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_SNAPDEV:
2188 err = zvol_set_snapdev(dsname, intval);
2190 case ZFS_PROP_VERSION:
2194 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2197 err = zfs_set_version(zsb, intval);
2198 zfs_sb_rele(zsb, FTAG);
2200 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2203 zc = kmem_zalloc(sizeof (zfs_cmd_t),
2204 KM_SLEEP | KM_NODEBUG);
2205 (void) strcpy(zc->zc_name, dsname);
2206 (void) zfs_ioc_userspace_upgrade(zc);
2207 kmem_free(zc, sizeof (zfs_cmd_t));
2211 case ZFS_PROP_COMPRESSION:
2213 if (intval == ZIO_COMPRESS_LZ4) {
2214 zfeature_info_t *feature =
2215 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2219 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2222 dp = spa->spa_dsl_pool;
2225 * Setting the LZ4 compression algorithm activates
2228 if (!spa_feature_is_active(spa, feature)) {
2229 if ((err = zfs_prop_activate_feature(dp,
2231 spa_close(spa, FTAG);
2236 spa_close(spa, FTAG);
2239 * We still want the default set action to be performed in the
2240 * caller, we only performed zfeature settings here.
2254 * This function is best effort. If it fails to set any of the given properties,
2255 * it continues to set as many as it can and returns the first error
2256 * encountered. If the caller provides a non-NULL errlist, it also gives the
2257 * complete list of names of all the properties it failed to set along with the
2258 * corresponding error numbers. The caller is responsible for freeing the
2261 * If every property is set successfully, zero is returned and the list pointed
2262 * at by errlist is NULL.
2265 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2273 nvlist_t *genericnvl;
2277 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2278 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2279 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2283 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2284 const char *propname = nvpair_name(pair);
2285 zfs_prop_t prop = zfs_name_to_prop(propname);
2288 /* decode the property value */
2290 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2292 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2293 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2298 /* Validate value type */
2299 if (err == 0 && prop == ZPROP_INVAL) {
2300 if (zfs_prop_user(propname)) {
2301 if (nvpair_type(propval) != DATA_TYPE_STRING)
2303 } else if (zfs_prop_userquota(propname)) {
2304 if (nvpair_type(propval) !=
2305 DATA_TYPE_UINT64_ARRAY)
2310 } else if (err == 0) {
2311 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2312 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2314 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2317 VERIFY(nvpair_value_uint64(propval,
2320 switch (zfs_prop_get_type(prop)) {
2321 case PROP_TYPE_NUMBER:
2323 case PROP_TYPE_STRING:
2326 case PROP_TYPE_INDEX:
2327 if (zfs_prop_index_to_string(prop,
2328 intval, &unused) != 0)
2333 "unknown property type");
2340 /* Validate permissions */
2342 err = zfs_check_settable(dsname, pair, CRED());
2345 err = zfs_prop_set_special(dsname, source, pair);
2348 * For better performance we build up a list of
2349 * properties to set in a single transaction.
2351 err = nvlist_add_nvpair(genericnvl, pair);
2352 } else if (err != 0 && nvl != retrynvl) {
2354 * This may be a spurious error caused by
2355 * receiving quota and reservation out of order.
2356 * Try again in a second pass.
2358 err = nvlist_add_nvpair(retrynvl, pair);
2363 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2366 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2371 if (!nvlist_empty(genericnvl) &&
2372 dsl_props_set(dsname, source, genericnvl) != 0) {
2374 * If this fails, we still want to set as many properties as we
2375 * can, so try setting them individually.
2378 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2379 const char *propname = nvpair_name(pair);
2383 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2385 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2386 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2390 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2391 VERIFY(nvpair_value_string(propval,
2393 err = dsl_prop_set(dsname, propname, source, 1,
2394 strlen(strval) + 1, strval);
2396 VERIFY(nvpair_value_uint64(propval,
2398 err = dsl_prop_set(dsname, propname, source, 8,
2403 VERIFY(nvlist_add_int32(errors, propname,
2408 nvlist_free(genericnvl);
2409 nvlist_free(retrynvl);
2411 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2412 nvlist_free(errors);
2415 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2418 if (errlist == NULL)
2419 nvlist_free(errors);
2427 * Check that all the properties are valid user properties.
2430 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2432 nvpair_t *pair = NULL;
2435 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2436 const char *propname = nvpair_name(pair);
2439 if (!zfs_prop_user(propname) ||
2440 nvpair_type(pair) != DATA_TYPE_STRING)
2443 if ((error = zfs_secpolicy_write_perms(fsname,
2444 ZFS_DELEG_PERM_USERPROP, CRED())))
2447 if (strlen(propname) >= ZAP_MAXNAMELEN)
2448 return (ENAMETOOLONG);
2450 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2451 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2458 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2462 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2465 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2466 if (nvlist_exists(skipped, nvpair_name(pair)))
2469 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2474 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2478 nvlist_t *cleared_props = NULL;
2479 props_skip(props, skipped, &cleared_props);
2480 if (!nvlist_empty(cleared_props)) {
2482 * Acts on local properties until the dataset has received
2483 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2485 zprop_source_t flags = (ZPROP_SRC_NONE |
2486 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2487 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2489 nvlist_free(cleared_props);
2495 * zc_name name of filesystem
2496 * zc_value name of property to set
2497 * zc_nvlist_src{_size} nvlist of properties to apply
2498 * zc_cookie received properties flag
2501 * zc_nvlist_dst{_size} error for each unapplied received property
2504 zfs_ioc_set_prop(zfs_cmd_t *zc)
2507 boolean_t received = zc->zc_cookie;
2508 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2510 nvlist_t *errors = NULL;
2513 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2514 zc->zc_iflags, &nvl)) != 0)
2518 nvlist_t *origprops;
2521 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2522 if (dsl_prop_get_received(os, &origprops) == 0) {
2523 (void) clear_received_props(os,
2524 zc->zc_name, origprops, nvl);
2525 nvlist_free(origprops);
2528 dsl_prop_set_hasrecvd(os);
2529 dmu_objset_rele(os, FTAG);
2533 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2535 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2536 (void) put_nvlist(zc, errors);
2539 nvlist_free(errors);
2546 * zc_name name of filesystem
2547 * zc_value name of property to inherit
2548 * zc_cookie revert to received value if TRUE
2553 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2555 const char *propname = zc->zc_value;
2556 zfs_prop_t prop = zfs_name_to_prop(propname);
2557 boolean_t received = zc->zc_cookie;
2558 zprop_source_t source = (received
2559 ? ZPROP_SRC_NONE /* revert to received value, if any */
2560 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2569 * zfs_prop_set_special() expects properties in the form of an
2570 * nvpair with type info.
2572 if (prop == ZPROP_INVAL) {
2573 if (!zfs_prop_user(propname))
2576 type = PROP_TYPE_STRING;
2577 } else if (prop == ZFS_PROP_VOLSIZE ||
2578 prop == ZFS_PROP_VERSION) {
2581 type = zfs_prop_get_type(prop);
2584 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2587 case PROP_TYPE_STRING:
2588 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2590 case PROP_TYPE_NUMBER:
2591 case PROP_TYPE_INDEX:
2592 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2599 pair = nvlist_next_nvpair(dummy, NULL);
2600 err = zfs_prop_set_special(zc->zc_name, source, pair);
2603 return (err); /* special property already handled */
2606 * Only check this in the non-received case. We want to allow
2607 * 'inherit -S' to revert non-inheritable properties like quota
2608 * and reservation to the received or default values even though
2609 * they are not considered inheritable.
2611 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2615 /* the property name has been validated by zfs_secpolicy_inherit() */
2616 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2620 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2627 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2628 zc->zc_iflags, &props)))
2632 * If the only property is the configfile, then just do a spa_lookup()
2633 * to handle the faulted case.
2635 pair = nvlist_next_nvpair(props, NULL);
2636 if (pair != NULL && strcmp(nvpair_name(pair),
2637 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2638 nvlist_next_nvpair(props, pair) == NULL) {
2639 mutex_enter(&spa_namespace_lock);
2640 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2641 spa_configfile_set(spa, props, B_FALSE);
2642 spa_config_sync(spa, B_FALSE, B_TRUE);
2644 mutex_exit(&spa_namespace_lock);
2651 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2656 error = spa_prop_set(spa, props);
2659 spa_close(spa, FTAG);
2665 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2669 nvlist_t *nvp = NULL;
2671 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2673 * If the pool is faulted, there may be properties we can still
2674 * get (such as altroot and cachefile), so attempt to get them
2677 mutex_enter(&spa_namespace_lock);
2678 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2679 error = spa_prop_get(spa, &nvp);
2680 mutex_exit(&spa_namespace_lock);
2682 error = spa_prop_get(spa, &nvp);
2683 spa_close(spa, FTAG);
2686 if (error == 0 && zc->zc_nvlist_dst != 0)
2687 error = put_nvlist(zc, nvp);
2697 * zc_name name of volume
2702 zfs_ioc_create_minor(zfs_cmd_t *zc)
2704 return (zvol_create_minor(zc->zc_name));
2709 * zc_name name of volume
2714 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2716 return (zvol_remove_minor(zc->zc_name));
2721 * zc_name name of filesystem
2722 * zc_nvlist_src{_size} nvlist of delegated permissions
2723 * zc_perm_action allow/unallow flag
2728 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2731 nvlist_t *fsaclnv = NULL;
2733 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2734 zc->zc_iflags, &fsaclnv)) != 0)
2738 * Verify nvlist is constructed correctly
2740 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2741 nvlist_free(fsaclnv);
2746 * If we don't have PRIV_SYS_MOUNT, then validate
2747 * that user is allowed to hand out each permission in
2751 error = secpolicy_zfs(CRED());
2753 if (zc->zc_perm_action == B_FALSE) {
2754 error = dsl_deleg_can_allow(zc->zc_name,
2757 error = dsl_deleg_can_unallow(zc->zc_name,
2763 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2765 nvlist_free(fsaclnv);
2771 * zc_name name of filesystem
2774 * zc_nvlist_src{_size} nvlist of delegated permissions
2777 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2782 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2783 error = put_nvlist(zc, nvp);
2792 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2794 zfs_creat_t *zct = arg;
2796 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2799 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2803 * createprops list of properties requested by creator
2804 * default_zplver zpl version to use if unspecified in createprops
2805 * fuids_ok fuids allowed in this version of the spa?
2806 * os parent objset pointer (NULL if root fs)
2809 * zplprops values for the zplprops we attach to the master node object
2810 * is_ci true if requested file system will be purely case-insensitive
2812 * Determine the settings for utf8only, normalization and
2813 * casesensitivity. Specific values may have been requested by the
2814 * creator and/or we can inherit values from the parent dataset. If
2815 * the file system is of too early a vintage, a creator can not
2816 * request settings for these properties, even if the requested
2817 * setting is the default value. We don't actually want to create dsl
2818 * properties for these, so remove them from the source nvlist after
2822 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2823 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2824 nvlist_t *zplprops, boolean_t *is_ci)
2826 uint64_t sense = ZFS_PROP_UNDEFINED;
2827 uint64_t norm = ZFS_PROP_UNDEFINED;
2828 uint64_t u8 = ZFS_PROP_UNDEFINED;
2831 ASSERT(zplprops != NULL);
2834 * Pull out creator prop choices, if any.
2837 (void) nvlist_lookup_uint64(createprops,
2838 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2839 (void) nvlist_lookup_uint64(createprops,
2840 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2841 (void) nvlist_remove_all(createprops,
2842 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2843 (void) nvlist_lookup_uint64(createprops,
2844 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2845 (void) nvlist_remove_all(createprops,
2846 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2847 (void) nvlist_lookup_uint64(createprops,
2848 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2849 (void) nvlist_remove_all(createprops,
2850 zfs_prop_to_name(ZFS_PROP_CASE));
2854 * If the zpl version requested is whacky or the file system
2855 * or pool is version is too "young" to support normalization
2856 * and the creator tried to set a value for one of the props,
2859 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2860 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2861 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2862 (zplver < ZPL_VERSION_NORMALIZATION &&
2863 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2864 sense != ZFS_PROP_UNDEFINED)))
2868 * Put the version in the zplprops
2870 VERIFY(nvlist_add_uint64(zplprops,
2871 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2873 if (norm == ZFS_PROP_UNDEFINED &&
2874 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
2876 VERIFY(nvlist_add_uint64(zplprops,
2877 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2880 * If we're normalizing, names must always be valid UTF-8 strings.
2884 if (u8 == ZFS_PROP_UNDEFINED &&
2885 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
2887 VERIFY(nvlist_add_uint64(zplprops,
2888 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2890 if (sense == ZFS_PROP_UNDEFINED &&
2891 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
2893 VERIFY(nvlist_add_uint64(zplprops,
2894 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2897 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2903 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2904 nvlist_t *zplprops, boolean_t *is_ci)
2906 boolean_t fuids_ok, sa_ok;
2907 uint64_t zplver = ZPL_VERSION;
2908 objset_t *os = NULL;
2909 char parentname[MAXNAMELEN];
2915 (void) strlcpy(parentname, dataset, sizeof (parentname));
2916 cp = strrchr(parentname, '/');
2920 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2923 spa_vers = spa_version(spa);
2924 spa_close(spa, FTAG);
2926 zplver = zfs_zpl_version_map(spa_vers);
2927 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2928 sa_ok = (zplver >= ZPL_VERSION_SA);
2931 * Open parent object set so we can inherit zplprop values.
2933 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2936 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2938 dmu_objset_rele(os, FTAG);
2943 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2944 nvlist_t *zplprops, boolean_t *is_ci)
2948 uint64_t zplver = ZPL_VERSION;
2951 zplver = zfs_zpl_version_map(spa_vers);
2952 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2953 sa_ok = (zplver >= ZPL_VERSION_SA);
2955 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2956 createprops, zplprops, is_ci);
2962 * zc_objset_type type of objset to create (fs vs zvol)
2963 * zc_name name of new objset
2964 * zc_value name of snapshot to clone from (may be empty)
2965 * zc_nvlist_src{_size} nvlist of properties to apply
2970 zfs_ioc_create(zfs_cmd_t *zc)
2975 nvlist_t *nvprops = NULL;
2976 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2977 dmu_objset_type_t type = zc->zc_objset_type;
2982 cbfunc = zfs_create_cb;
2986 cbfunc = zvol_create_cb;
2993 if (strchr(zc->zc_name, '@') ||
2994 strchr(zc->zc_name, '%'))
2997 if (zc->zc_nvlist_src != 0 &&
2998 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2999 zc->zc_iflags, &nvprops)) != 0)
3002 zct.zct_zplprops = NULL;
3003 zct.zct_props = nvprops;
3005 if (zc->zc_value[0] != '\0') {
3007 * We're creating a clone of an existing snapshot.
3009 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3010 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
3011 nvlist_free(nvprops);
3015 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
3017 nvlist_free(nvprops);
3021 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
3022 dmu_objset_rele(clone, FTAG);
3024 nvlist_free(nvprops);
3028 boolean_t is_insensitive = B_FALSE;
3030 if (cbfunc == NULL) {
3031 nvlist_free(nvprops);
3035 if (type == DMU_OST_ZVOL) {
3036 uint64_t volsize, volblocksize;
3038 if (nvprops == NULL ||
3039 nvlist_lookup_uint64(nvprops,
3040 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
3042 nvlist_free(nvprops);
3046 if ((error = nvlist_lookup_uint64(nvprops,
3047 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3048 &volblocksize)) != 0 && error != ENOENT) {
3049 nvlist_free(nvprops);
3054 volblocksize = zfs_prop_default_numeric(
3055 ZFS_PROP_VOLBLOCKSIZE);
3057 if ((error = zvol_check_volblocksize(
3058 volblocksize)) != 0 ||
3059 (error = zvol_check_volsize(volsize,
3060 volblocksize)) != 0) {
3061 nvlist_free(nvprops);
3064 } else if (type == DMU_OST_ZFS) {
3068 * We have to have normalization and
3069 * case-folding flags correct when we do the
3070 * file system creation, so go figure them out
3073 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3074 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3075 error = zfs_fill_zplprops(zc->zc_name, nvprops,
3076 zct.zct_zplprops, &is_insensitive);
3078 nvlist_free(nvprops);
3079 nvlist_free(zct.zct_zplprops);
3083 error = dmu_objset_create(zc->zc_name, type,
3084 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3085 nvlist_free(zct.zct_zplprops);
3089 * It would be nice to do this atomically.
3092 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3095 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3097 nvlist_free(nvprops);
3103 * zc_name name of filesystem
3104 * zc_value short name of snapshot
3105 * zc_cookie recursive flag
3106 * zc_nvlist_src[_size] property list
3109 * zc_value short snapname (i.e. part after the '@')
3112 zfs_ioc_snapshot(zfs_cmd_t *zc)
3114 nvlist_t *nvprops = NULL;
3116 boolean_t recursive = zc->zc_cookie;
3118 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3121 if (zc->zc_nvlist_src != 0 &&
3122 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3123 zc->zc_iflags, &nvprops)) != 0)
3126 error = zfs_check_userprops(zc->zc_name, nvprops);
3130 if (!nvlist_empty(nvprops) &&
3131 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3136 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3137 nvprops, recursive, B_FALSE, -1);
3140 nvlist_free(nvprops);
3146 * name dataset name, or when 'arg == NULL' the full snapshot name
3147 * arg short snapshot name (i.e. part after the '@')
3150 zfs_unmount_snap(const char *name, void *arg)
3152 zfs_sb_t *zsb = NULL;
3160 dsname = strdup(name);
3161 snapname = strdup(arg);
3163 ptr = strchr(name, '@');
3165 dsname = strdup(name);
3166 dsname[ptr - name] = '\0';
3167 snapname = strdup(ptr + 1);
3173 fullname = kmem_asprintf("%s@%s", dsname, snapname);
3175 error = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
3177 error = zfsctl_unmount_snapshot(zsb, fullname, MNT_FORCE);
3178 zfs_sb_rele(zsb, FTAG);
3180 /* Allow ENOENT for consistency with upstream */
3181 if (error == ENOENT)
3194 * zc_name name of filesystem, snaps must be under it
3195 * zc_nvlist_src[_size] full names of snapshots to destroy
3196 * zc_defer_destroy mark for deferred destroy
3199 * zc_name on failure, name of failed snapshot
3202 zfs_ioc_destroy_snaps_nvl(zfs_cmd_t *zc)
3208 if ((err = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3209 zc->zc_iflags, &nvl)) != 0)
3212 len = strlen(zc->zc_name);
3213 for (pair = nvlist_next_nvpair(nvl, NULL); pair != NULL;
3214 pair = nvlist_next_nvpair(nvl, pair)) {
3215 const char *name = nvpair_name(pair);
3217 * The snap name must be underneath the zc_name. This ensures
3218 * that our permission checks were legitimate.
3220 if (strncmp(zc->zc_name, name, len) != 0 ||
3221 (name[len] != '@' && name[len] != '/')) {
3226 (void) zfs_unmount_snap(name, NULL);
3227 (void) zvol_remove_minor(name);
3230 err = dmu_snapshots_destroy_nvl(nvl, zc->zc_defer_destroy,
3238 * zc_name name of dataset to destroy
3239 * zc_objset_type type of objset
3240 * zc_defer_destroy mark for deferred destroy
3245 zfs_ioc_destroy(zfs_cmd_t *zc)
3248 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3249 err = zfs_unmount_snap(zc->zc_name, NULL);
3254 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3255 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3256 (void) zvol_remove_minor(zc->zc_name);
3262 * zc_name name of dataset to rollback (to most recent snapshot)
3267 zfs_ioc_rollback(zfs_cmd_t *zc)
3269 dsl_dataset_t *ds, *clone;
3274 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3278 /* must not be a snapshot */
3279 if (dsl_dataset_is_snapshot(ds)) {
3280 dsl_dataset_rele(ds, FTAG);
3284 /* must have a most recent snapshot */
3285 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3286 dsl_dataset_rele(ds, FTAG);
3291 * Create clone of most recent snapshot.
3293 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3294 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3298 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3305 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
3306 error = zfs_suspend_fs(zsb);
3310 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3311 error = dsl_dataset_clone_swap(clone, ds,
3313 dsl_dataset_disown(ds, FTAG);
3318 resume_err = zfs_resume_fs(zsb, zc->zc_name);
3319 error = error ? error : resume_err;
3321 deactivate_super(zsb->z_sb);
3323 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3324 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3325 dsl_dataset_disown(ds, FTAG);
3333 * Destroy clone (which also closes it).
3335 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3338 strfree(clone_name);
3340 dsl_dataset_rele(ds, FTAG);
3346 * zc_name old name of dataset
3347 * zc_value new name of dataset
3348 * zc_cookie recursive flag (only valid for snapshots)
3353 zfs_ioc_rename(zfs_cmd_t *zc)
3355 boolean_t recursive = zc->zc_cookie & 1;
3358 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3359 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3360 strchr(zc->zc_value, '%'))
3364 * Unmount snapshot unless we're doing a recursive rename,
3365 * in which case the dataset code figures out which snapshots
3368 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3369 zc->zc_objset_type == DMU_OST_ZFS) {
3370 err = zfs_unmount_snap(zc->zc_name, NULL);
3375 err = dmu_objset_rename(zc->zc_name, zc->zc_value, recursive);
3376 if ((err == 0) && (zc->zc_objset_type == DMU_OST_ZVOL)) {
3377 (void) zvol_remove_minor(zc->zc_name);
3378 (void) zvol_create_minor(zc->zc_value);
3385 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3387 const char *propname = nvpair_name(pair);
3388 boolean_t issnap = (strchr(dsname, '@') != NULL);
3389 zfs_prop_t prop = zfs_name_to_prop(propname);
3393 if (prop == ZPROP_INVAL) {
3394 if (zfs_prop_user(propname)) {
3395 if ((err = zfs_secpolicy_write_perms(dsname,
3396 ZFS_DELEG_PERM_USERPROP, cr)))
3401 if (!issnap && zfs_prop_userquota(propname)) {
3402 const char *perm = NULL;
3403 const char *uq_prefix =
3404 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3405 const char *gq_prefix =
3406 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3408 if (strncmp(propname, uq_prefix,
3409 strlen(uq_prefix)) == 0) {
3410 perm = ZFS_DELEG_PERM_USERQUOTA;
3411 } else if (strncmp(propname, gq_prefix,
3412 strlen(gq_prefix)) == 0) {
3413 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3415 /* USERUSED and GROUPUSED are read-only */
3419 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3430 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3432 * dsl_prop_get_all_impl() returns properties in this
3436 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3437 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3442 * Check that this value is valid for this pool version
3445 case ZFS_PROP_COMPRESSION:
3447 * If the user specified gzip compression, make sure
3448 * the SPA supports it. We ignore any errors here since
3449 * we'll catch them later.
3451 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3452 nvpair_value_uint64(pair, &intval) == 0) {
3453 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3454 intval <= ZIO_COMPRESS_GZIP_9 &&
3455 zfs_earlier_version(dsname,
3456 SPA_VERSION_GZIP_COMPRESSION)) {
3460 if (intval == ZIO_COMPRESS_ZLE &&
3461 zfs_earlier_version(dsname,
3462 SPA_VERSION_ZLE_COMPRESSION))
3465 if (intval == ZIO_COMPRESS_LZ4) {
3466 zfeature_info_t *feature =
3468 SPA_FEATURE_LZ4_COMPRESS];
3471 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3474 if (!spa_feature_is_enabled(spa, feature)) {
3475 spa_close(spa, FTAG);
3478 spa_close(spa, FTAG);
3482 * If this is a bootable dataset then
3483 * verify that the compression algorithm
3484 * is supported for booting. We must return
3485 * something other than ENOTSUP since it
3486 * implies a downrev pool version.
3488 if (zfs_is_bootfs(dsname) &&
3489 !BOOTFS_COMPRESS_VALID(intval)) {
3495 case ZFS_PROP_COPIES:
3496 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3500 case ZFS_PROP_DEDUP:
3501 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3505 case ZFS_PROP_SHARESMB:
3506 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3510 case ZFS_PROP_ACLINHERIT:
3511 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3512 nvpair_value_uint64(pair, &intval) == 0) {
3513 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3514 zfs_earlier_version(dsname,
3515 SPA_VERSION_PASSTHROUGH_X))
3523 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3527 * Activates a feature on a pool in response to a property setting. This
3528 * creates a new sync task which modifies the pool to reflect the feature
3532 zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature)
3536 /* EBUSY here indicates that the feature is already active */
3537 err = dsl_sync_task_do(dp, zfs_prop_activate_feature_check,
3538 zfs_prop_activate_feature_sync, dp->dp_spa, feature, 2);
3540 if (err != 0 && err != EBUSY)
3547 * Checks for a race condition to make sure we don't increment a feature flag
3552 zfs_prop_activate_feature_check(void *arg1, void *arg2, dmu_tx_t *tx)
3555 zfeature_info_t *feature = arg2;
3557 if (!spa_feature_is_active(spa, feature))
3564 * The callback invoked on feature activation in the sync task caused by
3565 * zfs_prop_activate_feature.
3568 zfs_prop_activate_feature_sync(void *arg1, void *arg2, dmu_tx_t *tx)
3571 zfeature_info_t *feature = arg2;
3573 spa_feature_incr(spa, feature, tx);
3577 * Removes properties from the given props list that fail permission checks
3578 * needed to clear them and to restore them in case of a receive error. For each
3579 * property, make sure we have both set and inherit permissions.
3581 * Returns the first error encountered if any permission checks fail. If the
3582 * caller provides a non-NULL errlist, it also gives the complete list of names
3583 * of all the properties that failed a permission check along with the
3584 * corresponding error numbers. The caller is responsible for freeing the
3587 * If every property checks out successfully, zero is returned and the list
3588 * pointed at by errlist is NULL.
3591 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3594 nvpair_t *pair, *next_pair;
3601 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3603 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3604 (void) strcpy(zc->zc_name, dataset);
3605 pair = nvlist_next_nvpair(props, NULL);
3606 while (pair != NULL) {
3607 next_pair = nvlist_next_nvpair(props, pair);
3609 (void) strcpy(zc->zc_value, nvpair_name(pair));
3610 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3611 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3612 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3613 VERIFY(nvlist_add_int32(errors,
3614 zc->zc_value, err) == 0);
3618 kmem_free(zc, sizeof (zfs_cmd_t));
3620 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3621 nvlist_free(errors);
3624 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3627 if (errlist == NULL)
3628 nvlist_free(errors);
3636 propval_equals(nvpair_t *p1, nvpair_t *p2)
3638 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3639 /* dsl_prop_get_all_impl() format */
3641 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3642 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3646 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3648 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3649 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3653 if (nvpair_type(p1) != nvpair_type(p2))
3656 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3657 char *valstr1, *valstr2;
3659 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3660 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3661 return (strcmp(valstr1, valstr2) == 0);
3663 uint64_t intval1, intval2;
3665 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3666 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3667 return (intval1 == intval2);
3672 * Remove properties from props if they are not going to change (as determined
3673 * by comparison with origprops). Remove them from origprops as well, since we
3674 * do not need to clear or restore properties that won't change.
3677 props_reduce(nvlist_t *props, nvlist_t *origprops)
3679 nvpair_t *pair, *next_pair;
3681 if (origprops == NULL)
3682 return; /* all props need to be received */
3684 pair = nvlist_next_nvpair(props, NULL);
3685 while (pair != NULL) {
3686 const char *propname = nvpair_name(pair);
3689 next_pair = nvlist_next_nvpair(props, pair);
3691 if ((nvlist_lookup_nvpair(origprops, propname,
3692 &match) != 0) || !propval_equals(pair, match))
3693 goto next; /* need to set received value */
3695 /* don't clear the existing received value */
3696 (void) nvlist_remove_nvpair(origprops, match);
3697 /* don't bother receiving the property */
3698 (void) nvlist_remove_nvpair(props, pair);
3705 static boolean_t zfs_ioc_recv_inject_err;
3710 * zc_name name of containing filesystem
3711 * zc_nvlist_src{_size} nvlist of properties to apply
3712 * zc_value name of snapshot to create
3713 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3714 * zc_cookie file descriptor to recv from
3715 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3716 * zc_guid force flag
3717 * zc_cleanup_fd cleanup-on-exit file descriptor
3718 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3721 * zc_cookie number of bytes read
3722 * zc_nvlist_dst{_size} error for each unapplied received property
3723 * zc_obj zprop_errflags_t
3724 * zc_action_handle handle for this guid/ds mapping
3727 zfs_ioc_recv(zfs_cmd_t *zc)
3731 dmu_recv_cookie_t drc;
3732 boolean_t force = (boolean_t)zc->zc_guid;
3735 int props_error = 0;
3738 nvlist_t *props = NULL; /* sent properties */
3739 nvlist_t *origprops = NULL; /* existing properties */
3740 objset_t *origin = NULL;
3742 char tofs[ZFS_MAXNAMELEN];
3743 boolean_t first_recvd_props = B_FALSE;
3745 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3746 strchr(zc->zc_value, '@') == NULL ||
3747 strchr(zc->zc_value, '%'))
3750 (void) strcpy(tofs, zc->zc_value);
3751 tosnap = strchr(tofs, '@');
3754 if (zc->zc_nvlist_src != 0 &&
3755 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3756 zc->zc_iflags, &props)) != 0)
3766 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3768 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3769 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3770 !dsl_prop_get_hasrecvd(os)) {
3771 first_recvd_props = B_TRUE;
3775 * If new received properties are supplied, they are to
3776 * completely replace the existing received properties, so stash
3777 * away the existing ones.
3779 if (dsl_prop_get_received(os, &origprops) == 0) {
3780 nvlist_t *errlist = NULL;
3782 * Don't bother writing a property if its value won't
3783 * change (and avoid the unnecessary security checks).
3785 * The first receive after SPA_VERSION_RECVD_PROPS is a
3786 * special case where we blow away all local properties
3789 if (!first_recvd_props)
3790 props_reduce(props, origprops);
3791 if (zfs_check_clearable(tofs, origprops,
3793 (void) nvlist_merge(errors, errlist, 0);
3794 nvlist_free(errlist);
3797 dmu_objset_rele(os, FTAG);
3800 if (zc->zc_string[0]) {
3801 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3806 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3807 &zc->zc_begin_record, force, origin, &drc);
3809 dmu_objset_rele(origin, FTAG);
3814 * Set properties before we receive the stream so that they are applied
3815 * to the new data. Note that we must call dmu_recv_stream() if
3816 * dmu_recv_begin() succeeds.
3821 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3822 if (drc.drc_newfs) {
3823 if (spa_version(os->os_spa) >=
3824 SPA_VERSION_RECVD_PROPS)
3825 first_recvd_props = B_TRUE;
3826 } else if (origprops != NULL) {
3827 if (clear_received_props(os, tofs, origprops,
3828 first_recvd_props ? NULL : props) != 0)
3829 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3831 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3833 dsl_prop_set_hasrecvd(os);
3834 } else if (!drc.drc_newfs) {
3835 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3838 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3840 (void) nvlist_merge(errors, errlist, 0);
3841 nvlist_free(errlist);
3844 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3846 * Caller made zc->zc_nvlist_dst less than the minimum expected
3847 * size or supplied an invalid address.
3849 props_error = EINVAL;
3853 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3854 &zc->zc_action_handle);
3857 zfs_sb_t *zsb = NULL;
3859 if (get_zfs_sb(tofs, &zsb) == 0) {
3863 error = zfs_suspend_fs(zsb);
3865 * If the suspend fails, then the recv_end will
3866 * likely also fail, and clean up after itself.
3868 end_err = dmu_recv_end(&drc);
3870 error = zfs_resume_fs(zsb, tofs);
3871 error = error ? error : end_err;
3872 deactivate_super(zsb->z_sb);
3874 error = dmu_recv_end(&drc);
3878 zc->zc_cookie = off - fp->f_offset;
3879 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3883 if (zfs_ioc_recv_inject_err) {
3884 zfs_ioc_recv_inject_err = B_FALSE;
3889 * On error, restore the original props.
3891 if (error && props) {
3892 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3893 if (clear_received_props(os, tofs, props, NULL) != 0) {
3895 * We failed to clear the received properties.
3896 * Since we may have left a $recvd value on the
3897 * system, we can't clear the $hasrecvd flag.
3899 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3900 } else if (first_recvd_props) {
3901 dsl_prop_unset_hasrecvd(os);
3903 dmu_objset_rele(os, FTAG);
3904 } else if (!drc.drc_newfs) {
3905 /* We failed to clear the received properties. */
3906 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3909 if (origprops == NULL && !drc.drc_newfs) {
3910 /* We failed to stash the original properties. */
3911 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3915 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3916 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3917 * explictly if we're restoring local properties cleared in the
3918 * first new-style receive.
3920 if (origprops != NULL &&
3921 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3922 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3923 origprops, NULL) != 0) {
3925 * We stashed the original properties but failed to
3928 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3933 nvlist_free(origprops);
3934 nvlist_free(errors);
3938 error = props_error;
3945 * zc_name name of snapshot to send
3946 * zc_cookie file descriptor to send stream to
3947 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3948 * zc_sendobj objsetid of snapshot to send
3949 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3950 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
3951 * output size in zc_objset_type.
3956 zfs_ioc_send(zfs_cmd_t *zc)
3958 objset_t *fromsnap = NULL;
3963 dsl_dataset_t *dsfrom = NULL;
3966 boolean_t estimate = (zc->zc_guid != 0);
3968 error = spa_open(zc->zc_name, &spa, FTAG);
3972 dp = spa_get_dsl(spa);
3973 rw_enter(&dp->dp_config_rwlock, RW_READER);
3974 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3975 rw_exit(&dp->dp_config_rwlock);
3977 spa_close(spa, FTAG);
3981 error = dmu_objset_from_ds(ds, &tosnap);
3983 dsl_dataset_rele(ds, FTAG);
3984 spa_close(spa, FTAG);
3988 if (zc->zc_fromobj != 0) {
3989 rw_enter(&dp->dp_config_rwlock, RW_READER);
3990 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3991 rw_exit(&dp->dp_config_rwlock);
3992 spa_close(spa, FTAG);
3994 dsl_dataset_rele(ds, FTAG);
3997 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3999 dsl_dataset_rele(dsfrom, FTAG);
4000 dsl_dataset_rele(ds, FTAG);
4004 spa_close(spa, FTAG);
4008 error = dmu_send_estimate(tosnap, fromsnap, zc->zc_obj,
4009 &zc->zc_objset_type);
4011 file_t *fp = getf(zc->zc_cookie);
4013 dsl_dataset_rele(ds, FTAG);
4015 dsl_dataset_rele(dsfrom, FTAG);
4020 error = dmu_send(tosnap, fromsnap, zc->zc_obj,
4021 zc->zc_cookie, fp->f_vnode, &off);
4023 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4025 releasef(zc->zc_cookie);
4028 dsl_dataset_rele(dsfrom, FTAG);
4029 dsl_dataset_rele(ds, FTAG);
4035 * zc_name name of snapshot on which to report progress
4036 * zc_cookie file descriptor of send stream
4039 * zc_cookie number of bytes written in send stream thus far
4042 zfs_ioc_send_progress(zfs_cmd_t *zc)
4045 dmu_sendarg_t *dsp = NULL;
4048 if ((error = dsl_dataset_hold(zc->zc_name, FTAG, &ds)) != 0)
4051 mutex_enter(&ds->ds_sendstream_lock);
4054 * Iterate over all the send streams currently active on this dataset.
4055 * If there's one which matches the specified file descriptor _and_ the
4056 * stream was started by the current process, return the progress of
4060 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4061 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4062 if (dsp->dsa_outfd == zc->zc_cookie &&
4063 dsp->dsa_proc->group_leader == curproc->group_leader)
4068 zc->zc_cookie = *(dsp->dsa_off);
4072 mutex_exit(&ds->ds_sendstream_lock);
4073 dsl_dataset_rele(ds, FTAG);
4078 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4082 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4083 &zc->zc_inject_record);
4086 zc->zc_guid = (uint64_t)id;
4092 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4094 return (zio_clear_fault((int)zc->zc_guid));
4098 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4100 int id = (int)zc->zc_guid;
4103 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4104 &zc->zc_inject_record);
4112 zfs_ioc_error_log(zfs_cmd_t *zc)
4116 size_t count = (size_t)zc->zc_nvlist_dst_size;
4118 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4121 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4124 zc->zc_nvlist_dst_size = count;
4126 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4128 spa_close(spa, FTAG);
4134 zfs_ioc_clear(zfs_cmd_t *zc)
4141 * On zpool clear we also fix up missing slogs
4143 mutex_enter(&spa_namespace_lock);
4144 spa = spa_lookup(zc->zc_name);
4146 mutex_exit(&spa_namespace_lock);
4149 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4150 /* we need to let spa_open/spa_load clear the chains */
4151 spa_set_log_state(spa, SPA_LOG_CLEAR);
4153 spa->spa_last_open_failed = 0;
4154 mutex_exit(&spa_namespace_lock);
4156 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4157 error = spa_open(zc->zc_name, &spa, FTAG);
4160 nvlist_t *config = NULL;
4162 if (zc->zc_nvlist_src == 0)
4165 if ((error = get_nvlist(zc->zc_nvlist_src,
4166 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4167 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4169 if (config != NULL) {
4172 if ((err = put_nvlist(zc, config)) != 0)
4174 nvlist_free(config);
4176 nvlist_free(policy);
4183 spa_vdev_state_enter(spa, SCL_NONE);
4185 if (zc->zc_guid == 0) {
4188 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4190 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4191 spa_close(spa, FTAG);
4196 vdev_clear(spa, vd);
4198 (void) spa_vdev_state_exit(spa, NULL, 0);
4201 * Resume any suspended I/Os.
4203 if (zio_resume(spa) != 0)
4206 spa_close(spa, FTAG);
4212 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4217 error = spa_open(zc->zc_name, &spa, FTAG);
4221 spa_vdev_state_enter(spa, SCL_NONE);
4224 * If a resilver is already in progress then set the
4225 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4226 * the scan as a side effect of the reopen. Otherwise, let
4227 * vdev_open() decided if a resilver is required.
4229 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4230 vdev_reopen(spa->spa_root_vdev);
4231 spa->spa_scrub_reopen = B_FALSE;
4233 (void) spa_vdev_state_exit(spa, NULL, 0);
4234 spa_close(spa, FTAG);
4239 * zc_name name of filesystem
4240 * zc_value name of origin snapshot
4243 * zc_string name of conflicting snapshot, if there is one
4246 zfs_ioc_promote(zfs_cmd_t *zc)
4251 * We don't need to unmount *all* the origin fs's snapshots, but
4254 cp = strchr(zc->zc_value, '@');
4257 (void) dmu_objset_find(zc->zc_value,
4258 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4259 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4263 * Retrieve a single {user|group}{used|quota}@... property.
4266 * zc_name name of filesystem
4267 * zc_objset_type zfs_userquota_prop_t
4268 * zc_value domain name (eg. "S-1-234-567-89")
4269 * zc_guid RID/UID/GID
4272 * zc_cookie property value
4275 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4280 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4283 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4287 error = zfs_userspace_one(zsb,
4288 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4289 zfs_sb_rele(zsb, FTAG);
4296 * zc_name name of filesystem
4297 * zc_cookie zap cursor
4298 * zc_objset_type zfs_userquota_prop_t
4299 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4302 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4303 * zc_cookie zap cursor
4306 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4309 int bufsize = zc->zc_nvlist_dst_size;
4316 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4320 buf = vmem_alloc(bufsize, KM_SLEEP);
4322 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4323 buf, &zc->zc_nvlist_dst_size);
4326 error = xcopyout(buf,
4327 (void *)(uintptr_t)zc->zc_nvlist_dst,
4328 zc->zc_nvlist_dst_size);
4330 vmem_free(buf, bufsize);
4331 zfs_sb_rele(zsb, FTAG);
4338 * zc_name name of filesystem
4344 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4350 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4351 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4353 * If userused is not enabled, it may be because the
4354 * objset needs to be closed & reopened (to grow the
4355 * objset_phys_t). Suspend/resume the fs will do that.
4357 error = zfs_suspend_fs(zsb);
4359 error = zfs_resume_fs(zsb, zc->zc_name);
4362 error = dmu_objset_userspace_upgrade(zsb->z_os);
4363 deactivate_super(zsb->z_sb);
4365 /* XXX kind of reading contents without owning */
4366 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4370 error = dmu_objset_userspace_upgrade(os);
4371 dmu_objset_rele(os, FTAG);
4378 zfs_ioc_share(zfs_cmd_t *zc)
4383 ace_t full_access[] = {
4384 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4389 * zc_name name of containing filesystem
4390 * zc_obj object # beyond which we want next in-use object #
4393 * zc_obj next in-use object #
4396 zfs_ioc_next_obj(zfs_cmd_t *zc)
4398 objset_t *os = NULL;
4401 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4405 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4406 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4408 dmu_objset_rele(os, FTAG);
4414 * zc_name name of filesystem
4415 * zc_value prefix name for snapshot
4416 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4421 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4426 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4427 (u_longlong_t)ddi_get_lbolt64());
4429 if (strlen(snap_name) >= MAXNAMELEN) {
4434 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4435 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4441 (void) strcpy(zc->zc_value, snap_name);
4448 * zc_name name of "to" snapshot
4449 * zc_value name of "from" snapshot
4450 * zc_cookie file descriptor to write diff data on
4453 * dmu_diff_record_t's to the file descriptor
4456 zfs_ioc_diff(zfs_cmd_t *zc)
4464 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4468 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4470 dmu_objset_rele(tosnap, FTAG);
4474 fp = getf(zc->zc_cookie);
4476 dmu_objset_rele(fromsnap, FTAG);
4477 dmu_objset_rele(tosnap, FTAG);
4483 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4485 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4487 releasef(zc->zc_cookie);
4489 dmu_objset_rele(fromsnap, FTAG);
4490 dmu_objset_rele(tosnap, FTAG);
4495 * Remove all ACL files in shares dir
4497 #ifdef HAVE_SMB_SHARE
4499 zfs_smb_acl_purge(znode_t *dzp)
4502 zap_attribute_t zap;
4503 zfs_sb_t *zsb = ZTOZSB(dzp);
4506 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4507 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4508 zap_cursor_advance(&zc)) {
4509 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4513 zap_cursor_fini(&zc);
4516 #endif /* HAVE_SMB_SHARE */
4519 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4521 #ifdef HAVE_SMB_SHARE
4524 vnode_t *resourcevp = NULL;
4533 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4534 NO_FOLLOW, NULL, &vp)) != 0)
4537 /* Now make sure mntpnt and dataset are ZFS */
4539 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4540 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4541 zc->zc_name) != 0)) {
4551 * Create share dir if its missing.
4553 mutex_enter(&zsb->z_lock);
4554 if (zsb->z_shares_dir == 0) {
4557 tx = dmu_tx_create(zsb->z_os);
4558 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4560 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4561 error = dmu_tx_assign(tx, TXG_WAIT);
4565 error = zfs_create_share_dir(zsb, tx);
4569 mutex_exit(&zsb->z_lock);
4575 mutex_exit(&zsb->z_lock);
4577 ASSERT(zsb->z_shares_dir);
4578 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4584 switch (zc->zc_cookie) {
4585 case ZFS_SMB_ACL_ADD:
4586 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4587 vattr.va_mode = S_IFREG|0777;
4591 vsec.vsa_mask = VSA_ACE;
4592 vsec.vsa_aclentp = &full_access;
4593 vsec.vsa_aclentsz = sizeof (full_access);
4594 vsec.vsa_aclcnt = 1;
4596 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4597 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4599 VN_RELE(resourcevp);
4602 case ZFS_SMB_ACL_REMOVE:
4603 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4607 case ZFS_SMB_ACL_RENAME:
4608 if ((error = get_nvlist(zc->zc_nvlist_src,
4609 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4614 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4615 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4618 VN_RELE(ZTOV(sharedir));
4620 nvlist_free(nvlist);
4623 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4625 nvlist_free(nvlist);
4628 case ZFS_SMB_ACL_PURGE:
4629 error = zfs_smb_acl_purge(sharedir);
4638 VN_RELE(ZTOV(sharedir));
4645 #endif /* HAVE_SMB_SHARE */
4650 * zc_name name of filesystem
4651 * zc_value short name of snap
4652 * zc_string user-supplied tag for this hold
4653 * zc_cookie recursive flag
4654 * zc_temphold set if hold is temporary
4655 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4656 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4657 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4662 zfs_ioc_hold(zfs_cmd_t *zc)
4664 boolean_t recursive = zc->zc_cookie;
4671 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4674 if (zc->zc_sendobj == 0) {
4675 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4676 zc->zc_string, recursive, zc->zc_temphold,
4677 zc->zc_cleanup_fd));
4683 error = spa_open(zc->zc_name, &spa, FTAG);
4687 dp = spa_get_dsl(spa);
4688 rw_enter(&dp->dp_config_rwlock, RW_READER);
4689 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4690 rw_exit(&dp->dp_config_rwlock);
4691 spa_close(spa, FTAG);
4696 * Until we have a hold on this snapshot, it's possible that
4697 * zc_sendobj could've been destroyed and reused as part
4698 * of a later txg. Make sure we're looking at the right object.
4700 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4701 dsl_dataset_rele(ds, FTAG);
4705 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4706 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4708 dsl_dataset_rele(ds, FTAG);
4713 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4717 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4720 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4722 dsl_dataset_rele(ds, FTAG);
4729 * zc_name name of dataset from which we're releasing a user hold
4730 * zc_value short name of snap
4731 * zc_string user-supplied tag for this hold
4732 * zc_cookie recursive flag
4737 zfs_ioc_release(zfs_cmd_t *zc)
4739 boolean_t recursive = zc->zc_cookie;
4741 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4744 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4745 zc->zc_string, recursive));
4750 * zc_name name of filesystem
4753 * zc_nvlist_src{_size} nvlist of snapshot holds
4756 zfs_ioc_get_holds(zfs_cmd_t *zc)
4761 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4762 error = put_nvlist(zc, nvp);
4771 * zc_guid flags (ZEVENT_NONBLOCK)
4774 * zc_nvlist_dst next nvlist event
4775 * zc_cookie dropped events since last get
4776 * zc_cleanup_fd cleanup-on-exit file descriptor
4779 zfs_ioc_events_next(zfs_cmd_t *zc)
4782 nvlist_t *event = NULL;
4784 uint64_t dropped = 0;
4787 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4792 error = zfs_zevent_next(ze, &event,
4793 &zc->zc_nvlist_dst_size, &dropped);
4794 if (event != NULL) {
4795 zc->zc_cookie = dropped;
4796 error = put_nvlist(zc, event);
4800 if (zc->zc_guid & ZEVENT_NONBLOCK)
4803 if ((error == 0) || (error != ENOENT))
4806 error = zfs_zevent_wait(ze);
4811 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4818 * zc_cookie cleared events count
4821 zfs_ioc_events_clear(zfs_cmd_t *zc)
4825 zfs_zevent_drain_all(&count);
4826 zc->zc_cookie = count;
4833 * zc_name name of new filesystem or snapshot
4834 * zc_value full name of old snapshot
4837 * zc_cookie space in bytes
4838 * zc_objset_type compressed space in bytes
4839 * zc_perm_action uncompressed space in bytes
4842 zfs_ioc_space_written(zfs_cmd_t *zc)
4845 dsl_dataset_t *new, *old;
4847 error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
4850 error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
4852 dsl_dataset_rele(new, FTAG);
4856 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
4857 &zc->zc_objset_type, &zc->zc_perm_action);
4858 dsl_dataset_rele(old, FTAG);
4859 dsl_dataset_rele(new, FTAG);
4865 * zc_name full name of last snapshot
4866 * zc_value full name of first snapshot
4869 * zc_cookie space in bytes
4870 * zc_objset_type compressed space in bytes
4871 * zc_perm_action uncompressed space in bytes
4874 zfs_ioc_space_snaps(zfs_cmd_t *zc)
4877 dsl_dataset_t *new, *old;
4879 error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
4882 error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
4884 dsl_dataset_rele(new, FTAG);
4888 error = dsl_dataset_space_wouldfree(old, new, &zc->zc_cookie,
4889 &zc->zc_objset_type, &zc->zc_perm_action);
4890 dsl_dataset_rele(old, FTAG);
4891 dsl_dataset_rele(new, FTAG);
4896 * pool create, destroy, and export don't log the history as part of
4897 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4898 * do the logging of those commands.
4900 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4901 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4903 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4905 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4907 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4909 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4911 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4913 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4915 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4916 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4917 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4918 POOL_CHECK_READONLY },
4919 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4920 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4921 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4923 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4924 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4925 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4926 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4927 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4928 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4929 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4930 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4931 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4932 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4933 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4934 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4935 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4936 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4937 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4938 POOL_CHECK_SUSPENDED },
4939 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4941 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4942 POOL_CHECK_SUSPENDED },
4943 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4944 POOL_CHECK_SUSPENDED },
4945 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4946 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4947 { zfs_ioc_create_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4949 { zfs_ioc_remove_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4951 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4952 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4953 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4954 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4955 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4956 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4957 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4958 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4959 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4960 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4961 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_FALSE,
4963 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4965 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4967 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4969 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4971 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4973 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4974 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4975 { zfs_ioc_destroy_snaps_nvl, zfs_secpolicy_destroy_recursive,
4976 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4977 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4978 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4979 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4981 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4982 POOL_CHECK_SUSPENDED },
4983 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4984 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4985 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4987 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4988 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4989 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4991 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4993 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4994 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4995 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4997 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4998 B_FALSE, POOL_CHECK_NONE },
4999 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
5000 B_FALSE, POOL_CHECK_NONE },
5001 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5002 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5003 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
5004 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5005 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
5006 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5007 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5008 POOL_CHECK_SUSPENDED },
5009 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5011 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5012 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5013 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5015 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
5017 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
5018 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5019 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
5020 POOL_CHECK_SUSPENDED },
5021 { zfs_ioc_events_next, zfs_secpolicy_config, NO_NAME, B_FALSE,
5023 { zfs_ioc_events_clear, zfs_secpolicy_config, NO_NAME, B_FALSE,
5025 { zfs_ioc_pool_reguid, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5026 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
5027 { zfs_ioc_space_written, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5028 POOL_CHECK_SUSPENDED },
5029 { zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5030 POOL_CHECK_SUSPENDED },
5031 { zfs_ioc_pool_reopen, zfs_secpolicy_config, POOL_NAME, B_TRUE,
5032 POOL_CHECK_SUSPENDED },
5033 { zfs_ioc_send_progress, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
5038 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5039 zfs_ioc_poolcheck_t check)
5044 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5046 if (check & POOL_CHECK_NONE)
5049 error = spa_open(name, &spa, FTAG);
5051 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5053 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5055 spa_close(spa, FTAG);
5061 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
5065 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5067 for (zs = list_head(&zfsdev_state_list); zs != NULL;
5068 zs = list_next(&zfsdev_state_list, zs)) {
5069 if (zs->zs_minor == minor) {
5071 case ZST_ONEXIT: return (zs->zs_onexit);
5072 case ZST_ZEVENT: return (zs->zs_zevent);
5073 case ZST_ALL: return (zs);
5082 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
5086 mutex_enter(&zfsdev_state_lock);
5087 ptr = zfsdev_get_state_impl(minor, which);
5088 mutex_exit(&zfsdev_state_lock);
5094 zfsdev_getminor(struct file *filp)
5096 ASSERT(filp != NULL);
5097 ASSERT(filp->private_data != NULL);
5099 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
5103 * Find a free minor number. The zfsdev_state_list is expected to
5104 * be short since it is only a list of currently open file handles.
5107 zfsdev_minor_alloc(void)
5109 static minor_t last_minor = 0;
5112 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5114 for (m = last_minor + 1; m != last_minor; m++) {
5115 if (m > ZFSDEV_MAX_MINOR)
5117 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
5127 zfsdev_state_init(struct file *filp)
5132 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5134 minor = zfsdev_minor_alloc();
5138 zs = kmem_zalloc( sizeof(zfsdev_state_t), KM_SLEEP);
5143 zs->zs_minor = minor;
5144 filp->private_data = zs;
5146 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
5147 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
5149 list_insert_tail(&zfsdev_state_list, zs);
5155 zfsdev_state_destroy(struct file *filp)
5159 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5160 ASSERT(filp->private_data != NULL);
5162 zs = filp->private_data;
5163 zfs_onexit_destroy(zs->zs_onexit);
5164 zfs_zevent_destroy(zs->zs_zevent);
5166 list_remove(&zfsdev_state_list, zs);
5167 kmem_free(zs, sizeof(zfsdev_state_t));
5173 zfsdev_open(struct inode *ino, struct file *filp)
5177 mutex_enter(&zfsdev_state_lock);
5178 error = zfsdev_state_init(filp);
5179 mutex_exit(&zfsdev_state_lock);
5185 zfsdev_release(struct inode *ino, struct file *filp)
5189 mutex_enter(&zfsdev_state_lock);
5190 error = zfsdev_state_destroy(filp);
5191 mutex_exit(&zfsdev_state_lock);
5197 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5201 int error, rc, flag = 0;
5203 vec = cmd - ZFS_IOC;
5204 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5207 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
5209 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5213 if ((error == 0) && !(flag & FKIOCTL))
5214 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, CRED());
5217 * Ensure that all pool/dataset names are valid before we pass down to
5221 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5222 zc->zc_iflags = flag & FKIOCTL;
5223 switch (zfs_ioc_vec[vec].zvec_namecheck) {
5225 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5227 error = pool_status_check(zc->zc_name,
5228 zfs_ioc_vec[vec].zvec_namecheck,
5229 zfs_ioc_vec[vec].zvec_pool_check);
5233 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5235 error = pool_status_check(zc->zc_name,
5236 zfs_ioc_vec[vec].zvec_namecheck,
5237 zfs_ioc_vec[vec].zvec_pool_check);
5246 error = zfs_ioc_vec[vec].zvec_func(zc);
5248 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5252 if (zfs_ioc_vec[vec].zvec_his_log)
5253 zfs_log_history(zc);
5256 kmem_free(zc, sizeof (zfs_cmd_t));
5260 #ifdef CONFIG_COMPAT
5262 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
5264 return zfsdev_ioctl(filp, cmd, arg);
5267 #define zfsdev_compat_ioctl NULL
5270 static const struct file_operations zfsdev_fops = {
5271 .open = zfsdev_open,
5272 .release = zfsdev_release,
5273 .unlocked_ioctl = zfsdev_ioctl,
5274 .compat_ioctl = zfsdev_compat_ioctl,
5275 .owner = THIS_MODULE,
5278 static struct miscdevice zfs_misc = {
5279 .minor = MISC_DYNAMIC_MINOR,
5281 .fops = &zfsdev_fops,
5289 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
5290 list_create(&zfsdev_state_list, sizeof (zfsdev_state_t),
5291 offsetof(zfsdev_state_t, zs_next));
5293 error = misc_register(&zfs_misc);
5295 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
5307 error = misc_deregister(&zfs_misc);
5309 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5311 mutex_destroy(&zfsdev_state_lock);
5312 list_destroy(&zfsdev_state_list);
5315 uint_t zfs_fsyncer_key;
5316 extern uint_t rrw_tsd_key;
5319 #define ZFS_DEBUG_STR " (DEBUG mode)"
5321 #define ZFS_DEBUG_STR ""
5329 spa_init(FREAD | FWRITE);
5332 if ((error = zvol_init()) != 0)
5335 if ((error = zfs_attach()) != 0)
5338 tsd_create(&zfs_fsyncer_key, NULL);
5339 tsd_create(&rrw_tsd_key, NULL);
5341 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
5342 "ZFS pool version %s, ZFS filesystem version %s\n",
5343 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
5344 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5353 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
5354 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
5355 ZFS_DEBUG_STR, error);
5368 tsd_destroy(&zfs_fsyncer_key);
5369 tsd_destroy(&rrw_tsd_key);
5371 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
5372 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
5378 spl_module_init(_init);
5379 spl_module_exit(_fini);
5381 MODULE_DESCRIPTION("ZFS");
5382 MODULE_AUTHOR(ZFS_META_AUTHOR);
5383 MODULE_LICENSE(ZFS_META_LICENSE);
5384 #endif /* HAVE_SPL */