_CASE));
2851 * If the zpl version requested is whacky or the file system
2852 * or pool is version is too "young" to support normalization
2853 * and the creator tried to set a value for one of the props,
2856 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2857 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2858 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2859 (zplver < ZPL_VERSION_NORMALIZATION &&
2860 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2861 sense != ZFS_PROP_UNDEFINED)))
2865 * Put the version in the zplprops
2867 VERIFY(nvlist_add_uint64(zplprops,
2868 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2870 if (norm == ZFS_PROP_UNDEFINED &&
2871 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
2873 VERIFY(nvlist_add_uint64(zplprops,
2874 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2877 * If we're normalizing, names must always be valid UTF-8 strings.
2881 if (u8 == ZFS_PROP_UNDEFINED &&
2882 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
2884 VERIFY(nvlist_add_uint64(zplprops,
2885 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2887 if (sense == ZFS_PROP_UNDEFINED &&
2888 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
2890 VERIFY(nvlist_add_uint64(zplprops,
2891 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2894 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2900 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2901 nvlist_t *zplprops, boolean_t *is_ci)
2903 boolean_t fuids_ok, sa_ok;
2904 uint64_t zplver = ZPL_VERSION;
2905 objset_t *os = NULL;
2906 char parentname[MAXNAMELEN];
2912 (void) strlcpy(parentname, dataset, sizeof (parentname));
2913 cp = strrchr(parentname, '/');
2917 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2920 spa_vers = spa_version(spa);
2921 spa_close(spa, FTAG);
2923 zplver = zfs_zpl_version_map(spa_vers);
2924 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2925 sa_ok = (zplver >= ZPL_VERSION_SA);
2928 * Open parent object set so we can inherit zplprop values.
2930 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2933 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2935 dmu_objset_rele(os, FTAG);
2940 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2941 nvlist_t *zplprops, boolean_t *is_ci)
2945 uint64_t zplver = ZPL_VERSION;
2948 zplver = zfs_zpl_version_map(spa_vers);
2949 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2950 sa_ok = (zplver >= ZPL_VERSION_SA);
2952 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2953 createprops, zplprops, is_ci);
2959 * zc_objset_type type of objset to create (fs vs zvol)
2960 * zc_name name of new objset
2961 * zc_value name of snapshot to clone from (may be empty)
2962 * zc_nvlist_src{_size} nvlist of properties to apply
2967 zfs_ioc_create(zfs_cmd_t *zc)
2972 nvlist_t *nvprops = NULL;
2973 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2974 dmu_objset_type_t type = zc->zc_objset_type;
2979 cbfunc = zfs_create_cb;
2983 cbfunc = zvol_create_cb;
2990 if (strchr(zc->zc_name, '@') ||
2991 strchr(zc->zc_name, '%'))
2994 if (zc->zc_nvlist_src != 0 &&
2995 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2996 zc->zc_iflags, &nvprops)) != 0)
2999 zct.zct_zplprops = NULL;
3000 zct.zct_props = nvprops;
3002 if (zc->zc_value[0] != '\0') {
3004 * We're creating a clone of an existing snapshot.
3006 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3007 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
3008 nvlist_free(nvprops);
3012 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
3014 nvlist_free(nvprops);
3018 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
3019 dmu_objset_rele(clone, FTAG);
3021 nvlist_free(nvprops);
3025 boolean_t is_insensitive = B_FALSE;
3027 if (cbfunc == NULL) {
3028 nvlist_free(nvprops);
3032 if (type == DMU_OST_ZVOL) {
3033 uint64_t volsize, volblocksize;
3035 if (nvprops == NULL ||
3036 nvlist_lookup_uint64(nvprops,
3037 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
3039 nvlist_free(nvprops);
3043 if ((error = nvlist_lookup_uint64(nvprops,
3044 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3045 &volblocksize)) != 0 && error != ENOENT) {
3046 nvlist_free(nvprops);
3051 volblocksize = zfs_prop_default_numeric(
3052 ZFS_PROP_VOLBLOCKSIZE);
3054 if ((error = zvol_check_volblocksize(
3055 volblocksize)) != 0 ||
3056 (error = zvol_check_volsize(volsize,
3057 volblocksize)) != 0) {
3058 nvlist_free(nvprops);
3061 } else if (type == DMU_OST_ZFS) {
3065 * We have to have normalization and
3066 * case-folding flags correct when we do the
3067 * file system creation, so go figure them out
3070 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3071 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3072 error = zfs_fill_zplprops(zc->zc_name, nvprops,
3073 zct.zct_zplprops, &is_insensitive);
3075 nvlist_free(nvprops);
3076 nvlist_free(zct.zct_zplprops);
3080 error = dmu_objset_create(zc->zc_name, type,
3081 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3082 nvlist_free(zct.zct_zplprops);
3086 * It would be nice to do this atomically.
3089 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3092 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3094 nvlist_free(nvprops);
3100 * zc_name name of filesystem
3101 * zc_value short name of snapshot
3102 * zc_cookie recursive flag
3103 * zc_nvlist_src[_size] property list
3106 * zc_value short snapname (i.e. part after the '@')
3109 zfs_ioc_snapshot(zfs_cmd_t *zc)
3111 nvlist_t *nvprops = NULL;
3113 boolean_t recursive = zc->zc_cookie;
3115 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3118 if (zc->zc_nvlist_src != 0 &&
3119 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3120 zc->zc_iflags, &nvprops)) != 0)
3123 error = zfs_check_userprops(zc->zc_name, nvprops);
3127 if (!nvlist_empty(nvprops) &&
3128 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3133 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3134 nvprops, recursive, B_FALSE, -1);
3137 nvlist_free(nvprops);
3143 * name dataset name, or when 'arg == NULL' the full snapshot name
3144 * arg short snapshot name (i.e. part after the '@')
3147 zfs_unmount_snap(const char *name, void *arg)
3149 zfs_sb_t *zsb = NULL;
3157 dsname = strdup(name);
3158 snapname = strdup(arg);
3160 ptr = strchr(name, '@');
3162 dsname = strdup(name);
3163 dsname[ptr - name] = '\0';
3164 snapname = strdup(ptr + 1);
3170 fullname = kmem_asprintf("%s@%s", dsname, snapname);
3172 error = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
3174 error = zfsctl_unmount_snapshot(zsb, fullname, MNT_FORCE);
3175 zfs_sb_rele(zsb, FTAG);
3177 /* Allow ENOENT for consistency with upstream */
3178 if (error == ENOENT)
3191 * zc_name name of filesystem, snaps must be under it
3192 * zc_nvlist_src[_size] full names of snapshots to destroy
3193 * zc_defer_destroy mark for deferred destroy
3196 * zc_name on failure, name of failed snapshot
3199 zfs_ioc_destroy_snaps_nvl(zfs_cmd_t *zc)
3205 if ((err = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3206 zc->zc_iflags, &nvl)) != 0)
3209 len = strlen(zc->zc_name);
3210 for (pair = nvlist_next_nvpair(nvl, NULL); pair != NULL;
3211 pair = nvlist_next_nvpair(nvl, pair)) {
3212 const char *name = nvpair_name(pair);
3214 * The snap name must be underneath the zc_name. This ensures
3215 * that our permission checks were legitimate.
3217 if (strncmp(zc->zc_name, name, len) != 0 ||
3218 (name[len] != '@' && name[len] != '/')) {
3223 (void) zfs_unmount_snap(name, NULL);
3224 (void) zvol_remove_minor(name);
3227 err = dmu_snapshots_destroy_nvl(nvl, zc->zc_defer_destroy,
3235 * zc_name name of dataset to destroy
3236 * zc_objset_type type of objset
3237 * zc_defer_destroy mark for deferred destroy
3242 zfs_ioc_destroy(zfs_cmd_t *zc)
3245 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3246 err = zfs_unmount_snap(zc->zc_name, NULL);
3251 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3252 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3253 (void) zvol_remove_minor(zc->zc_name);
3259 * zc_name name of dataset to rollback (to most recent snapshot)
3264 zfs_ioc_rollback(zfs_cmd_t *zc)
3266 dsl_dataset_t *ds, *clone;
3271 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3275 /* must not be a snapshot */
3276 if (dsl_dataset_is_snapshot(ds)) {
3277 dsl_dataset_rele(ds, FTAG);
3281 /* must have a most recent snapshot */
3282 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3283 dsl_dataset_rele(ds, FTAG);
3288 * Create clone of most recent snapshot.
3290 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3291 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3295 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3302 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
3303 error = zfs_suspend_fs(zsb);
3307 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3308 error = dsl_dataset_clone_swap(clone, ds,
3310 dsl_dataset_disown(ds, FTAG);
3315 resume_err = zfs_resume_fs(zsb, zc->zc_name);
3316 error = error ? error : resume_err;
3318 deactivate_super(zsb->z_sb);
3320 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3321 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3322 dsl_dataset_disown(ds, FTAG);
3330 * Destroy clone (which also closes it).
3332 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3335 strfree(clone_name);
3337 dsl_dataset_rele(ds, FTAG);
3343 * zc_name old name of dataset
3344 * zc_value new name of dataset
3345 * zc_cookie recursive flag (only valid for snapshots)
3350 zfs_ioc_rename(zfs_cmd_t *zc)
3352 boolean_t recursive = zc->zc_cookie & 1;
3355 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3356 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3357 strchr(zc->zc_value, '%'))
3361 * Unmount snapshot unless we're doing a recursive rename,
3362 * in which case the dataset code figures out which snapshots
3365 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3366 zc->zc_objset_type == DMU_OST_ZFS) {
3367 err = zfs_unmount_snap(zc->zc_name, NULL);
3372 err = dmu_objset_rename(zc->zc_name, zc->zc_value, recursive);
3373 if ((err == 0) && (zc->zc_objset_type == DMU_OST_ZVOL)) {
3374 (void) zvol_remove_minor(zc->zc_name);
3375 (void) zvol_create_minor(zc->zc_value);
3382 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3384 const char *propname = nvpair_name(pair);
3385 boolean_t issnap = (strchr(dsname, '@') != NULL);
3386 zfs_prop_t prop = zfs_name_to_prop(propname);
3390 if (prop == ZPROP_INVAL) {
3391 if (zfs_prop_user(propname)) {
3392 if ((err = zfs_secpolicy_write_perms(dsname,
3393 ZFS_DELEG_PERM_USERPROP, cr)))
3398 if (!issnap && zfs_prop_userquota(propname)) {
3399 const char *perm = NULL;
3400 const char *uq_prefix =
3401 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3402 const char *gq_prefix =
3403 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3405 if (strncmp(propname, uq_prefix,
3406 strlen(uq_prefix)) == 0) {
3407 perm = ZFS_DELEG_PERM_USERQUOTA;
3408 } else if (strncmp(propname, gq_prefix,
3409 strlen(gq_prefix)) == 0) {
3410 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3412 /* USERUSED and GROUPUSED are read-only */
3416 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3427 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3429 * dsl_prop_get_all_impl() returns properties in this
3433 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3434 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3439 * Check that this value is valid for this pool version
3442 case ZFS_PROP_COMPRESSION:
3444 * If the user specified gzip compression, make sure
3445 * the SPA supports it. We ignore any errors here since
3446 * we'll catch them later.
3448 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3449 nvpair_value_uint64(pair, &intval) == 0) {
3450 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3451 intval <= ZIO_COMPRESS_GZIP_9 &&
3452 zfs_earlier_version(dsname,
3453 SPA_VERSION_GZIP_COMPRESSION)) {
3457 if (intval == ZIO_COMPRESS_ZLE &&
3458 zfs_earlier_version(dsname,
3459 SPA_VERSION_ZLE_COMPRESSION))
3462 if (intval == ZIO_COMPRESS_LZ4) {
3463 zfeature_info_t *feature =
3465 SPA_FEATURE_LZ4_COMPRESS];
3468 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3471 if (!spa_feature_is_enabled(spa, feature)) {
3472 spa_close(spa, FTAG);
3475 spa_close(spa, FTAG);
3479 * If this is a bootable dataset then
3480 * verify that the compression algorithm
3481 * is supported for booting. We must return
3482 * something other than ENOTSUP since it
3483 * implies a downrev pool version.
3485 if (zfs_is_bootfs(dsname) &&
3486 !BOOTFS_COMPRESS_VALID(intval)) {
3492 case ZFS_PROP_COPIES:
3493 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3497 case ZFS_PROP_DEDUP:
3498 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3502 case ZFS_PROP_SHARESMB:
3503 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3507 case ZFS_PROP_ACLINHERIT:
3508 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3509 nvpair_value_uint64(pair, &intval) == 0) {
3510 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3511 zfs_earlier_version(dsname,
3512 SPA_VERSION_PASSTHROUGH_X))
3520 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3524 * Activates a feature on a pool in response to a property setting. This
3525 * creates a new sync task which modifies the pool to reflect the feature
3529 zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature)
3533 /* EBUSY here indicates that the feature is already active */
3534 err = dsl_sync_task_do(dp, zfs_prop_activate_feature_check,
3535 zfs_prop_activate_feature_sync, dp->dp_spa, feature, 2);
3537 if (err != 0 && err != EBUSY)
3544 * Checks for a race condition to make sure we don't increment a feature flag
3549 zfs_prop_activate_feature_check(void *arg1, void *arg2, dmu_tx_t *tx)
3552 zfeature_info_t *feature = arg2;
3554 if (!spa_feature_is_active(spa, feature))
3561 * The callback invoked on feature activation in the sync task caused by
3562 * zfs_prop_activate_feature.
3565 zfs_prop_activate_feature_sync(void *arg1, void *arg2, dmu_tx_t *tx)
3568 zfeature_info_t *feature = arg2;
3570 spa_feature_incr(spa, feature, tx);
3574 * Removes properties from the given props list that fail permission checks
3575 * needed to clear them and to restore them in case of a receive error. For each
3576 * property, make sure we have both set and inherit permissions.
3578 * Returns the first error encountered if any permission checks fail. If the
3579 * caller provides a non-NULL errlist, it also gives the complete list of names
3580 * of all the properties that failed a permission check along with the
3581 * corresponding error numbers. The caller is responsible for freeing the
3584 * If every property checks out successfully, zero is returned and the list
3585 * pointed at by errlist is NULL.
3588 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3591 nvpair_t *pair, *next_pair;
3598 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3600 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3601 (void) strcpy(zc->zc_name, dataset);
3602 pair = nvlist_next_nvpair(props, NULL);
3603 while (pair != NULL) {
3604 next_pair = nvlist_next_nvpair(props, pair);
3606 (void) strcpy(zc->zc_value, nvpair_name(pair));
3607 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3608 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3609 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3610 VERIFY(nvlist_add_int32(errors,
3611 zc->zc_value, err) == 0);
3615 kmem_free(zc, sizeof (zfs_cmd_t));
3617 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3618 nvlist_free(errors);
3621 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3624 if (errlist == NULL)
3625 nvlist_free(errors);
3633 propval_equals(nvpair_t *p1, nvpair_t *p2)
3635 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3636 /* dsl_prop_get_all_impl() format */
3638 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3639 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3643 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3645 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3646 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3650 if (nvpair_type(p1) != nvpair_type(p2))
3653 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3654 char *valstr1, *valstr2;
3656 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3657 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3658 return (strcmp(valstr1, valstr2) == 0);
3660 uint64_t intval1, intval2;
3662 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3663 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3664 return (intval1 == intval2);
3669 * Remove properties from props if they are not going to change (as determined
3670 * by comparison with origprops). Remove them from origprops as well, since we
3671 * do not need to clear or restore properties that won't change.
3674 props_reduce(nvlist_t *props, nvlist_t *origprops)
3676 nvpair_t *pair, *next_pair;
3678 if (origprops == NULL)
3679 return; /* all props need to be received */
3681 pair = nvlist_next_nvpair(props, NULL);
3682 while (pair != NULL) {
3683 const char *propname = nvpair_name(pair);
3686 next_pair = nvlist_next_nvpair(props, pair);
3688 if ((nvlist_lookup_nvpair(origprops, propname,
3689 &match) != 0) || !propval_equals(pair, match))
3690 goto next; /* need to set received value */
3692 /* don't clear the existing received value */
3693 (void) nvlist_remove_nvpair(origprops, match);
3694 /* don't bother receiving the property */
3695 (void) nvlist_remove_nvpair(props, pair);
3702 static boolean_t zfs_ioc_recv_inject_err;
3707 * zc_name name of containing filesystem
3708 * zc_nvlist_src{_size} nvlist of properties to apply
3709 * zc_value name of snapshot to create
3710 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3711 * zc_cookie file descriptor to recv from
3712 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3713 * zc_guid force flag
3714 * zc_cleanup_fd cleanup-on-exit file descriptor
3715 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3718 * zc_cookie number of bytes read
3719 * zc_nvlist_dst{_size} error for each unapplied received property
3720 * zc_obj zprop_errflags_t
3721 * zc_action_handle handle for this guid/ds mapping
3724 zfs_ioc_recv(zfs_cmd_t *zc)
3728 dmu_recv_cookie_t drc;
3729 boolean_t force = (boolean_t)zc->zc_guid;
3732 int props_error = 0;
3735 nvlist_t *props = NULL; /* sent properties */
3736 nvlist_t *origprops = NULL; /* existing properties */
3737 objset_t *origin = NULL;
3739 char tofs[ZFS_MAXNAMELEN];
3740 boolean_t first_recvd_props = B_FALSE;
3742 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3743 strchr(zc->zc_value, '@') == NULL ||
3744 strchr(zc->zc_value, '%'))
3747 (void) strcpy(tofs, zc->zc_value);
3748 tosnap = strchr(tofs, '@');
3751 if (zc->zc_nvlist_src != 0 &&
3752 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3753 zc->zc_iflags, &props)) != 0)
3763 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3765 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3766 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3767 !dsl_prop_get_hasrecvd(os)) {
3768 first_recvd_props = B_TRUE;
3772 * If new received properties are supplied, they are to
3773 * completely replace the existing received properties, so stash
3774 * away the existing ones.
3776 if (dsl_prop_get_received(os, &origprops) == 0) {
3777 nvlist_t *errlist = NULL;
3779 * Don't bother writing a property if its value won't
3780 * change (and avoid the unnecessary security checks).
3782 * The first receive after SPA_VERSION_RECVD_PROPS is a
3783 * special case where we blow away all local properties
3786 if (!first_recvd_props)
3787 props_reduce(props, origprops);
3788 if (zfs_check_clearable(tofs, origprops,
3790 (void) nvlist_merge(errors, errlist, 0);
3791 nvlist_free(errlist);
3794 dmu_objset_rele(os, FTAG);
3797 if (zc->zc_string[0]) {
3798 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3803 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3804 &zc->zc_begin_record, force, origin, &drc);
3806 dmu_objset_rele(origin, FTAG);
3811 * Set properties before we receive the stream so that they are applied
3812 * to the new data. Note that we must call dmu_recv_stream() if
3813 * dmu_recv_begin() succeeds.
3818 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3819 if (drc.drc_newfs) {
3820 if (spa_version(os->os_spa) >=
3821 SPA_VERSION_RECVD_PROPS)
3822 first_recvd_props = B_TRUE;
3823 } else if (origprops != NULL) {
3824 if (clear_received_props(os, tofs, origprops,
3825 first_recvd_props ? NULL : props) != 0)
3826 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3828 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3830 dsl_prop_set_hasrecvd(os);
3831 } else if (!drc.drc_newfs) {
3832 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3835 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3837 (void) nvlist_merge(errors, errlist, 0);
3838 nvlist_free(errlist);
3841 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3843 * Caller made zc->zc_nvlist_dst less than the minimum expected
3844 * size or supplied an invalid address.
3846 props_error = EINVAL;
3850 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3851 &zc->zc_action_handle);
3854 zfs_sb_t *zsb = NULL;
3856 if (get_zfs_sb(tofs, &zsb) == 0) {
3860 error = zfs_suspend_fs(zsb);
3862 * If the suspend fails, then the recv_end will
3863 * likely also fail, and clean up after itself.
3865 end_err = dmu_recv_end(&drc);
3867 error = zfs_resume_fs(zsb, tofs);
3868 error = error ? error : end_err;
3869 deactivate_super(zsb->z_sb);
3871 error = dmu_recv_end(&drc);
3875 zc->zc_cookie = off - fp->f_offset;
3876 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3880 if (zfs_ioc_recv_inject_err) {
3881 zfs_ioc_recv_inject_err = B_FALSE;
3886 * On error, restore the original props.
3888 if (error && props) {
3889 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3890 if (clear_received_props(os, tofs, props, NULL) != 0) {
3892 * We failed to clear the received properties.
3893 * Since we may have left a $recvd value on the
3894 * system, we can't clear the $hasrecvd flag.
3896 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3897 } else if (first_recvd_props) {
3898 dsl_prop_unset_hasrecvd(os);
3900 dmu_objset_rele(os, FTAG);
3901 } else if (!drc.drc_newfs) {
3902 /* We failed to clear the received properties. */
3903 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3906 if (origprops == NULL && !drc.drc_newfs) {
3907 /* We failed to stash the original properties. */
3908 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3912 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3913 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3914 * explictly if we're restoring local properties cleared in the
3915 * first new-style receive.
3917 if (origprops != NULL &&
3918 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3919 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3920 origprops, NULL) != 0) {
3922 * We stashed the original properties but failed to
3925 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3930 nvlist_free(origprops);
3931 nvlist_free(errors);
3935 error = props_error;
3942 * zc_name name of snapshot to send
3943 * zc_cookie file descriptor to send stream to
3944 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3945 * zc_sendobj objsetid of snapshot to send
3946 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3947 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
3948 * output size in zc_objset_type.
3953 zfs_ioc_send(zfs_cmd_t *zc)
3955 objset_t *fromsnap = NULL;
3960 dsl_dataset_t *dsfrom = NULL;
3963 boolean_t estimate = (zc->zc_guid != 0);
3965 error = spa_open(zc->zc_name, &spa, FTAG);
3969 dp = spa_get_dsl(spa);
3970 rw_enter(&dp->dp_config_rwlock, RW_READER);
3971 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3972 rw_exit(&dp->dp_config_rwlock);
3974 spa_close(spa, FTAG);
3978 error = dmu_objset_from_ds(ds, &tosnap);
3980 dsl_dataset_rele(ds, FTAG);
3981 spa_close(spa, FTAG);
3985 if (zc->zc_fromobj != 0) {
3986 rw_enter(&dp->dp_config_rwlock, RW_READER);
3987 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3988 rw_exit(&dp->dp_config_rwlock);
3989 spa_close(spa, FTAG);
3991 dsl_dataset_rele(ds, FTAG);
3994 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3996 dsl_dataset_rele(dsfrom, FTAG);
3997 dsl_dataset_rele(ds, FTAG);
4001 spa_close(spa, FTAG);
4005 error = dmu_send_estimate(tosnap, fromsnap, zc->zc_obj,
4006 &zc->zc_objset_type);
4008 file_t *fp = getf(zc->zc_cookie);
4010 dsl_dataset_rele(ds, FTAG);
4012 dsl_dataset_rele(dsfrom, FTAG);
4017 error = dmu_send(tosnap, fromsnap, zc->zc_obj,
4018 zc->zc_cookie, fp->f_vnode, &off);
4020 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4022 releasef(zc->zc_cookie);
4025 dsl_dataset_rele(dsfrom, FTAG);
4026 dsl_dataset_rele(ds, FTAG);
4032 * zc_name name of snapshot on which to report progress
4033 * zc_cookie file descriptor of send stream
4036 * zc_cookie number of bytes written in send stream thus far
4039 zfs_ioc_send_progress(zfs_cmd_t *zc)
4042 dmu_sendarg_t *dsp = NULL;
4045 if ((error = dsl_dataset_hold(zc->zc_name, FTAG, &ds)) != 0)
4048 mutex_enter(&ds->ds_sendstream_lock);
4051 * Iterate over all the send streams currently active on this dataset.
4052 * If there's one which matches the specified file descriptor _and_ the
4053 * stream was started by the current process, return the progress of
4057 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4058 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4059 if (dsp->dsa_outfd == zc->zc_cookie &&
4060 dsp->dsa_proc->group_leader == curproc->group_leader)
4065 zc->zc_cookie = *(dsp->dsa_off);
4069 mutex_exit(&ds->ds_sendstream_lock);
4070 dsl_dataset_rele(ds, FTAG);
4075 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4079 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4080 &zc->zc_inject_record);
4083 zc->zc_guid = (uint64_t)id;
4089 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4091 return (zio_clear_fault((int)zc->zc_guid));
4095 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4097 int id = (int)zc->zc_guid;
4100 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4101 &zc->zc_inject_record);
4109 zfs_ioc_error_log(zfs_cmd_t *zc)
4113 size_t count = (size_t)zc->zc_nvlist_dst_size;
4115 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4118 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4121 zc->zc_nvlist_dst_size = count;
4123 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4125 spa_close(spa, FTAG);
4131 zfs_ioc_clear(zfs_cmd_t *zc)
4138 * On zpool clear we also fix up missing slogs
4140 mutex_enter(&spa_namespace_lock);
4141 spa = spa_lookup(zc->zc_name);
4143 mutex_exit(&spa_namespace_lock);
4146 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4147 /* we need to let spa_open/spa_load clear the chains */
4148 spa_set_log_state(spa, SPA_LOG_CLEAR);
4150 spa->spa_last_open_failed = 0;
4151 mutex_exit(&spa_namespace_lock);
4153 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4154 error = spa_open(zc->zc_name, &spa, FTAG);
4157 nvlist_t *config = NULL;
4159 if (zc->zc_nvlist_src == 0)
4162 if ((error = get_nvlist(zc->zc_nvlist_src,
4163 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4164 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4166 if (config != NULL) {
4169 if ((err = put_nvlist(zc, config)) != 0)
4171 nvlist_free(config);
4173 nvlist_free(policy);
4180 spa_vdev_state_enter(spa, SCL_NONE);
4182 if (zc->zc_guid == 0) {
4185 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4187 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4188 spa_close(spa, FTAG);
4193 vdev_clear(spa, vd);
4195 (void) spa_vdev_state_exit(spa, NULL, 0);
4198 * Resume any suspended I/Os.
4200 if (zio_resume(spa) != 0)
4203 spa_close(spa, FTAG);
4209 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4214 error = spa_open(zc->zc_name, &spa, FTAG);
4218 spa_vdev_state_enter(spa, SCL_NONE);
4221 * If a resilver is already in progress then set the
4222 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4223 * the scan as a side effect of the reopen. Otherwise, let
4224 * vdev_open() decided if a resilver is required.
4226 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4227 vdev_reopen(spa->spa_root_vdev);
4228 spa->spa_scrub_reopen = B_FALSE;
4230 (void) spa_vdev_state_exit(spa, NULL, 0);
4231 spa_close(spa, FTAG);
4236 * zc_name name of filesystem
4237 * zc_value name of origin snapshot
4240 * zc_string name of conflicting snapshot, if there is one
4243 zfs_ioc_promote(zfs_cmd_t *zc)
4248 * We don't need to unmount *all* the origin fs's snapshots, but
4251 cp = strchr(zc->zc_value, '@');
4254 (void) dmu_objset_find(zc->zc_value,
4255 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4256 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4260 * Retrieve a single {user|group}{used|quota}@... property.
4263 * zc_name name of filesystem
4264 * zc_objset_type zfs_userquota_prop_t
4265 * zc_value domain name (eg. "S-1-234-567-89")
4266 * zc_guid RID/UID/GID
4269 * zc_cookie property value
4272 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4277 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4280 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4284 error = zfs_userspace_one(zsb,
4285 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4286 zfs_sb_rele(zsb, FTAG);
4293 * zc_name name of filesystem
4294 * zc_cookie zap cursor
4295 * zc_objset_type zfs_userquota_prop_t
4296 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4299 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4300 * zc_cookie zap cursor
4303 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4306 int bufsize = zc->zc_nvlist_dst_size;
4313 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4317 buf = vmem_alloc(bufsize, KM_SLEEP);
4319 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4320 buf, &zc->zc_nvlist_dst_size);
4323 error = xcopyout(buf,
4324 (void *)(uintptr_t)zc->zc_nvlist_dst,
4325 zc->zc_nvlist_dst_size);
4327 vmem_free(buf, bufsize);
4328 zfs_sb_rele(zsb, FTAG);
4335 * zc_name name of filesystem
4341 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4347 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4348 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4350 * If userused is not enabled, it may be because the
4351 * objset needs to be closed & reopened (to grow the
4352 * objset_phys_t). Suspend/resume the fs will do that.
4354 error = zfs_suspend_fs(zsb);
4356 error = zfs_resume_fs(zsb, zc->zc_name);
4359 error = dmu_objset_userspace_upgrade(zsb->z_os);
4360 deactivate_super(zsb->z_sb);
4362 /* XXX kind of reading contents without owning */
4363 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4367 error = dmu_objset_userspace_upgrade(os);
4368 dmu_objset_rele(os, FTAG);
4375 zfs_ioc_share(zfs_cmd_t *zc)
4380 ace_t full_access[] = {
4381 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4386 * zc_name name of containing filesystem
4387 * zc_obj object # beyond which we want next in-use object #
4390 * zc_obj next in-use object #
4393 zfs_ioc_next_obj(zfs_cmd_t *zc)
4395 objset_t *os = NULL;
4398 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4402 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4403 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4405 dmu_objset_rele(os, FTAG);
4411 * zc_name name of filesystem
4412 * zc_value prefix name for snapshot
4413 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4418 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4423 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4424 (u_longlong_t)ddi_get_lbolt64());
4426 if (strlen(snap_name) >= MAXNAMELEN) {
4431 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4432 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4438 (void) strcpy(zc->zc_value, snap_name);
4445 * zc_name name of "to" snapshot
4446 * zc_value name of "from" snapshot
4447 * zc_cookie file descriptor to write diff data on
4450 * dmu_diff_record_t's to the file descriptor
4453 zfs_ioc_diff(zfs_cmd_t *zc)
4461 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4465 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4467 dmu_objset_rele(tosnap, FTAG);
4471 fp = getf(zc->zc_cookie);
4473 dmu_objset_rele(fromsnap, FTAG);
4474 dmu_objset_rele(tosnap, FTAG);
4480 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4482 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4484 releasef(zc->zc_cookie);
4486 dmu_objset_rele(fromsnap, FTAG);
4487 dmu_objset_rele(tosnap, FTAG);
4492 * Remove all ACL files in shares dir
4494 #ifdef HAVE_SMB_SHARE
4496 zfs_smb_acl_purge(znode_t *dzp)
4499 zap_attribute_t zap;
4500 zfs_sb_t *zsb = ZTOZSB(dzp);
4503 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4504 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4505 zap_cursor_advance(&zc)) {
4506 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4510 zap_cursor_fini(&zc);
4513 #endif /* HAVE_SMB_SHARE */
4516 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4518 #ifdef HAVE_SMB_SHARE
4521 vnode_t *resourcevp = NULL;
4530 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4531 NO_FOLLOW, NULL, &vp)) != 0)
4534 /* Now make sure mntpnt and dataset are ZFS */
4536 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4537 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4538 zc->zc_name) != 0)) {
4548 * Create share dir if its missing.
4550 mutex_enter(&zsb->z_lock);
4551 if (zsb->z_shares_dir == 0) {
4554 tx = dmu_tx_create(zsb->z_os);
4555 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4557 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4558 error = dmu_tx_assign(tx, TXG_WAIT);
4562 error = zfs_create_share_dir(zsb, tx);
4566 mutex_exit(&zsb->z_lock);
4572 mutex_exit(&zsb->z_lock);
4574 ASSERT(zsb->z_shares_dir);
4575 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4581 switch (zc->zc_cookie) {
4582 case ZFS_SMB_ACL_ADD:
4583 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4584 vattr.va_mode = S_IFREG|0777;
4588 vsec.vsa_mask = VSA_ACE;
4589 vsec.vsa_aclentp = &full_access;
4590 vsec.vsa_aclentsz = sizeof (full_access);
4591 vsec.vsa_aclcnt = 1;
4593 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4594 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4596 VN_RELE(resourcevp);
4599 case ZFS_SMB_ACL_REMOVE:
4600 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4604 case ZFS_SMB_ACL_RENAME:
4605 if ((error = get_nvlist(zc->zc_nvlist_src,
4606 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4611 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4612 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4615 VN_RELE(ZTOV(sharedir));
4617 nvlist_free(nvlist);
4620 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4622 nvlist_free(nvlist);
4625 case ZFS_SMB_ACL_PURGE:
4626 error = zfs_smb_acl_purge(sharedir);
4635 VN_RELE(ZTOV(sharedir));
4642 #endif /* HAVE_SMB_SHARE */
4647 * zc_name name of filesystem
4648 * zc_value short name of snap
4649 * zc_string user-supplied tag for this hold
4650 * zc_cookie recursive flag
4651 * zc_temphold set if hold is temporary
4652 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4653 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4654 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4659 zfs_ioc_hold(zfs_cmd_t *zc)
4661 boolean_t recursive = zc->zc_cookie;
4668 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4671 if (zc->zc_sendobj == 0) {
4672 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4673 zc->zc_string, recursive, zc->zc_temphold,
4674 zc->zc_cleanup_fd));
4680 error = spa_open(zc->zc_name, &spa, FTAG);
4684 dp = spa_get_dsl(spa);
4685 rw_enter(&dp->dp_config_rwlock, RW_READER);
4686 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4687 rw_exit(&dp->dp_config_rwlock);
4688 spa_close(spa, FTAG);
4693 * Until we have a hold on this snapshot, it's possible that
4694 * zc_sendobj could've been destroyed and reused as part
4695 * of a later txg. Make sure we're looking at the right object.
4697 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4698 dsl_dataset_rele(ds, FTAG);
4702 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4703 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4705 dsl_dataset_rele(ds, FTAG);
4710 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4714 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4717 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4719 dsl_dataset_rele(ds, FTAG);
4726 * zc_name name of dataset from which we're releasing a user hold
4727 * zc_value short name of snap
4728 * zc_string user-supplied tag for this hold
4729 * zc_cookie recursive flag
4734 zfs_ioc_release(zfs_cmd_t *zc)
4736 boolean_t recursive = zc->zc_cookie;
4738 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4741 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4742 zc->zc_string, recursive));
4747 * zc_name name of filesystem
4750 * zc_nvlist_src{_size} nvlist of snapshot holds
4753 zfs_ioc_get_holds(zfs_cmd_t *zc)
4758 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4759 error = put_nvlist(zc, nvp);
4768 * zc_guid flags (ZEVENT_NONBLOCK)
4771 * zc_nvlist_dst next nvlist event
4772 * zc_cookie dropped events since last get
4773 * zc_cleanup_fd cleanup-on-exit file descriptor
4776 zfs_ioc_events_next(zfs_cmd_t *zc)
4779 nvlist_t *event = NULL;
4781 uint64_t dropped = 0;
4784 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4789 error = zfs_zevent_next(ze, &event,
4790 &zc->zc_nvlist_dst_size, &dropped);
4791 if (event != NULL) {
4792 zc->zc_cookie = dropped;
4793 error = put_nvlist(zc, event);
4797 if (zc->zc_guid & ZEVENT_NONBLOCK)
4800 if ((error == 0) || (error != ENOENT))
4803 error = zfs_zevent_wait(ze);
4808 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4815 * zc_cookie cleared events count
4818 zfs_ioc_events_clear(zfs_cmd_t *zc)
4822 zfs_zevent_drain_all(&count);
4823 zc->zc_cookie = count;
4830 * zc_name name of new filesystem or snapshot
4831 * zc_value full name of old snapshot
4834 * zc_cookie space in bytes
4835 * zc_objset_type compressed space in bytes
4836 * zc_perm_action uncompressed space in bytes
4839 zfs_ioc_space_written(zfs_cmd_t *zc)
4842 dsl_dataset_t *new, *old;
4844 error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
4847 error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
4849 dsl_dataset_rele(new, FTAG);
4853 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
4854 &zc->zc_objset_type, &zc->zc_perm_action);
4855 dsl_dataset_rele(old, FTAG);
4856 dsl_dataset_rele(new, FTAG);
4862 * zc_name full name of last snapshot
4863 * zc_value full name of first snapshot
4866 * zc_cookie space in bytes
4867 * zc_objset_type compressed space in bytes
4868 * zc_perm_action uncompressed space in bytes
4871 zfs_ioc_space_snaps(zfs_cmd_t *zc)
4874 dsl_dataset_t *new, *old;
4876 error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
4879 error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
4881 dsl_dataset_rele(new, FTAG);
4885 error = dsl_dataset_space_wouldfree(old, new, &zc->zc_cookie,
4886 &zc->zc_objset_type, &zc->zc_perm_action);
4887 dsl_dataset_rele(old, FTAG);
4888 dsl_dataset_rele(new, FTAG);
4893 * pool create, destroy, and export don't log the history as part of
4894 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4895 * do the logging of those commands.
4897 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4898 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4900 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4902 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4904 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4906 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4908 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4910 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4912 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4913 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4914 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4915 POOL_CHECK_READONLY },
4916 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4917 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4918 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4920 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4921 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4922 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4923 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4924 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4925 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4926 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4927 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4928 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4929 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4930 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4931 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4932 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4933 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4934 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4935 POOL_CHECK_SUSPENDED },
4936 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4938 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4939 POOL_CHECK_SUSPENDED },
4940 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4941 POOL_CHECK_SUSPENDED },
4942 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4943 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4944 { zfs_ioc_create_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4946 { zfs_ioc_remove_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4948 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4949 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4950 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4951 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4952 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4953 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4954 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4955 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4956 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4957 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4958 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_FALSE,
4960 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4962 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4964 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4966 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4968 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4970 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4971 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4972 { zfs_ioc_destroy_snaps_nvl, zfs_secpolicy_destroy_recursive,
4973 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4974 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4975 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4976 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4978 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4979 POOL_CHECK_SUSPENDED },
4980 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4981 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4982 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4984 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4985 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4986 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4988 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4990 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4991 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4992 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4994 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4995 B_FALSE, POOL_CHECK_NONE },
4996 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
4997 B_FALSE, POOL_CHECK_NONE },
4998 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4999 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5000 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
5001 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5002 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
5003 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5004 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5005 POOL_CHECK_SUSPENDED },
5006 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5008 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5009 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5010 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5012 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
5014 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
5015 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5016 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
5017 POOL_CHECK_SUSPENDED },
5018 { zfs_ioc_events_next, zfs_secpolicy_config, NO_NAME, B_FALSE,
5020 { zfs_ioc_events_clear, zfs_secpolicy_config, NO_NAME, B_FALSE,
5022 { zfs_ioc_pool_reguid, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5023 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5024 { zfs_ioc_space_written, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5025 POOL_CHECK_SUSPENDED },
5026 { zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5027 POOL_CHECK_SUSPENDED },
5028 { zfs_ioc_pool_reopen, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5029 POOL_CHECK_SUSPENDED },
5030 { zfs_ioc_send_progress, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5035 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5036 zfs_ioc_poolcheck_t check)
5041 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5043 if (check & POOL_CHECK_NONE)
5046 error = spa_open(name, &spa, FTAG);
5048 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5050 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5052 spa_close(spa, FTAG);
5058 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
5062 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5064 for (zs = list_head(&zfsdev_state_list); zs != NULL;
5065 zs = list_next(&zfsdev_state_list, zs)) {
5066 if (zs->zs_minor == minor) {
5068 case ZST_ONEXIT: return (zs->zs_onexit);
5069 case ZST_ZEVENT: return (zs->zs_zevent);
5070 case ZST_ALL: return (zs);
5079 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
5083 mutex_enter(&zfsdev_state_lock);
5084 ptr = zfsdev_get_state_impl(minor, which);
5085 mutex_exit(&zfsdev_state_lock);
5091 zfsdev_getminor(struct file *filp)
5093 ASSERT(filp != NULL);
5094 ASSERT(filp->private_data != NULL);
5096 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
5100 * Find a free minor number. The zfsdev_state_list is expected to
5101 * be short since it is only a list of currently open file handles.
5104 zfsdev_minor_alloc(void)
5106 static minor_t last_minor = 0;
5109 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5111 for (m = last_minor + 1; m != last_minor; m++) {
5112 if (m > ZFSDEV_MAX_MINOR)
5114 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
5124 zfsdev_state_init(struct file *filp)
5129 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5131 minor = zfsdev_minor_alloc();
5135 zs = kmem_zalloc( sizeof(zfsdev_state_t), KM_SLEEP);
5140 zs->zs_minor = minor;
5141 filp->private_data = zs;
5143 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
5144 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
5146 list_insert_tail(&zfsdev_state_list, zs);
5152 zfsdev_state_destroy(struct file *filp)
5156 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5157 ASSERT(filp->private_data != NULL);
5159 zs = filp->private_data;
5160 zfs_onexit_destroy(zs->zs_onexit);
5161 zfs_zevent_destroy(zs->zs_zevent);
5163 list_remove(&zfsdev_state_list, zs);
5164 kmem_free(zs, sizeof(zfsdev_state_t));
5170 zfsdev_open(struct inode *ino, struct file *filp)
5174 mutex_enter(&zfsdev_state_lock);
5175 error = zfsdev_state_init(filp);
5176 mutex_exit(&zfsdev_state_lock);
5182 zfsdev_release(struct inode *ino, struct file *filp)
5186 mutex_enter(&zfsdev_state_lock);
5187 error = zfsdev_state_destroy(filp);
5188 mutex_exit(&zfsdev_state_lock);
5194 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5198 int error, rc, flag = 0;
5200 vec = cmd - ZFS_IOC;
5201 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5204 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
5206 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5210 if ((error == 0) && !(flag & FKIOCTL))
5211 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, CRED());
5214 * Ensure that all pool/dataset names are valid before we pass down to
5218 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5219 zc->zc_iflags = flag & FKIOCTL;
5220 switch (zfs_ioc_vec[vec].zvec_namecheck) {
5222 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5224 error = pool_status_check(zc->zc_name,
5225 zfs_ioc_vec[vec].zvec_namecheck,
5226 zfs_ioc_vec[vec].zvec_pool_check);
5230 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5232 error = pool_status_check(zc->zc_name,
5233 zfs_ioc_vec[vec].zvec_namecheck,
5234 zfs_ioc_vec[vec].zvec_pool_check);
5243 error = zfs_ioc_vec[vec].zvec_func(zc);
5245 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5249 if (zfs_ioc_vec[vec].zvec_his_log)
5250 zfs_log_history(zc);
5253 kmem_free(zc, sizeof (zfs_cmd_t));
5257 #ifdef CONFIG_COMPAT
5259 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5261 return zfsdev_ioctl(filp, cmd, arg);
5264 #define zfsdev_compat_ioctl NULL
5267 static const struct file_operations zfsdev_fops = {
5268 .open = zfsdev_open,
5269 .release = zfsdev_release,
5270 .unlocked_ioctl = zfsdev_ioctl,
5271 .compat_ioctl = zfsdev_compat_ioctl,
5272 .owner = THIS_MODULE,
5275 static struct miscdevice zfs_misc = {
5276 .minor = MISC_DYNAMIC_MINOR,
5278 .fops = &zfsdev_fops,
5286 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
5287 list_create(&zfsdev_state_list, sizeof (zfsdev_state_t),
5288 offsetof(zfsdev_state_t, zs_next));
5290 error = misc_register(&zfs_misc);
5292 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
5304 error = misc_deregister(&zfs_misc);
5306 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5308 mutex_destroy(&zfsdev_state_lock);
5309 list_destroy(&zfsdev_state_list);
5312 uint_t zfs_fsyncer_key;
5313 extern uint_t rrw_tsd_key;
5316 #define ZFS_DEBUG_STR " (DEBUG mode)"
5318 #define ZFS_DEBUG_STR ""
5326 spa_init(FREAD | FWRITE);
5329 if ((error = zvol_init()) != 0)
5332 if ((error = zfs_attach()) != 0)
5335 tsd_create(&zfs_fsyncer_key, NULL);
5336 tsd_create(&rrw_tsd_key, NULL);
5338 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
5339 "ZFS pool version %s, ZFS filesystem version %s\n",
5340 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
5341 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5350 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
5351 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
5352 ZFS_DEBUG_STR, error);
5365 tsd_destroy(&zfs_fsyncer_key);
5366 tsd_destroy(&rrw_tsd_key);
5368 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
5369 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
5375 spl_module_init(_init);
5376 spl_module_exit(_fini);
5378 MODULE_DESCRIPTION("ZFS");
5379 MODULE_AUTHOR(ZFS_META_AUTHOR);
5380 MODULE_LICENSE(ZFS_META_LICENSE);
5381 #endif /* HAVE_SPL */