4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 #include <sys/types.h>
26 #include <sys/param.h>
27 #include <sys/errno.h>
30 #include <sys/modctl.h>
35 #include <sys/cmn_err.h>
37 #include <sys/zfs_ioctl.h>
38 #include <sys/zfs_vfsops.h>
39 #include <sys/zfs_znode.h>
42 #include <sys/spa_impl.h>
44 #include <sys/priv_impl.h>
46 #include <sys/dsl_dir.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_prop.h>
49 #include <sys/dsl_deleg.h>
50 #include <sys/dmu_objset.h>
52 #include <sys/sunddi.h>
53 #include <sys/sunldi.h>
54 #include <sys/policy.h>
56 #include <sys/nvpair.h>
57 #include <sys/pathname.h>
58 #include <sys/mount.h>
60 #include <sys/fs/zfs.h>
61 #include <sys/zfs_dir.h>
62 #include <sys/zfs_onexit.h>
64 #include <sys/dsl_scan.h>
65 #include <sharefs/share.h>
66 #include <sys/dmu_objset.h>
67 #include <sys/fm/util.h>
69 #include <linux/miscdevice.h>
71 #include "zfs_namecheck.h"
73 #include "zfs_deleg.h"
74 #include "zfs_comutil.h"
76 kmutex_t zfsdev_state_lock;
77 list_t zfsdev_state_list;
79 extern void zfs_init(void);
80 extern void zfs_fini(void);
82 typedef int zfs_ioc_func_t(zfs_cmd_t *);
83 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
89 } zfs_ioc_namecheck_t;
92 POOL_CHECK_NONE = 1 << 0,
93 POOL_CHECK_SUSPENDED = 1 << 1,
94 POOL_CHECK_READONLY = 1 << 2
95 } zfs_ioc_poolcheck_t;
97 typedef struct zfs_ioc_vec {
98 zfs_ioc_func_t *zvec_func;
99 zfs_secpolicy_func_t *zvec_secpolicy;
100 zfs_ioc_namecheck_t zvec_namecheck;
101 boolean_t zvec_his_log;
102 zfs_ioc_poolcheck_t zvec_pool_check;
105 /* This array is indexed by zfs_userquota_prop_t */
106 static const char *userquota_perms[] = {
107 ZFS_DELEG_PERM_USERUSED,
108 ZFS_DELEG_PERM_USERQUOTA,
109 ZFS_DELEG_PERM_GROUPUSED,
110 ZFS_DELEG_PERM_GROUPQUOTA,
113 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
114 static int zfs_check_settable(const char *name, nvpair_t *property,
116 static int zfs_check_clearable(char *dataset, nvlist_t *props,
118 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
120 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **);
122 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
124 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
131 * Get rid of annoying "../common/" prefix to filename.
133 newfile = strrchr(file, '/');
134 if (newfile != NULL) {
135 newfile = newfile + 1; /* Get rid of leading / */
141 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
145 * To get this data, use the zfs-dprintf probe as so:
146 * dtrace -q -n 'zfs-dprintf \
147 * /stringof(arg0) == "dbuf.c"/ \
148 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
150 * arg1 = function name
154 DTRACE_PROBE4(zfs__dprintf,
155 char *, newfile, char *, func, int, line, char *, buf);
159 history_str_free(char *buf)
161 kmem_free(buf, HIS_MAX_RECORD_LEN);
165 history_str_get(zfs_cmd_t *zc)
169 if (zc->zc_history == 0)
172 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP | KM_NODEBUG);
173 if (copyinstr((void *)(uintptr_t)zc->zc_history,
174 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
175 history_str_free(buf);
179 buf[HIS_MAX_RECORD_LEN -1] = '\0';
185 * Check to see if the named dataset is currently defined as bootable
188 zfs_is_bootfs(const char *name)
192 if (dmu_objset_hold(name, FTAG, &os) == 0) {
194 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
195 dmu_objset_rele(os, FTAG);
202 * zfs_earlier_version
204 * Return non-zero if the spa version is less than requested version.
207 zfs_earlier_version(const char *name, int version)
211 if (spa_open(name, &spa, FTAG) == 0) {
212 if (spa_version(spa) < version) {
213 spa_close(spa, FTAG);
216 spa_close(spa, FTAG);
222 * zpl_earlier_version
224 * Return TRUE if the ZPL version is less than requested version.
227 zpl_earlier_version(const char *name, int version)
230 boolean_t rc = B_TRUE;
232 if (dmu_objset_hold(name, FTAG, &os) == 0) {
235 if (dmu_objset_type(os) != DMU_OST_ZFS) {
236 dmu_objset_rele(os, FTAG);
239 /* XXX reading from non-owned objset */
240 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
241 rc = zplversion < version;
242 dmu_objset_rele(os, FTAG);
248 zfs_log_history(zfs_cmd_t *zc)
253 if ((buf = history_str_get(zc)) == NULL)
256 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
257 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
258 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
259 spa_close(spa, FTAG);
261 history_str_free(buf);
265 * Policy for top-level read operations (list pools). Requires no privileges,
266 * and can be used in the local zone, as there is no associated dataset.
270 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
276 * Policy for dataset read operations (list children, get statistics). Requires
277 * no privileges, but must be visible in the local zone.
281 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
283 if (INGLOBALZONE(curproc) ||
284 zone_dataset_visible(zc->zc_name, NULL))
291 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
296 * The dataset must be visible by this zone -- check this first
297 * so they don't see EPERM on something they shouldn't know about.
299 if (!INGLOBALZONE(curproc) &&
300 !zone_dataset_visible(dataset, &writable))
303 if (INGLOBALZONE(curproc)) {
305 * If the fs is zoned, only root can access it from the
308 if (secpolicy_zfs(cr) && zoned)
312 * If we are in a local zone, the 'zoned' property must be set.
317 /* must be writable by this zone */
325 zfs_dozonecheck(const char *dataset, cred_t *cr)
329 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
332 return (zfs_dozonecheck_impl(dataset, zoned, cr));
336 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
340 rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
341 if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
342 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
345 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
347 return (zfs_dozonecheck_impl(dataset, zoned, cr));
351 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
355 error = zfs_dozonecheck(name, cr);
357 error = secpolicy_zfs(cr);
359 error = dsl_deleg_access(name, perm, cr);
365 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
366 const char *perm, cred_t *cr)
370 error = zfs_dozonecheck_ds(name, ds, cr);
372 error = secpolicy_zfs(cr);
374 error = dsl_deleg_access_impl(ds, perm, cr);
380 * Policy for setting the security label property.
382 * Returns 0 for success, non-zero for access and other errors.
385 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
388 char ds_hexsl[MAXNAMELEN];
389 bslabel_t ds_sl, new_sl;
390 boolean_t new_default = FALSE;
392 int needed_priv = -1;
395 /* First get the existing dataset label. */
396 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
397 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
401 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
404 /* The label must be translatable */
405 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
409 * In a non-global zone, disallow attempts to set a label that
410 * doesn't match that of the zone; otherwise no other checks
413 if (!INGLOBALZONE(curproc)) {
414 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
420 * For global-zone datasets (i.e., those whose zoned property is
421 * "off", verify that the specified new label is valid for the
424 if (dsl_prop_get_integer(name,
425 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
428 if (zfs_check_global_label(name, strval) != 0)
433 * If the existing dataset label is nondefault, check if the
434 * dataset is mounted (label cannot be changed while mounted).
435 * Get the zfs_sb_t; if there isn't one, then the dataset isn't
436 * mounted (or isn't a dataset, doesn't exist, ...).
438 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
440 static char *setsl_tag = "setsl_tag";
443 * Try to own the dataset; abort if there is any error,
444 * (e.g., already mounted, in use, or other error).
446 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
451 dmu_objset_disown(os, setsl_tag);
454 needed_priv = PRIV_FILE_DOWNGRADE_SL;
458 if (hexstr_to_label(strval, &new_sl) != 0)
461 if (blstrictdom(&ds_sl, &new_sl))
462 needed_priv = PRIV_FILE_DOWNGRADE_SL;
463 else if (blstrictdom(&new_sl, &ds_sl))
464 needed_priv = PRIV_FILE_UPGRADE_SL;
466 /* dataset currently has a default label */
468 needed_priv = PRIV_FILE_UPGRADE_SL;
472 if (needed_priv != -1)
473 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
477 #endif /* HAVE_MLSLABEL */
481 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
487 * Check permissions for special properties.
494 * Disallow setting of 'zoned' from within a local zone.
496 if (!INGLOBALZONE(curproc))
501 if (!INGLOBALZONE(curproc)) {
503 char setpoint[MAXNAMELEN];
505 * Unprivileged users are allowed to modify the
506 * quota on things *under* (ie. contained by)
507 * the thing they own.
509 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
512 if (!zoned || strlen(dsname) <= strlen(setpoint))
517 case ZFS_PROP_MLSLABEL:
518 if (!is_system_labeled())
521 if (nvpair_value_string(propval, &strval) == 0) {
524 err = zfs_set_slabel_policy(dsname, strval, CRED());
531 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
535 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
539 error = zfs_dozonecheck(zc->zc_name, cr);
544 * permission to set permissions will be evaluated later in
545 * dsl_deleg_can_allow()
551 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
553 return (zfs_secpolicy_write_perms(zc->zc_name,
554 ZFS_DELEG_PERM_ROLLBACK, cr));
558 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
567 * Generate the current snapshot name from the given objsetid, then
568 * use that name for the secpolicy/zone checks.
570 cp = strchr(zc->zc_name, '@');
573 error = spa_open(zc->zc_name, &spa, FTAG);
577 dp = spa_get_dsl(spa);
578 rw_enter(&dp->dp_config_rwlock, RW_READER);
579 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
580 rw_exit(&dp->dp_config_rwlock);
581 spa_close(spa, FTAG);
585 dsl_dataset_name(ds, zc->zc_name);
587 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
588 ZFS_DELEG_PERM_SEND, cr);
589 dsl_dataset_rele(ds, FTAG);
594 #ifdef HAVE_SMB_SHARE
596 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
601 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
602 NO_FOLLOW, NULL, &vp)) != 0)
605 /* Now make sure mntpnt and dataset are ZFS */
607 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
608 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
609 zc->zc_name) != 0)) {
615 return (dsl_deleg_access(zc->zc_name,
616 ZFS_DELEG_PERM_SHARE, cr));
618 #endif /* HAVE_SMB_SHARE */
621 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
623 #ifdef HAVE_SMB_SHARE
624 if (!INGLOBALZONE(curproc))
627 if (secpolicy_nfs(cr) == 0) {
630 return (zfs_secpolicy_deleg_share(zc, cr));
634 #endif /* HAVE_SMB_SHARE */
638 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
640 #ifdef HAVE_SMB_SHARE
641 if (!INGLOBALZONE(curproc))
644 if (secpolicy_smb(cr) == 0) {
647 return (zfs_secpolicy_deleg_share(zc, cr));
651 #endif /* HAVE_SMB_SHARE */
655 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
660 * Remove the @bla or /bla from the end of the name to get the parent.
662 (void) strncpy(parent, datasetname, parentsize);
663 cp = strrchr(parent, '@');
667 cp = strrchr(parent, '/');
677 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
681 if ((error = zfs_secpolicy_write_perms(name,
682 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
685 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
689 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
691 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
695 * Destroying snapshots with delegated permissions requires
696 * descendent mount and destroy permissions.
697 * Reassemble the full filesystem@snap name so dsl_deleg_access()
698 * can do the correct permission check.
700 * Since this routine is used when doing a recursive destroy of snapshots
701 * and destroying snapshots requires descendent permissions, a successfull
702 * check of the top level snapshot applies to snapshots of all descendent
705 * The target snapshot may not exist when doing a recursive destroy.
706 * In this case fallback to permissions of the parent dataset.
709 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, cred_t *cr)
714 dsname = kmem_asprintf("%s@%s", zc->zc_name, zc->zc_value);
716 error = zfs_secpolicy_destroy_perms(dsname, cr);
718 error = zfs_secpolicy_destroy_perms(zc->zc_name, cr);
725 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
727 char parentname[MAXNAMELEN];
730 if ((error = zfs_secpolicy_write_perms(from,
731 ZFS_DELEG_PERM_RENAME, cr)) != 0)
734 if ((error = zfs_secpolicy_write_perms(from,
735 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
738 if ((error = zfs_get_parent(to, parentname,
739 sizeof (parentname))) != 0)
742 if ((error = zfs_secpolicy_write_perms(parentname,
743 ZFS_DELEG_PERM_CREATE, cr)) != 0)
746 if ((error = zfs_secpolicy_write_perms(parentname,
747 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
754 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
756 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
760 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
762 char parentname[MAXNAMELEN];
766 error = zfs_secpolicy_write_perms(zc->zc_name,
767 ZFS_DELEG_PERM_PROMOTE, cr);
771 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
774 dsl_dataset_t *pclone = NULL;
776 dd = clone->os_dsl_dataset->ds_dir;
778 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
779 error = dsl_dataset_hold_obj(dd->dd_pool,
780 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
781 rw_exit(&dd->dd_pool->dp_config_rwlock);
783 dmu_objset_rele(clone, FTAG);
787 error = zfs_secpolicy_write_perms(zc->zc_name,
788 ZFS_DELEG_PERM_MOUNT, cr);
790 dsl_dataset_name(pclone, parentname);
791 dmu_objset_rele(clone, FTAG);
792 dsl_dataset_rele(pclone, FTAG);
794 error = zfs_secpolicy_write_perms(parentname,
795 ZFS_DELEG_PERM_PROMOTE, cr);
801 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
805 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
806 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
809 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
810 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
813 return (zfs_secpolicy_write_perms(zc->zc_name,
814 ZFS_DELEG_PERM_CREATE, cr));
818 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
820 return (zfs_secpolicy_write_perms(name,
821 ZFS_DELEG_PERM_SNAPSHOT, cr));
825 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
828 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
832 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
834 char parentname[MAXNAMELEN];
837 if ((error = zfs_get_parent(zc->zc_name, parentname,
838 sizeof (parentname))) != 0)
841 if (zc->zc_value[0] != '\0') {
842 if ((error = zfs_secpolicy_write_perms(zc->zc_value,
843 ZFS_DELEG_PERM_CLONE, cr)) != 0)
847 if ((error = zfs_secpolicy_write_perms(parentname,
848 ZFS_DELEG_PERM_CREATE, cr)) != 0)
851 error = zfs_secpolicy_write_perms(parentname,
852 ZFS_DELEG_PERM_MOUNT, cr);
858 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
859 * SYS_CONFIG privilege, which is not available in a local zone.
863 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
865 if (secpolicy_sys_config(cr, B_FALSE) != 0)
872 * Policy for object to name lookups.
876 zfs_secpolicy_diff(zfs_cmd_t *zc, cred_t *cr)
880 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
883 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
888 * Policy for fault injection. Requires all privileges.
892 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
894 return (secpolicy_zinject(cr));
898 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
900 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
902 if (prop == ZPROP_INVAL) {
903 if (!zfs_prop_user(zc->zc_value))
905 return (zfs_secpolicy_write_perms(zc->zc_name,
906 ZFS_DELEG_PERM_USERPROP, cr));
908 return (zfs_secpolicy_setprop(zc->zc_name, prop,
914 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
916 int err = zfs_secpolicy_read(zc, cr);
920 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
923 if (zc->zc_value[0] == 0) {
925 * They are asking about a posix uid/gid. If it's
926 * themself, allow it.
928 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
929 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
930 if (zc->zc_guid == crgetuid(cr))
933 if (groupmember(zc->zc_guid, cr))
938 return (zfs_secpolicy_write_perms(zc->zc_name,
939 userquota_perms[zc->zc_objset_type], cr));
943 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
945 int err = zfs_secpolicy_read(zc, cr);
949 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
952 return (zfs_secpolicy_write_perms(zc->zc_name,
953 userquota_perms[zc->zc_objset_type], cr));
957 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
959 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
964 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
966 return (zfs_secpolicy_write_perms(zc->zc_name,
967 ZFS_DELEG_PERM_HOLD, cr));
971 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
973 return (zfs_secpolicy_write_perms(zc->zc_name,
974 ZFS_DELEG_PERM_RELEASE, cr));
978 * Policy for allowing temporary snapshots to be taken or released
981 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, cred_t *cr)
984 * A temporary snapshot is the same as a snapshot,
985 * hold, destroy and release all rolled into one.
986 * Delegated diff alone is sufficient that we allow this.
990 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
991 ZFS_DELEG_PERM_DIFF, cr)) == 0)
994 error = zfs_secpolicy_snapshot(zc, cr);
996 error = zfs_secpolicy_hold(zc, cr);
998 error = zfs_secpolicy_release(zc, cr);
1000 error = zfs_secpolicy_destroy(zc, cr);
1005 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1008 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1012 nvlist_t *list = NULL;
1015 * Read in and unpack the user-supplied nvlist.
1020 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1022 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1024 kmem_free(packed, size);
1028 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1029 kmem_free(packed, size);
1033 kmem_free(packed, size);
1040 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
1044 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1046 if (size > zc->zc_nvlist_dst_size) {
1047 nvpair_t *more_errors;
1050 if (zc->zc_nvlist_dst_size < 1024)
1053 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
1054 more_errors = nvlist_prev_nvpair(*errors, NULL);
1057 nvpair_t *pair = nvlist_prev_nvpair(*errors,
1059 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
1061 VERIFY(nvlist_size(*errors, &size,
1062 NV_ENCODE_NATIVE) == 0);
1063 } while (size > zc->zc_nvlist_dst_size);
1065 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
1066 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
1067 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1068 ASSERT(size <= zc->zc_nvlist_dst_size);
1075 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1077 char *packed = NULL;
1081 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
1083 if (size > zc->zc_nvlist_dst_size) {
1086 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1087 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
1089 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1090 size, zc->zc_iflags) != 0)
1092 kmem_free(packed, size);
1095 zc->zc_nvlist_dst_size = size;
1100 get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
1105 error = dmu_objset_hold(dsname, FTAG, &os);
1108 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1109 dmu_objset_rele(os, FTAG);
1113 mutex_enter(&os->os_user_ptr_lock);
1114 *zsbp = dmu_objset_get_user(os);
1115 if (*zsbp && (*zsbp)->z_sb) {
1116 atomic_inc(&((*zsbp)->z_sb->s_active));
1120 mutex_exit(&os->os_user_ptr_lock);
1121 dmu_objset_rele(os, FTAG);
1126 * Find a zfs_sb_t for a mounted filesystem, or create our own, in which
1127 * case its z_sb will be NULL, and it will be opened as the owner.
1130 zfs_sb_hold(const char *name, void *tag, zfs_sb_t **zsbp, boolean_t writer)
1134 if (get_zfs_sb(name, zsbp) != 0)
1135 error = zfs_sb_create(name, zsbp);
1137 rrw_enter(&(*zsbp)->z_teardown_lock, (writer) ? RW_WRITER :
1139 if ((*zsbp)->z_unmounted) {
1141 * XXX we could probably try again, since the unmounting
1142 * thread should be just about to disassociate the
1143 * objset from the zfsvfs.
1145 rrw_exit(&(*zsbp)->z_teardown_lock, tag);
1153 zfs_sb_rele(zfs_sb_t *zsb, void *tag)
1155 rrw_exit(&zsb->z_teardown_lock, tag);
1158 deactivate_super(zsb->z_sb);
1160 dmu_objset_disown(zsb->z_os, zsb);
1166 zfs_ioc_pool_create(zfs_cmd_t *zc)
1169 nvlist_t *config, *props = NULL;
1170 nvlist_t *rootprops = NULL;
1171 nvlist_t *zplprops = NULL;
1174 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1175 zc->zc_iflags, &config)))
1178 if (zc->zc_nvlist_src_size != 0 && (error =
1179 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1180 zc->zc_iflags, &props))) {
1181 nvlist_free(config);
1186 nvlist_t *nvl = NULL;
1187 uint64_t version = SPA_VERSION;
1189 (void) nvlist_lookup_uint64(props,
1190 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1191 if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1193 goto pool_props_bad;
1195 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1197 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1199 nvlist_free(config);
1203 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1205 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1206 error = zfs_fill_zplprops_root(version, rootprops,
1209 goto pool_props_bad;
1212 buf = history_str_get(zc);
1214 error = spa_create(zc->zc_name, config, props, buf, zplprops);
1217 * Set the remaining root properties
1219 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1220 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1221 (void) spa_destroy(zc->zc_name);
1224 history_str_free(buf);
1227 nvlist_free(rootprops);
1228 nvlist_free(zplprops);
1229 nvlist_free(config);
1236 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1239 zfs_log_history(zc);
1240 error = spa_destroy(zc->zc_name);
1242 zvol_remove_minors(zc->zc_name);
1247 zfs_ioc_pool_import(zfs_cmd_t *zc)
1249 nvlist_t *config, *props = NULL;
1253 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1254 zc->zc_iflags, &config)) != 0)
1257 if (zc->zc_nvlist_src_size != 0 && (error =
1258 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1259 zc->zc_iflags, &props))) {
1260 nvlist_free(config);
1264 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1265 guid != zc->zc_guid)
1268 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1270 if (zc->zc_nvlist_dst != 0) {
1273 if ((err = put_nvlist(zc, config)) != 0)
1278 zvol_create_minors(zc->zc_name);
1280 nvlist_free(config);
1289 zfs_ioc_pool_export(zfs_cmd_t *zc)
1292 boolean_t force = (boolean_t)zc->zc_cookie;
1293 boolean_t hardforce = (boolean_t)zc->zc_guid;
1295 zfs_log_history(zc);
1296 error = spa_export(zc->zc_name, NULL, force, hardforce);
1298 zvol_remove_minors(zc->zc_name);
1303 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1308 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1311 error = put_nvlist(zc, configs);
1313 nvlist_free(configs);
1319 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1325 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1326 sizeof (zc->zc_value));
1328 if (config != NULL) {
1329 ret = put_nvlist(zc, config);
1330 nvlist_free(config);
1333 * The config may be present even if 'error' is non-zero.
1334 * In this case we return success, and preserve the real errno
1337 zc->zc_cookie = error;
1346 * Try to import the given pool, returning pool stats as appropriate so that
1347 * user land knows which devices are available and overall pool health.
1350 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1352 nvlist_t *tryconfig, *config;
1355 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1356 zc->zc_iflags, &tryconfig)) != 0)
1359 config = spa_tryimport(tryconfig);
1361 nvlist_free(tryconfig);
1366 error = put_nvlist(zc, config);
1367 nvlist_free(config);
1374 * zc_name name of the pool
1375 * zc_cookie scan func (pool_scan_func_t)
1378 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1383 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1386 if (zc->zc_cookie == POOL_SCAN_NONE)
1387 error = spa_scan_stop(spa);
1389 error = spa_scan(spa, zc->zc_cookie);
1391 spa_close(spa, FTAG);
1397 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1402 error = spa_open(zc->zc_name, &spa, FTAG);
1405 spa_close(spa, FTAG);
1411 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1416 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1419 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1420 spa_close(spa, FTAG);
1424 spa_upgrade(spa, zc->zc_cookie);
1425 spa_close(spa, FTAG);
1431 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1438 if ((size = zc->zc_history_len) == 0)
1441 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1444 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1445 spa_close(spa, FTAG);
1449 hist_buf = vmem_alloc(size, KM_SLEEP);
1450 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1451 &zc->zc_history_len, hist_buf)) == 0) {
1452 error = ddi_copyout(hist_buf,
1453 (void *)(uintptr_t)zc->zc_history,
1454 zc->zc_history_len, zc->zc_iflags);
1457 spa_close(spa, FTAG);
1458 vmem_free(hist_buf, size);
1463 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1467 if ((error = dsl_dsobj_to_dsname(zc->zc_name,zc->zc_obj,zc->zc_value)))
1475 * zc_name name of filesystem
1476 * zc_obj object to find
1479 * zc_value name of object
1482 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1487 /* XXX reading from objset not owned */
1488 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1490 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1491 dmu_objset_rele(os, FTAG);
1494 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1495 sizeof (zc->zc_value));
1496 dmu_objset_rele(os, FTAG);
1503 * zc_name name of filesystem
1504 * zc_obj object to find
1507 * zc_stat stats on object
1508 * zc_value path to object
1511 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1516 /* XXX reading from objset not owned */
1517 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1519 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1520 dmu_objset_rele(os, FTAG);
1523 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1524 sizeof (zc->zc_value));
1525 dmu_objset_rele(os, FTAG);
1531 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1535 nvlist_t *config, **l2cache, **spares;
1536 uint_t nl2cache = 0, nspares = 0;
1538 error = spa_open(zc->zc_name, &spa, FTAG);
1542 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1543 zc->zc_iflags, &config);
1544 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1545 &l2cache, &nl2cache);
1547 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1551 * A root pool with concatenated devices is not supported.
1552 * Thus, can not add a device to a root pool.
1554 * Intent log device can not be added to a rootpool because
1555 * during mountroot, zil is replayed, a seperated log device
1556 * can not be accessed during the mountroot time.
1558 * l2cache and spare devices are ok to be added to a rootpool.
1560 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1561 nvlist_free(config);
1562 spa_close(spa, FTAG);
1567 error = spa_vdev_add(spa, config);
1568 nvlist_free(config);
1570 spa_close(spa, FTAG);
1576 * zc_name name of the pool
1577 * zc_nvlist_conf nvlist of devices to remove
1578 * zc_cookie to stop the remove?
1581 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1586 error = spa_open(zc->zc_name, &spa, FTAG);
1589 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1590 spa_close(spa, FTAG);
1595 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1599 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1601 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1603 switch (zc->zc_cookie) {
1604 case VDEV_STATE_ONLINE:
1605 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1608 case VDEV_STATE_OFFLINE:
1609 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1612 case VDEV_STATE_FAULTED:
1613 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1614 zc->zc_obj != VDEV_AUX_EXTERNAL)
1615 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1617 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1620 case VDEV_STATE_DEGRADED:
1621 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1622 zc->zc_obj != VDEV_AUX_EXTERNAL)
1623 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1625 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1631 zc->zc_cookie = newstate;
1632 spa_close(spa, FTAG);
1637 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1640 int replacing = zc->zc_cookie;
1644 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1647 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1648 zc->zc_iflags, &config)) == 0) {
1649 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1650 nvlist_free(config);
1653 spa_close(spa, FTAG);
1658 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1663 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1666 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1668 spa_close(spa, FTAG);
1673 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1676 nvlist_t *config, *props = NULL;
1678 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1680 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1683 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1684 zc->zc_iflags, &config))) {
1685 spa_close(spa, FTAG);
1689 if (zc->zc_nvlist_src_size != 0 && (error =
1690 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1691 zc->zc_iflags, &props))) {
1692 spa_close(spa, FTAG);
1693 nvlist_free(config);
1697 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1699 spa_close(spa, FTAG);
1701 nvlist_free(config);
1708 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1711 char *path = zc->zc_value;
1712 uint64_t guid = zc->zc_guid;
1715 error = spa_open(zc->zc_name, &spa, FTAG);
1719 error = spa_vdev_setpath(spa, guid, path);
1720 spa_close(spa, FTAG);
1725 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1728 char *fru = zc->zc_value;
1729 uint64_t guid = zc->zc_guid;
1732 error = spa_open(zc->zc_name, &spa, FTAG);
1736 error = spa_vdev_setfru(spa, guid, fru);
1737 spa_close(spa, FTAG);
1742 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1747 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1749 if (zc->zc_nvlist_dst != 0 &&
1750 (error = dsl_prop_get_all(os, &nv)) == 0) {
1751 dmu_objset_stats(os, nv);
1753 * NB: zvol_get_stats() will read the objset contents,
1754 * which we aren't supposed to do with a
1755 * DS_MODE_USER hold, because it could be
1756 * inconsistent. So this is a bit of a workaround...
1757 * XXX reading with out owning
1759 if (!zc->zc_objset_stats.dds_inconsistent) {
1760 if (dmu_objset_type(os) == DMU_OST_ZVOL)
1761 error = zvol_get_stats(os, nv);
1764 error = put_nvlist(zc, nv);
1773 * zc_name name of filesystem
1774 * zc_nvlist_dst_size size of buffer for property nvlist
1777 * zc_objset_stats stats
1778 * zc_nvlist_dst property nvlist
1779 * zc_nvlist_dst_size size of property nvlist
1782 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1784 objset_t *os = NULL;
1787 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1790 error = zfs_ioc_objset_stats_impl(zc, os);
1792 dmu_objset_rele(os, FTAG);
1799 * zc_name name of filesystem
1800 * zc_nvlist_dst_size size of buffer for property nvlist
1803 * zc_nvlist_dst received property nvlist
1804 * zc_nvlist_dst_size size of received property nvlist
1806 * Gets received properties (distinct from local properties on or after
1807 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1808 * local property values.
1811 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1813 objset_t *os = NULL;
1817 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1821 * Without this check, we would return local property values if the
1822 * caller has not already received properties on or after
1823 * SPA_VERSION_RECVD_PROPS.
1825 if (!dsl_prop_get_hasrecvd(os)) {
1826 dmu_objset_rele(os, FTAG);
1830 if (zc->zc_nvlist_dst != 0 &&
1831 (error = dsl_prop_get_received(os, &nv)) == 0) {
1832 error = put_nvlist(zc, nv);
1836 dmu_objset_rele(os, FTAG);
1841 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1847 * zfs_get_zplprop() will either find a value or give us
1848 * the default value (if there is one).
1850 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1852 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1858 * zc_name name of filesystem
1859 * zc_nvlist_dst_size size of buffer for zpl property nvlist
1862 * zc_nvlist_dst zpl property nvlist
1863 * zc_nvlist_dst_size size of zpl property nvlist
1866 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1871 /* XXX reading without owning */
1872 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1875 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1878 * NB: nvl_add_zplprop() will read the objset contents,
1879 * which we aren't supposed to do with a DS_MODE_USER
1880 * hold, because it could be inconsistent.
1882 if (zc->zc_nvlist_dst != 0 &&
1883 !zc->zc_objset_stats.dds_inconsistent &&
1884 dmu_objset_type(os) == DMU_OST_ZFS) {
1887 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1888 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1889 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1890 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1891 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1892 err = put_nvlist(zc, nv);
1897 dmu_objset_rele(os, FTAG);
1902 dataset_name_hidden(const char *name)
1905 * Skip over datasets that are not visible in this zone,
1906 * internal datasets (which have a $ in their name), and
1907 * temporary datasets (which have a % in their name).
1909 if (strchr(name, '$') != NULL)
1911 if (strchr(name, '%') != NULL)
1913 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1920 * zc_name name of filesystem
1921 * zc_cookie zap cursor
1922 * zc_nvlist_dst_size size of buffer for property nvlist
1925 * zc_name name of next filesystem
1926 * zc_cookie zap cursor
1927 * zc_objset_stats stats
1928 * zc_nvlist_dst property nvlist
1929 * zc_nvlist_dst_size size of property nvlist
1932 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1937 size_t orig_len = strlen(zc->zc_name);
1940 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
1941 if (error == ENOENT)
1946 p = strrchr(zc->zc_name, '/');
1947 if (p == NULL || p[1] != '\0')
1948 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1949 p = zc->zc_name + strlen(zc->zc_name);
1952 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
1953 * but is not declared void because its called by dmu_objset_find().
1955 if (zc->zc_cookie == 0) {
1956 uint64_t cookie = 0;
1957 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1959 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1960 (void) dmu_objset_prefetch(p, NULL);
1964 error = dmu_dir_list_next(os,
1965 sizeof (zc->zc_name) - (p - zc->zc_name), p,
1966 NULL, &zc->zc_cookie);
1967 if (error == ENOENT)
1969 } while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1970 !(zc->zc_iflags & FKIOCTL));
1971 dmu_objset_rele(os, FTAG);
1974 * If it's an internal dataset (ie. with a '$' in its name),
1975 * don't try to get stats for it, otherwise we'll return ENOENT.
1977 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1978 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1979 if (error == ENOENT) {
1980 /* We lost a race with destroy, get the next one. */
1981 zc->zc_name[orig_len] = '\0';
1990 * zc_name name of filesystem
1991 * zc_cookie zap cursor
1992 * zc_nvlist_dst_size size of buffer for property nvlist
1995 * zc_name name of next snapshot
1996 * zc_objset_stats stats
1997 * zc_nvlist_dst property nvlist
1998 * zc_nvlist_dst_size size of property nvlist
2001 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2007 if (zc->zc_cookie == 0)
2008 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
2009 NULL, DS_FIND_SNAPSHOTS);
2011 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2013 return (error == ENOENT ? ESRCH : error);
2016 * A dataset name of maximum length cannot have any snapshots,
2017 * so exit immediately.
2019 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2020 dmu_objset_rele(os, FTAG);
2024 error = dmu_snapshot_list_next(os,
2025 sizeof (zc->zc_name) - strlen(zc->zc_name),
2026 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2031 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2034 * Since we probably don't have a hold on this snapshot,
2035 * it's possible that the objsetid could have been destroyed
2036 * and reused for a new objset. It's OK if this happens during
2037 * a zfs send operation, since the new createtxg will be
2038 * beyond the range we're interested in.
2040 rw_enter(&dp->dp_config_rwlock, RW_READER);
2041 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2042 rw_exit(&dp->dp_config_rwlock);
2044 if (error == ENOENT) {
2045 /* Racing with destroy, get the next one. */
2046 *strchr(zc->zc_name, '@') = '\0';
2047 dmu_objset_rele(os, FTAG);
2053 error = dmu_objset_from_ds(ds, &ossnap);
2055 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2056 dsl_dataset_rele(ds, FTAG);
2058 } else if (error == ENOENT) {
2062 dmu_objset_rele(os, FTAG);
2063 /* if we failed, undo the @ that we tacked on to zc_name */
2065 *strchr(zc->zc_name, '@') = '\0';
2070 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2072 const char *propname = nvpair_name(pair);
2074 unsigned int vallen;
2077 zfs_userquota_prop_t type;
2083 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2085 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2086 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2092 * A correctly constructed propname is encoded as
2093 * userquota@<rid>-<domain>.
2095 if ((dash = strchr(propname, '-')) == NULL ||
2096 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2105 err = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
2107 err = zfs_set_userquota(zsb, type, domain, rid, quota);
2108 zfs_sb_rele(zsb, FTAG);
2115 * If the named property is one that has a special function to set its value,
2116 * return 0 on success and a positive error code on failure; otherwise if it is
2117 * not one of the special properties handled by this function, return -1.
2119 * XXX: It would be better for callers of the property interface if we handled
2120 * these special cases in dsl_prop.c (in the dsl layer).
2123 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2126 const char *propname = nvpair_name(pair);
2127 zfs_prop_t prop = zfs_name_to_prop(propname);
2131 if (prop == ZPROP_INVAL) {
2132 if (zfs_prop_userquota(propname))
2133 return (zfs_prop_set_userquota(dsname, pair));
2137 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2139 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2140 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2144 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2147 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2150 case ZFS_PROP_QUOTA:
2151 err = dsl_dir_set_quota(dsname, source, intval);
2153 case ZFS_PROP_REFQUOTA:
2154 err = dsl_dataset_set_quota(dsname, source, intval);
2156 case ZFS_PROP_RESERVATION:
2157 err = dsl_dir_set_reservation(dsname, source, intval);
2159 case ZFS_PROP_REFRESERVATION:
2160 err = dsl_dataset_set_reservation(dsname, source, intval);
2162 case ZFS_PROP_VOLSIZE:
2163 err = zvol_set_volsize(dsname, intval);
2165 case ZFS_PROP_VERSION:
2169 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2172 err = zfs_set_version(zsb, intval);
2173 zfs_sb_rele(zsb, FTAG);
2175 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2178 zc = kmem_zalloc(sizeof (zfs_cmd_t),
2179 KM_SLEEP | KM_NODEBUG);
2180 (void) strcpy(zc->zc_name, dsname);
2181 (void) zfs_ioc_userspace_upgrade(zc);
2182 kmem_free(zc, sizeof (zfs_cmd_t));
2195 * This function is best effort. If it fails to set any of the given properties,
2196 * it continues to set as many as it can and returns the first error
2197 * encountered. If the caller provides a non-NULL errlist, it also gives the
2198 * complete list of names of all the properties it failed to set along with the
2199 * corresponding error numbers. The caller is responsible for freeing the
2202 * If every property is set successfully, zero is returned and the list pointed
2203 * at by errlist is NULL.
2206 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2214 nvlist_t *genericnvl;
2218 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2219 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2220 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2224 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2225 const char *propname = nvpair_name(pair);
2226 zfs_prop_t prop = zfs_name_to_prop(propname);
2229 /* decode the property value */
2231 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2233 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2234 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2239 /* Validate value type */
2240 if (err == 0 && prop == ZPROP_INVAL) {
2241 if (zfs_prop_user(propname)) {
2242 if (nvpair_type(propval) != DATA_TYPE_STRING)
2244 } else if (zfs_prop_userquota(propname)) {
2245 if (nvpair_type(propval) !=
2246 DATA_TYPE_UINT64_ARRAY)
2249 } else if (err == 0) {
2250 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2251 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2253 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2256 VERIFY(nvpair_value_uint64(propval,
2259 switch (zfs_prop_get_type(prop)) {
2260 case PROP_TYPE_NUMBER:
2262 case PROP_TYPE_STRING:
2265 case PROP_TYPE_INDEX:
2266 if (zfs_prop_index_to_string(prop,
2267 intval, &unused) != 0)
2272 "unknown property type");
2279 /* Validate permissions */
2281 err = zfs_check_settable(dsname, pair, CRED());
2284 err = zfs_prop_set_special(dsname, source, pair);
2287 * For better performance we build up a list of
2288 * properties to set in a single transaction.
2290 err = nvlist_add_nvpair(genericnvl, pair);
2291 } else if (err != 0 && nvl != retrynvl) {
2293 * This may be a spurious error caused by
2294 * receiving quota and reservation out of order.
2295 * Try again in a second pass.
2297 err = nvlist_add_nvpair(retrynvl, pair);
2302 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2305 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2310 if (!nvlist_empty(genericnvl) &&
2311 dsl_props_set(dsname, source, genericnvl) != 0) {
2313 * If this fails, we still want to set as many properties as we
2314 * can, so try setting them individually.
2317 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2318 const char *propname = nvpair_name(pair);
2322 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2324 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2325 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2329 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2330 VERIFY(nvpair_value_string(propval,
2332 err = dsl_prop_set(dsname, propname, source, 1,
2333 strlen(strval) + 1, strval);
2335 VERIFY(nvpair_value_uint64(propval,
2337 err = dsl_prop_set(dsname, propname, source, 8,
2342 VERIFY(nvlist_add_int32(errors, propname,
2347 nvlist_free(genericnvl);
2348 nvlist_free(retrynvl);
2350 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2351 nvlist_free(errors);
2354 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2357 if (errlist == NULL)
2358 nvlist_free(errors);
2366 * Check that all the properties are valid user properties.
2369 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2371 nvpair_t *pair = NULL;
2374 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2375 const char *propname = nvpair_name(pair);
2378 if (!zfs_prop_user(propname) ||
2379 nvpair_type(pair) != DATA_TYPE_STRING)
2382 if ((error = zfs_secpolicy_write_perms(fsname,
2383 ZFS_DELEG_PERM_USERPROP, CRED())))
2386 if (strlen(propname) >= ZAP_MAXNAMELEN)
2387 return (ENAMETOOLONG);
2389 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2390 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2397 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2401 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2404 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2405 if (nvlist_exists(skipped, nvpair_name(pair)))
2408 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2413 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2417 nvlist_t *cleared_props = NULL;
2418 props_skip(props, skipped, &cleared_props);
2419 if (!nvlist_empty(cleared_props)) {
2421 * Acts on local properties until the dataset has received
2422 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2424 zprop_source_t flags = (ZPROP_SRC_NONE |
2425 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2426 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2428 nvlist_free(cleared_props);
2434 * zc_name name of filesystem
2435 * zc_value name of property to set
2436 * zc_nvlist_src{_size} nvlist of properties to apply
2437 * zc_cookie received properties flag
2440 * zc_nvlist_dst{_size} error for each unapplied received property
2443 zfs_ioc_set_prop(zfs_cmd_t *zc)
2446 boolean_t received = zc->zc_cookie;
2447 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2449 nvlist_t *errors = NULL;
2452 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2453 zc->zc_iflags, &nvl)) != 0)
2457 nvlist_t *origprops;
2460 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2461 if (dsl_prop_get_received(os, &origprops) == 0) {
2462 (void) clear_received_props(os,
2463 zc->zc_name, origprops, nvl);
2464 nvlist_free(origprops);
2467 dsl_prop_set_hasrecvd(os);
2468 dmu_objset_rele(os, FTAG);
2472 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2474 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2475 (void) put_nvlist(zc, errors);
2478 nvlist_free(errors);
2485 * zc_name name of filesystem
2486 * zc_value name of property to inherit
2487 * zc_cookie revert to received value if TRUE
2492 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2494 const char *propname = zc->zc_value;
2495 zfs_prop_t prop = zfs_name_to_prop(propname);
2496 boolean_t received = zc->zc_cookie;
2497 zprop_source_t source = (received
2498 ? ZPROP_SRC_NONE /* revert to received value, if any */
2499 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2508 * zfs_prop_set_special() expects properties in the form of an
2509 * nvpair with type info.
2511 if (prop == ZPROP_INVAL) {
2512 if (!zfs_prop_user(propname))
2515 type = PROP_TYPE_STRING;
2516 } else if (prop == ZFS_PROP_VOLSIZE ||
2517 prop == ZFS_PROP_VERSION) {
2520 type = zfs_prop_get_type(prop);
2523 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2526 case PROP_TYPE_STRING:
2527 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2529 case PROP_TYPE_NUMBER:
2530 case PROP_TYPE_INDEX:
2531 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2538 pair = nvlist_next_nvpair(dummy, NULL);
2539 err = zfs_prop_set_special(zc->zc_name, source, pair);
2542 return (err); /* special property already handled */
2545 * Only check this in the non-received case. We want to allow
2546 * 'inherit -S' to revert non-inheritable properties like quota
2547 * and reservation to the received or default values even though
2548 * they are not considered inheritable.
2550 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2554 /* the property name has been validated by zfs_secpolicy_inherit() */
2555 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2559 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2566 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2567 zc->zc_iflags, &props)))
2571 * If the only property is the configfile, then just do a spa_lookup()
2572 * to handle the faulted case.
2574 pair = nvlist_next_nvpair(props, NULL);
2575 if (pair != NULL && strcmp(nvpair_name(pair),
2576 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2577 nvlist_next_nvpair(props, pair) == NULL) {
2578 mutex_enter(&spa_namespace_lock);
2579 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2580 spa_configfile_set(spa, props, B_FALSE);
2581 spa_config_sync(spa, B_FALSE, B_TRUE);
2583 mutex_exit(&spa_namespace_lock);
2590 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2595 error = spa_prop_set(spa, props);
2598 spa_close(spa, FTAG);
2604 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2608 nvlist_t *nvp = NULL;
2610 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2612 * If the pool is faulted, there may be properties we can still
2613 * get (such as altroot and cachefile), so attempt to get them
2616 mutex_enter(&spa_namespace_lock);
2617 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2618 error = spa_prop_get(spa, &nvp);
2619 mutex_exit(&spa_namespace_lock);
2621 error = spa_prop_get(spa, &nvp);
2622 spa_close(spa, FTAG);
2625 if (error == 0 && zc->zc_nvlist_dst != 0)
2626 error = put_nvlist(zc, nvp);
2636 * zc_name name of volume
2641 zfs_ioc_create_minor(zfs_cmd_t *zc)
2643 return (zvol_create_minor(zc->zc_name));
2648 * zc_name name of volume
2653 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2655 return (zvol_remove_minor(zc->zc_name));
2660 * zc_name name of filesystem
2661 * zc_nvlist_src{_size} nvlist of delegated permissions
2662 * zc_perm_action allow/unallow flag
2667 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2670 nvlist_t *fsaclnv = NULL;
2672 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2673 zc->zc_iflags, &fsaclnv)) != 0)
2677 * Verify nvlist is constructed correctly
2679 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2680 nvlist_free(fsaclnv);
2685 * If we don't have PRIV_SYS_MOUNT, then validate
2686 * that user is allowed to hand out each permission in
2690 error = secpolicy_zfs(CRED());
2692 if (zc->zc_perm_action == B_FALSE) {
2693 error = dsl_deleg_can_allow(zc->zc_name,
2696 error = dsl_deleg_can_unallow(zc->zc_name,
2702 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2704 nvlist_free(fsaclnv);
2710 * zc_name name of filesystem
2713 * zc_nvlist_src{_size} nvlist of delegated permissions
2716 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2721 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2722 error = put_nvlist(zc, nvp);
2729 #ifdef HAVE_SNAPSHOT
2731 * Search the vfs list for a specified resource. Returns a pointer to it
2732 * or NULL if no suitable entry is found. The caller of this routine
2733 * is responsible for releasing the returned vfs pointer.
2736 zfs_get_vfs(const char *resource)
2739 struct vfs *vfs_found = NULL;
2741 vfs_list_read_lock();
2744 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2749 vfsp = vfsp->vfs_next;
2750 } while (vfsp != rootvfs);
2754 #endif /* HAVE_SNAPSHOT */
2758 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2760 zfs_creat_t *zct = arg;
2762 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2765 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2769 * createprops list of properties requested by creator
2770 * default_zplver zpl version to use if unspecified in createprops
2771 * fuids_ok fuids allowed in this version of the spa?
2772 * os parent objset pointer (NULL if root fs)
2775 * zplprops values for the zplprops we attach to the master node object
2776 * is_ci true if requested file system will be purely case-insensitive
2778 * Determine the settings for utf8only, normalization and
2779 * casesensitivity. Specific values may have been requested by the
2780 * creator and/or we can inherit values from the parent dataset. If
2781 * the file system is of too early a vintage, a creator can not
2782 * request settings for these properties, even if the requested
2783 * setting is the default value. We don't actually want to create dsl
2784 * properties for these, so remove them from the source nvlist after
2788 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2789 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2790 nvlist_t *zplprops, boolean_t *is_ci)
2792 uint64_t sense = ZFS_PROP_UNDEFINED;
2793 uint64_t norm = ZFS_PROP_UNDEFINED;
2794 uint64_t u8 = ZFS_PROP_UNDEFINED;
2796 ASSERT(zplprops != NULL);
2799 * Pull out creator prop choices, if any.
2802 (void) nvlist_lookup_uint64(createprops,
2803 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2804 (void) nvlist_lookup_uint64(createprops,
2805 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2806 (void) nvlist_remove_all(createprops,
2807 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2808 (void) nvlist_lookup_uint64(createprops,
2809 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2810 (void) nvlist_remove_all(createprops,
2811 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2812 (void) nvlist_lookup_uint64(createprops,
2813 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2814 (void) nvlist_remove_all(createprops,
2815 zfs_prop_to_name(ZFS_PROP_CASE));
2819 * If the zpl version requested is whacky or the file system
2820 * or pool is version is too "young" to support normalization
2821 * and the creator tried to set a value for one of the props,
2824 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2825 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2826 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2827 (zplver < ZPL_VERSION_NORMALIZATION &&
2828 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2829 sense != ZFS_PROP_UNDEFINED)))
2833 * Put the version in the zplprops
2835 VERIFY(nvlist_add_uint64(zplprops,
2836 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2838 if (norm == ZFS_PROP_UNDEFINED)
2839 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2840 VERIFY(nvlist_add_uint64(zplprops,
2841 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2844 * If we're normalizing, names must always be valid UTF-8 strings.
2848 if (u8 == ZFS_PROP_UNDEFINED)
2849 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2850 VERIFY(nvlist_add_uint64(zplprops,
2851 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2853 if (sense == ZFS_PROP_UNDEFINED)
2854 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2855 VERIFY(nvlist_add_uint64(zplprops,
2856 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2859 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2865 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2866 nvlist_t *zplprops, boolean_t *is_ci)
2868 boolean_t fuids_ok, sa_ok;
2869 uint64_t zplver = ZPL_VERSION;
2870 objset_t *os = NULL;
2871 char parentname[MAXNAMELEN];
2877 (void) strlcpy(parentname, dataset, sizeof (parentname));
2878 cp = strrchr(parentname, '/');
2882 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2885 spa_vers = spa_version(spa);
2886 spa_close(spa, FTAG);
2888 zplver = zfs_zpl_version_map(spa_vers);
2889 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2890 sa_ok = (zplver >= ZPL_VERSION_SA);
2893 * Open parent object set so we can inherit zplprop values.
2895 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2898 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2900 dmu_objset_rele(os, FTAG);
2905 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2906 nvlist_t *zplprops, boolean_t *is_ci)
2910 uint64_t zplver = ZPL_VERSION;
2913 zplver = zfs_zpl_version_map(spa_vers);
2914 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2915 sa_ok = (zplver >= ZPL_VERSION_SA);
2917 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2918 createprops, zplprops, is_ci);
2924 * zc_objset_type type of objset to create (fs vs zvol)
2925 * zc_name name of new objset
2926 * zc_value name of snapshot to clone from (may be empty)
2927 * zc_nvlist_src{_size} nvlist of properties to apply
2932 zfs_ioc_create(zfs_cmd_t *zc)
2937 nvlist_t *nvprops = NULL;
2938 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2939 dmu_objset_type_t type = zc->zc_objset_type;
2944 cbfunc = zfs_create_cb;
2948 cbfunc = zvol_create_cb;
2955 if (strchr(zc->zc_name, '@') ||
2956 strchr(zc->zc_name, '%'))
2959 if (zc->zc_nvlist_src != 0 &&
2960 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2961 zc->zc_iflags, &nvprops)) != 0)
2964 zct.zct_zplprops = NULL;
2965 zct.zct_props = nvprops;
2967 if (zc->zc_value[0] != '\0') {
2969 * We're creating a clone of an existing snapshot.
2971 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2972 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2973 nvlist_free(nvprops);
2977 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2979 nvlist_free(nvprops);
2983 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2984 dmu_objset_rele(clone, FTAG);
2986 nvlist_free(nvprops);
2990 boolean_t is_insensitive = B_FALSE;
2992 if (cbfunc == NULL) {
2993 nvlist_free(nvprops);
2997 if (type == DMU_OST_ZVOL) {
2998 uint64_t volsize, volblocksize;
3000 if (nvprops == NULL ||
3001 nvlist_lookup_uint64(nvprops,
3002 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
3004 nvlist_free(nvprops);
3008 if ((error = nvlist_lookup_uint64(nvprops,
3009 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3010 &volblocksize)) != 0 && error != ENOENT) {
3011 nvlist_free(nvprops);
3016 volblocksize = zfs_prop_default_numeric(
3017 ZFS_PROP_VOLBLOCKSIZE);
3019 if ((error = zvol_check_volblocksize(
3020 volblocksize)) != 0 ||
3021 (error = zvol_check_volsize(volsize,
3022 volblocksize)) != 0) {
3023 nvlist_free(nvprops);
3026 } else if (type == DMU_OST_ZFS) {
3030 * We have to have normalization and
3031 * case-folding flags correct when we do the
3032 * file system creation, so go figure them out
3035 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3036 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3037 error = zfs_fill_zplprops(zc->zc_name, nvprops,
3038 zct.zct_zplprops, &is_insensitive);
3040 nvlist_free(nvprops);
3041 nvlist_free(zct.zct_zplprops);
3045 error = dmu_objset_create(zc->zc_name, type,
3046 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3047 nvlist_free(zct.zct_zplprops);
3051 * It would be nice to do this atomically.
3054 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3057 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3059 nvlist_free(nvprops);
3065 * zc_name name of filesystem
3066 * zc_value short name of snapshot
3067 * zc_cookie recursive flag
3068 * zc_nvlist_src[_size] property list
3071 * zc_value short snapname (i.e. part after the '@')
3074 zfs_ioc_snapshot(zfs_cmd_t *zc)
3076 nvlist_t *nvprops = NULL;
3078 boolean_t recursive = zc->zc_cookie;
3080 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3083 if (zc->zc_nvlist_src != 0 &&
3084 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3085 zc->zc_iflags, &nvprops)) != 0)
3088 error = zfs_check_userprops(zc->zc_name, nvprops);
3092 if (!nvlist_empty(nvprops) &&
3093 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3098 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3099 nvprops, recursive, B_FALSE, -1);
3102 nvlist_free(nvprops);
3107 zfs_unmount_snap(const char *name, void *arg)
3109 #ifdef HAVE_SNAPSHOT
3113 char *snapname = arg;
3114 char *fullname = kmem_asprintf("%s@%s", name, snapname);
3115 vfsp = zfs_get_vfs(fullname);
3117 } else if (strchr(name, '@')) {
3118 vfsp = zfs_get_vfs(name);
3123 * Always force the unmount for snapshots.
3125 int flag = MS_FORCE;
3128 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
3133 if ((err = dounmount(vfsp, flag, kcred)) != 0)
3136 #endif /* HAVE_SNAPSHOT */
3142 * zc_name name of filesystem
3143 * zc_value short name of snapshot
3144 * zc_defer_destroy mark for deferred destroy
3149 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
3153 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3155 err = dmu_objset_find(zc->zc_name,
3156 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
3159 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
3160 zc->zc_defer_destroy));
3165 * zc_name name of dataset to destroy
3166 * zc_objset_type type of objset
3167 * zc_defer_destroy mark for deferred destroy
3172 zfs_ioc_destroy(zfs_cmd_t *zc)
3175 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3176 err = zfs_unmount_snap(zc->zc_name, NULL);
3181 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3182 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3183 (void) zvol_remove_minor(zc->zc_name);
3189 * zc_name name of dataset to rollback (to most recent snapshot)
3194 zfs_ioc_rollback(zfs_cmd_t *zc)
3196 dsl_dataset_t *ds, *clone;
3201 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3205 /* must not be a snapshot */
3206 if (dsl_dataset_is_snapshot(ds)) {
3207 dsl_dataset_rele(ds, FTAG);
3211 /* must have a most recent snapshot */
3212 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3213 dsl_dataset_rele(ds, FTAG);
3218 * Create clone of most recent snapshot.
3220 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3221 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3225 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3232 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
3233 error = zfs_suspend_fs(zsb);
3237 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3238 error = dsl_dataset_clone_swap(clone, ds,
3240 dsl_dataset_disown(ds, FTAG);
3245 resume_err = zfs_resume_fs(zsb, zc->zc_name);
3246 error = error ? error : resume_err;
3248 deactivate_super(zsb->z_sb);
3250 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3251 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3252 dsl_dataset_disown(ds, FTAG);
3260 * Destroy clone (which also closes it).
3262 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3265 strfree(clone_name);
3267 dsl_dataset_rele(ds, FTAG);
3273 * zc_name old name of dataset
3274 * zc_value new name of dataset
3275 * zc_cookie recursive flag (only valid for snapshots)
3280 zfs_ioc_rename(zfs_cmd_t *zc)
3282 boolean_t recursive = zc->zc_cookie & 1;
3285 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3286 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3287 strchr(zc->zc_value, '%'))
3291 * Unmount snapshot unless we're doing a recursive rename,
3292 * in which case the dataset code figures out which snapshots
3295 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3296 zc->zc_objset_type == DMU_OST_ZFS) {
3297 err = zfs_unmount_snap(zc->zc_name, NULL);
3302 err = dmu_objset_rename(zc->zc_name, zc->zc_value, recursive);
3303 if ((err == 0) && (zc->zc_objset_type == DMU_OST_ZVOL)) {
3304 (void) zvol_remove_minor(zc->zc_name);
3305 (void) zvol_create_minor(zc->zc_value);
3312 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3314 const char *propname = nvpair_name(pair);
3315 boolean_t issnap = (strchr(dsname, '@') != NULL);
3316 zfs_prop_t prop = zfs_name_to_prop(propname);
3320 if (prop == ZPROP_INVAL) {
3321 if (zfs_prop_user(propname)) {
3322 if ((err = zfs_secpolicy_write_perms(dsname,
3323 ZFS_DELEG_PERM_USERPROP, cr)))
3328 if (!issnap && zfs_prop_userquota(propname)) {
3329 const char *perm = NULL;
3330 const char *uq_prefix =
3331 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3332 const char *gq_prefix =
3333 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3335 if (strncmp(propname, uq_prefix,
3336 strlen(uq_prefix)) == 0) {
3337 perm = ZFS_DELEG_PERM_USERQUOTA;
3338 } else if (strncmp(propname, gq_prefix,
3339 strlen(gq_prefix)) == 0) {
3340 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3342 /* USERUSED and GROUPUSED are read-only */
3346 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3357 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3359 * dsl_prop_get_all_impl() returns properties in this
3363 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3364 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3369 * Check that this value is valid for this pool version
3372 case ZFS_PROP_COMPRESSION:
3374 * If the user specified gzip compression, make sure
3375 * the SPA supports it. We ignore any errors here since
3376 * we'll catch them later.
3378 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3379 nvpair_value_uint64(pair, &intval) == 0) {
3380 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3381 intval <= ZIO_COMPRESS_GZIP_9 &&
3382 zfs_earlier_version(dsname,
3383 SPA_VERSION_GZIP_COMPRESSION)) {
3387 if (intval == ZIO_COMPRESS_ZLE &&
3388 zfs_earlier_version(dsname,
3389 SPA_VERSION_ZLE_COMPRESSION))
3393 * If this is a bootable dataset then
3394 * verify that the compression algorithm
3395 * is supported for booting. We must return
3396 * something other than ENOTSUP since it
3397 * implies a downrev pool version.
3399 if (zfs_is_bootfs(dsname) &&
3400 !BOOTFS_COMPRESS_VALID(intval)) {
3406 case ZFS_PROP_COPIES:
3407 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3411 case ZFS_PROP_DEDUP:
3412 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3416 case ZFS_PROP_SHARESMB:
3417 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3421 case ZFS_PROP_ACLINHERIT:
3422 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3423 nvpair_value_uint64(pair, &intval) == 0) {
3424 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3425 zfs_earlier_version(dsname,
3426 SPA_VERSION_PASSTHROUGH_X))
3434 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3438 * Removes properties from the given props list that fail permission checks
3439 * needed to clear them and to restore them in case of a receive error. For each
3440 * property, make sure we have both set and inherit permissions.
3442 * Returns the first error encountered if any permission checks fail. If the
3443 * caller provides a non-NULL errlist, it also gives the complete list of names
3444 * of all the properties that failed a permission check along with the
3445 * corresponding error numbers. The caller is responsible for freeing the
3448 * If every property checks out successfully, zero is returned and the list
3449 * pointed at by errlist is NULL.
3452 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3455 nvpair_t *pair, *next_pair;
3462 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3464 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3465 (void) strcpy(zc->zc_name, dataset);
3466 pair = nvlist_next_nvpair(props, NULL);
3467 while (pair != NULL) {
3468 next_pair = nvlist_next_nvpair(props, pair);
3470 (void) strcpy(zc->zc_value, nvpair_name(pair));
3471 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3472 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3473 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3474 VERIFY(nvlist_add_int32(errors,
3475 zc->zc_value, err) == 0);
3479 kmem_free(zc, sizeof (zfs_cmd_t));
3481 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3482 nvlist_free(errors);
3485 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3488 if (errlist == NULL)
3489 nvlist_free(errors);
3497 propval_equals(nvpair_t *p1, nvpair_t *p2)
3499 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3500 /* dsl_prop_get_all_impl() format */
3502 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3503 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3507 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3509 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3510 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3514 if (nvpair_type(p1) != nvpair_type(p2))
3517 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3518 char *valstr1, *valstr2;
3520 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3521 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3522 return (strcmp(valstr1, valstr2) == 0);
3524 uint64_t intval1, intval2;
3526 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3527 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3528 return (intval1 == intval2);
3533 * Remove properties from props if they are not going to change (as determined
3534 * by comparison with origprops). Remove them from origprops as well, since we
3535 * do not need to clear or restore properties that won't change.
3538 props_reduce(nvlist_t *props, nvlist_t *origprops)
3540 nvpair_t *pair, *next_pair;
3542 if (origprops == NULL)
3543 return; /* all props need to be received */
3545 pair = nvlist_next_nvpair(props, NULL);
3546 while (pair != NULL) {
3547 const char *propname = nvpair_name(pair);
3550 next_pair = nvlist_next_nvpair(props, pair);
3552 if ((nvlist_lookup_nvpair(origprops, propname,
3553 &match) != 0) || !propval_equals(pair, match))
3554 goto next; /* need to set received value */
3556 /* don't clear the existing received value */
3557 (void) nvlist_remove_nvpair(origprops, match);
3558 /* don't bother receiving the property */
3559 (void) nvlist_remove_nvpair(props, pair);
3566 static boolean_t zfs_ioc_recv_inject_err;
3571 * zc_name name of containing filesystem
3572 * zc_nvlist_src{_size} nvlist of properties to apply
3573 * zc_value name of snapshot to create
3574 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3575 * zc_cookie file descriptor to recv from
3576 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3577 * zc_guid force flag
3578 * zc_cleanup_fd cleanup-on-exit file descriptor
3579 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3582 * zc_cookie number of bytes read
3583 * zc_nvlist_dst{_size} error for each unapplied received property
3584 * zc_obj zprop_errflags_t
3585 * zc_action_handle handle for this guid/ds mapping
3588 zfs_ioc_recv(zfs_cmd_t *zc)
3592 dmu_recv_cookie_t drc;
3593 boolean_t force = (boolean_t)zc->zc_guid;
3596 int props_error = 0;
3599 nvlist_t *props = NULL; /* sent properties */
3600 nvlist_t *origprops = NULL; /* existing properties */
3601 objset_t *origin = NULL;
3603 char tofs[ZFS_MAXNAMELEN];
3604 boolean_t first_recvd_props = B_FALSE;
3606 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3607 strchr(zc->zc_value, '@') == NULL ||
3608 strchr(zc->zc_value, '%'))
3611 (void) strcpy(tofs, zc->zc_value);
3612 tosnap = strchr(tofs, '@');
3615 if (zc->zc_nvlist_src != 0 &&
3616 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3617 zc->zc_iflags, &props)) != 0)
3627 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3629 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3630 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3631 !dsl_prop_get_hasrecvd(os)) {
3632 first_recvd_props = B_TRUE;
3636 * If new received properties are supplied, they are to
3637 * completely replace the existing received properties, so stash
3638 * away the existing ones.
3640 if (dsl_prop_get_received(os, &origprops) == 0) {
3641 nvlist_t *errlist = NULL;
3643 * Don't bother writing a property if its value won't
3644 * change (and avoid the unnecessary security checks).
3646 * The first receive after SPA_VERSION_RECVD_PROPS is a
3647 * special case where we blow away all local properties
3650 if (!first_recvd_props)
3651 props_reduce(props, origprops);
3652 if (zfs_check_clearable(tofs, origprops,
3654 (void) nvlist_merge(errors, errlist, 0);
3655 nvlist_free(errlist);
3658 dmu_objset_rele(os, FTAG);
3661 if (zc->zc_string[0]) {
3662 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3667 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3668 &zc->zc_begin_record, force, origin, &drc);
3670 dmu_objset_rele(origin, FTAG);
3675 * Set properties before we receive the stream so that they are applied
3676 * to the new data. Note that we must call dmu_recv_stream() if
3677 * dmu_recv_begin() succeeds.
3682 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3683 if (drc.drc_newfs) {
3684 if (spa_version(os->os_spa) >=
3685 SPA_VERSION_RECVD_PROPS)
3686 first_recvd_props = B_TRUE;
3687 } else if (origprops != NULL) {
3688 if (clear_received_props(os, tofs, origprops,
3689 first_recvd_props ? NULL : props) != 0)
3690 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3692 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3694 dsl_prop_set_hasrecvd(os);
3695 } else if (!drc.drc_newfs) {
3696 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3699 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3701 (void) nvlist_merge(errors, errlist, 0);
3702 nvlist_free(errlist);
3705 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3707 * Caller made zc->zc_nvlist_dst less than the minimum expected
3708 * size or supplied an invalid address.
3710 props_error = EINVAL;
3714 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3715 &zc->zc_action_handle);
3718 zfs_sb_t *zsb = NULL;
3720 if (get_zfs_sb(tofs, &zsb) == 0) {
3724 error = zfs_suspend_fs(zsb);
3726 * If the suspend fails, then the recv_end will
3727 * likely also fail, and clean up after itself.
3729 end_err = dmu_recv_end(&drc);
3731 error = zfs_resume_fs(zsb, tofs);
3732 error = error ? error : end_err;
3733 deactivate_super(zsb->z_sb);
3735 error = dmu_recv_end(&drc);
3739 zc->zc_cookie = off - fp->f_offset;
3740 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3744 if (zfs_ioc_recv_inject_err) {
3745 zfs_ioc_recv_inject_err = B_FALSE;
3750 * On error, restore the original props.
3752 if (error && props) {
3753 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3754 if (clear_received_props(os, tofs, props, NULL) != 0) {
3756 * We failed to clear the received properties.
3757 * Since we may have left a $recvd value on the
3758 * system, we can't clear the $hasrecvd flag.
3760 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3761 } else if (first_recvd_props) {
3762 dsl_prop_unset_hasrecvd(os);
3764 dmu_objset_rele(os, FTAG);
3765 } else if (!drc.drc_newfs) {
3766 /* We failed to clear the received properties. */
3767 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3770 if (origprops == NULL && !drc.drc_newfs) {
3771 /* We failed to stash the original properties. */
3772 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3776 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3777 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3778 * explictly if we're restoring local properties cleared in the
3779 * first new-style receive.
3781 if (origprops != NULL &&
3782 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3783 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3784 origprops, NULL) != 0) {
3786 * We stashed the original properties but failed to
3789 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3794 nvlist_free(origprops);
3795 nvlist_free(errors);
3799 error = props_error;
3806 * zc_name name of snapshot to send
3807 * zc_cookie file descriptor to send stream to
3808 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3809 * zc_sendobj objsetid of snapshot to send
3810 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3815 zfs_ioc_send(zfs_cmd_t *zc)
3817 objset_t *fromsnap = NULL;
3823 dsl_dataset_t *dsfrom = NULL;
3827 error = spa_open(zc->zc_name, &spa, FTAG);
3831 dp = spa_get_dsl(spa);
3832 rw_enter(&dp->dp_config_rwlock, RW_READER);
3833 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3834 rw_exit(&dp->dp_config_rwlock);
3836 spa_close(spa, FTAG);
3840 error = dmu_objset_from_ds(ds, &tosnap);
3842 dsl_dataset_rele(ds, FTAG);
3843 spa_close(spa, FTAG);
3847 if (zc->zc_fromobj != 0) {
3848 rw_enter(&dp->dp_config_rwlock, RW_READER);
3849 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3850 rw_exit(&dp->dp_config_rwlock);
3851 spa_close(spa, FTAG);
3853 dsl_dataset_rele(ds, FTAG);
3856 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3858 dsl_dataset_rele(dsfrom, FTAG);
3859 dsl_dataset_rele(ds, FTAG);
3863 spa_close(spa, FTAG);
3866 fp = getf(zc->zc_cookie);
3868 dsl_dataset_rele(ds, FTAG);
3870 dsl_dataset_rele(dsfrom, FTAG);
3875 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3877 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3879 releasef(zc->zc_cookie);
3881 dsl_dataset_rele(dsfrom, FTAG);
3882 dsl_dataset_rele(ds, FTAG);
3887 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3891 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3892 &zc->zc_inject_record);
3895 zc->zc_guid = (uint64_t)id;
3901 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3903 return (zio_clear_fault((int)zc->zc_guid));
3907 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3909 int id = (int)zc->zc_guid;
3912 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3913 &zc->zc_inject_record);
3921 zfs_ioc_error_log(zfs_cmd_t *zc)
3925 size_t count = (size_t)zc->zc_nvlist_dst_size;
3927 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3930 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3933 zc->zc_nvlist_dst_size = count;
3935 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3937 spa_close(spa, FTAG);
3943 zfs_ioc_clear(zfs_cmd_t *zc)
3950 * On zpool clear we also fix up missing slogs
3952 mutex_enter(&spa_namespace_lock);
3953 spa = spa_lookup(zc->zc_name);
3955 mutex_exit(&spa_namespace_lock);
3958 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3959 /* we need to let spa_open/spa_load clear the chains */
3960 spa_set_log_state(spa, SPA_LOG_CLEAR);
3962 spa->spa_last_open_failed = 0;
3963 mutex_exit(&spa_namespace_lock);
3965 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3966 error = spa_open(zc->zc_name, &spa, FTAG);
3969 nvlist_t *config = NULL;
3971 if (zc->zc_nvlist_src == 0)
3974 if ((error = get_nvlist(zc->zc_nvlist_src,
3975 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3976 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3978 if (config != NULL) {
3981 if ((err = put_nvlist(zc, config)) != 0)
3983 nvlist_free(config);
3985 nvlist_free(policy);
3992 spa_vdev_state_enter(spa, SCL_NONE);
3994 if (zc->zc_guid == 0) {
3997 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3999 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4000 spa_close(spa, FTAG);
4005 vdev_clear(spa, vd);
4007 (void) spa_vdev_state_exit(spa, NULL, 0);
4010 * Resume any suspended I/Os.
4012 if (zio_resume(spa) != 0)
4015 spa_close(spa, FTAG);
4022 * zc_name name of filesystem
4023 * zc_value name of origin snapshot
4026 * zc_string name of conflicting snapshot, if there is one
4029 zfs_ioc_promote(zfs_cmd_t *zc)
4034 * We don't need to unmount *all* the origin fs's snapshots, but
4037 cp = strchr(zc->zc_value, '@');
4040 (void) dmu_objset_find(zc->zc_value,
4041 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4042 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4046 * Retrieve a single {user|group}{used|quota}@... property.
4049 * zc_name name of filesystem
4050 * zc_objset_type zfs_userquota_prop_t
4051 * zc_value domain name (eg. "S-1-234-567-89")
4052 * zc_guid RID/UID/GID
4055 * zc_cookie property value
4058 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4063 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4066 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4070 error = zfs_userspace_one(zsb,
4071 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4072 zfs_sb_rele(zsb, FTAG);
4079 * zc_name name of filesystem
4080 * zc_cookie zap cursor
4081 * zc_objset_type zfs_userquota_prop_t
4082 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4085 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4086 * zc_cookie zap cursor
4089 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4092 int bufsize = zc->zc_nvlist_dst_size;
4099 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4103 buf = vmem_alloc(bufsize, KM_SLEEP);
4105 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4106 buf, &zc->zc_nvlist_dst_size);
4109 error = xcopyout(buf,
4110 (void *)(uintptr_t)zc->zc_nvlist_dst,
4111 zc->zc_nvlist_dst_size);
4113 vmem_free(buf, bufsize);
4114 zfs_sb_rele(zsb, FTAG);
4121 * zc_name name of filesystem
4127 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4133 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4134 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4136 * If userused is not enabled, it may be because the
4137 * objset needs to be closed & reopened (to grow the
4138 * objset_phys_t). Suspend/resume the fs will do that.
4140 error = zfs_suspend_fs(zsb);
4142 error = zfs_resume_fs(zsb, zc->zc_name);
4145 error = dmu_objset_userspace_upgrade(zsb->z_os);
4146 deactivate_super(zsb->z_sb);
4148 /* XXX kind of reading contents without owning */
4149 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4153 error = dmu_objset_userspace_upgrade(os);
4154 dmu_objset_rele(os, FTAG);
4161 zfs_ioc_share(zfs_cmd_t *zc)
4166 ace_t full_access[] = {
4167 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4172 * zc_name name of containing filesystem
4173 * zc_obj object # beyond which we want next in-use object #
4176 * zc_obj next in-use object #
4179 zfs_ioc_next_obj(zfs_cmd_t *zc)
4181 objset_t *os = NULL;
4184 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4188 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4189 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4191 dmu_objset_rele(os, FTAG);
4197 * zc_name name of filesystem
4198 * zc_value prefix name for snapshot
4199 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4204 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4209 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4210 (u_longlong_t)ddi_get_lbolt64());
4212 if (strlen(snap_name) >= MAXNAMELEN) {
4217 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4218 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4224 (void) strcpy(zc->zc_value, snap_name);
4231 * zc_name name of "to" snapshot
4232 * zc_value name of "from" snapshot
4233 * zc_cookie file descriptor to write diff data on
4236 * dmu_diff_record_t's to the file descriptor
4239 zfs_ioc_diff(zfs_cmd_t *zc)
4247 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4251 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4253 dmu_objset_rele(tosnap, FTAG);
4257 fp = getf(zc->zc_cookie);
4259 dmu_objset_rele(fromsnap, FTAG);
4260 dmu_objset_rele(tosnap, FTAG);
4266 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4268 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4270 releasef(zc->zc_cookie);
4272 dmu_objset_rele(fromsnap, FTAG);
4273 dmu_objset_rele(tosnap, FTAG);
4278 * Remove all ACL files in shares dir
4280 #ifdef HAVE_SMB_SHARE
4282 zfs_smb_acl_purge(znode_t *dzp)
4285 zap_attribute_t zap;
4286 zfs_sb_t *zsb = ZTOZSB(dzp);
4289 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4290 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4291 zap_cursor_advance(&zc)) {
4292 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4296 zap_cursor_fini(&zc);
4299 #endif /* HAVE_SMB_SHARE */
4302 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4304 #ifdef HAVE_SMB_SHARE
4307 vnode_t *resourcevp = NULL;
4316 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4317 NO_FOLLOW, NULL, &vp)) != 0)
4320 /* Now make sure mntpnt and dataset are ZFS */
4322 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4323 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4324 zc->zc_name) != 0)) {
4334 * Create share dir if its missing.
4336 mutex_enter(&zsb->z_lock);
4337 if (zsb->z_shares_dir == 0) {
4340 tx = dmu_tx_create(zsb->z_os);
4341 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4343 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4344 error = dmu_tx_assign(tx, TXG_WAIT);
4348 error = zfs_create_share_dir(zsb, tx);
4352 mutex_exit(&zsb->z_lock);
4358 mutex_exit(&zsb->z_lock);
4360 ASSERT(zsb->z_shares_dir);
4361 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4367 switch (zc->zc_cookie) {
4368 case ZFS_SMB_ACL_ADD:
4369 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4370 vattr.va_mode = S_IFREG|0777;
4374 vsec.vsa_mask = VSA_ACE;
4375 vsec.vsa_aclentp = &full_access;
4376 vsec.vsa_aclentsz = sizeof (full_access);
4377 vsec.vsa_aclcnt = 1;
4379 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4380 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4382 VN_RELE(resourcevp);
4385 case ZFS_SMB_ACL_REMOVE:
4386 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4390 case ZFS_SMB_ACL_RENAME:
4391 if ((error = get_nvlist(zc->zc_nvlist_src,
4392 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4397 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4398 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4401 VN_RELE(ZTOV(sharedir));
4403 nvlist_free(nvlist);
4406 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4408 nvlist_free(nvlist);
4411 case ZFS_SMB_ACL_PURGE:
4412 error = zfs_smb_acl_purge(sharedir);
4421 VN_RELE(ZTOV(sharedir));
4428 #endif /* HAVE_SMB_SHARE */
4433 * zc_name name of filesystem
4434 * zc_value short name of snap
4435 * zc_string user-supplied tag for this hold
4436 * zc_cookie recursive flag
4437 * zc_temphold set if hold is temporary
4438 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4439 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4440 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4445 zfs_ioc_hold(zfs_cmd_t *zc)
4447 boolean_t recursive = zc->zc_cookie;
4454 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4457 if (zc->zc_sendobj == 0) {
4458 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4459 zc->zc_string, recursive, zc->zc_temphold,
4460 zc->zc_cleanup_fd));
4466 error = spa_open(zc->zc_name, &spa, FTAG);
4470 dp = spa_get_dsl(spa);
4471 rw_enter(&dp->dp_config_rwlock, RW_READER);
4472 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4473 rw_exit(&dp->dp_config_rwlock);
4474 spa_close(spa, FTAG);
4479 * Until we have a hold on this snapshot, it's possible that
4480 * zc_sendobj could've been destroyed and reused as part
4481 * of a later txg. Make sure we're looking at the right object.
4483 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4484 dsl_dataset_rele(ds, FTAG);
4488 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4489 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4491 dsl_dataset_rele(ds, FTAG);
4496 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4500 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4503 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4505 dsl_dataset_rele(ds, FTAG);
4512 * zc_name name of dataset from which we're releasing a user hold
4513 * zc_value short name of snap
4514 * zc_string user-supplied tag for this hold
4515 * zc_cookie recursive flag
4520 zfs_ioc_release(zfs_cmd_t *zc)
4522 boolean_t recursive = zc->zc_cookie;
4524 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4527 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4528 zc->zc_string, recursive));
4533 * zc_name name of filesystem
4536 * zc_nvlist_src{_size} nvlist of snapshot holds
4539 zfs_ioc_get_holds(zfs_cmd_t *zc)
4544 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4545 error = put_nvlist(zc, nvp);
4554 * zc_guid flags (ZEVENT_NONBLOCK)
4557 * zc_nvlist_dst next nvlist event
4558 * zc_cookie dropped events since last get
4559 * zc_cleanup_fd cleanup-on-exit file descriptor
4562 zfs_ioc_events_next(zfs_cmd_t *zc)
4565 nvlist_t *event = NULL;
4567 uint64_t dropped = 0;
4570 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4575 error = zfs_zevent_next(ze, &event,
4576 &zc->zc_nvlist_dst_size, &dropped);
4577 if (event != NULL) {
4578 zc->zc_cookie = dropped;
4579 error = put_nvlist(zc, event);
4583 if (zc->zc_guid & ZEVENT_NONBLOCK)
4586 if ((error == 0) || (error != ENOENT))
4589 error = zfs_zevent_wait(ze);
4594 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4601 * zc_cookie cleared events count
4604 zfs_ioc_events_clear(zfs_cmd_t *zc)
4608 zfs_zevent_drain_all(&count);
4609 zc->zc_cookie = count;
4615 * pool create, destroy, and export don't log the history as part of
4616 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4617 * do the logging of those commands.
4619 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4620 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4622 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4624 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4626 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4628 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4630 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4632 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4634 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4635 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4636 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4637 POOL_CHECK_READONLY },
4638 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4639 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4640 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4642 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4643 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4644 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4645 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4646 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4647 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4648 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4649 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4650 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4651 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4652 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4653 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4654 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4655 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4656 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4657 POOL_CHECK_SUSPENDED },
4658 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4660 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4661 POOL_CHECK_SUSPENDED },
4662 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4663 POOL_CHECK_SUSPENDED },
4664 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4665 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4666 { zfs_ioc_create_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4668 { zfs_ioc_remove_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4670 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4671 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4672 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4673 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4674 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4675 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4676 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4677 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4678 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4679 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4680 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE,
4682 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4684 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4686 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4688 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4690 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4692 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4693 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4694 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4695 B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4696 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4697 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4698 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4700 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4701 POOL_CHECK_SUSPENDED },
4702 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4703 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4704 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4706 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4707 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4708 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4710 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4712 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4713 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4714 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4716 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4717 B_FALSE, POOL_CHECK_NONE },
4718 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
4719 B_FALSE, POOL_CHECK_NONE },
4720 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4721 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4722 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
4723 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4724 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4725 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4726 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4727 POOL_CHECK_SUSPENDED },
4728 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4730 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4731 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4732 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4734 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4736 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
4737 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4738 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4739 POOL_CHECK_SUSPENDED },
4740 { zfs_ioc_events_next, zfs_secpolicy_config, NO_NAME, B_FALSE,
4742 { zfs_ioc_events_clear, zfs_secpolicy_config, NO_NAME, B_FALSE,
4747 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
4748 zfs_ioc_poolcheck_t check)
4753 ASSERT(type == POOL_NAME || type == DATASET_NAME);
4755 if (check & POOL_CHECK_NONE)
4758 error = spa_open(name, &spa, FTAG);
4760 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
4762 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
4764 spa_close(spa, FTAG);
4770 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
4774 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4776 for (zs = list_head(&zfsdev_state_list); zs != NULL;
4777 zs = list_next(&zfsdev_state_list, zs)) {
4778 if (zs->zs_minor == minor) {
4780 case ZST_ONEXIT: return (zs->zs_onexit);
4781 case ZST_ZEVENT: return (zs->zs_zevent);
4782 case ZST_ALL: return (zs);
4791 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
4795 mutex_enter(&zfsdev_state_lock);
4796 ptr = zfsdev_get_state_impl(minor, which);
4797 mutex_exit(&zfsdev_state_lock);
4803 zfsdev_getminor(struct file *filp)
4805 ASSERT(filp != NULL);
4806 ASSERT(filp->private_data != NULL);
4808 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
4812 * Find a free minor number. The zfsdev_state_list is expected to
4813 * be short since it is only a list of currently open file handles.
4816 zfsdev_minor_alloc(void)
4818 static minor_t last_minor = 0;
4821 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4823 for (m = last_minor + 1; m != last_minor; m++) {
4824 if (m > ZFSDEV_MAX_MINOR)
4826 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
4836 zfsdev_state_init(struct file *filp)
4841 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4843 minor = zfsdev_minor_alloc();
4847 zs = kmem_zalloc( sizeof(zfsdev_state_t), KM_SLEEP);
4852 zs->zs_minor = minor;
4853 filp->private_data = zs;
4855 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
4856 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
4858 list_insert_tail(&zfsdev_state_list, zs);
4864 zfsdev_state_destroy(struct file *filp)
4868 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4869 ASSERT(filp->private_data != NULL);
4871 zs = filp->private_data;
4872 zfs_onexit_destroy(zs->zs_onexit);
4873 zfs_zevent_destroy(zs->zs_zevent);
4875 list_remove(&zfsdev_state_list, zs);
4876 kmem_free(zs, sizeof(zfsdev_state_t));
4882 zfsdev_open(struct inode *ino, struct file *filp)
4886 mutex_enter(&zfsdev_state_lock);
4887 error = zfsdev_state_init(filp);
4888 mutex_exit(&zfsdev_state_lock);
4894 zfsdev_release(struct inode *ino, struct file *filp)
4898 mutex_enter(&zfsdev_state_lock);
4899 error = zfsdev_state_destroy(filp);
4900 mutex_exit(&zfsdev_state_lock);
4906 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4910 int error, rc, flag = 0;
4912 vec = cmd - ZFS_IOC;
4913 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4916 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
4918 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4922 if ((error == 0) && !(flag & FKIOCTL))
4923 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, CRED());
4926 * Ensure that all pool/dataset names are valid before we pass down to
4930 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4931 zc->zc_iflags = flag & FKIOCTL;
4932 switch (zfs_ioc_vec[vec].zvec_namecheck) {
4934 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4936 error = pool_status_check(zc->zc_name,
4937 zfs_ioc_vec[vec].zvec_namecheck,
4938 zfs_ioc_vec[vec].zvec_pool_check);
4942 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4944 error = pool_status_check(zc->zc_name,
4945 zfs_ioc_vec[vec].zvec_namecheck,
4946 zfs_ioc_vec[vec].zvec_pool_check);
4955 error = zfs_ioc_vec[vec].zvec_func(zc);
4957 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4961 if (zfs_ioc_vec[vec].zvec_his_log)
4962 zfs_log_history(zc);
4965 kmem_free(zc, sizeof (zfs_cmd_t));
4969 #ifdef CONFIG_COMPAT
4971 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4973 return zfsdev_ioctl(filp, cmd, arg);
4976 #define zfsdev_compat_ioctl NULL
4979 static const struct file_operations zfsdev_fops = {
4980 .open = zfsdev_open,
4981 .release = zfsdev_release,
4982 .unlocked_ioctl = zfsdev_ioctl,
4983 .compat_ioctl = zfsdev_compat_ioctl,
4984 .owner = THIS_MODULE,
4987 static struct miscdevice zfs_misc = {
4988 .minor = MISC_DYNAMIC_MINOR,
4990 .fops = &zfsdev_fops,
4998 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
4999 list_create(&zfsdev_state_list, sizeof (zfsdev_state_t),
5000 offsetof(zfsdev_state_t, zs_next));
5002 error = misc_register(&zfs_misc);
5004 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
5016 error = misc_deregister(&zfs_misc);
5018 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5020 mutex_destroy(&zfsdev_state_lock);
5021 list_destroy(&zfsdev_state_list);
5024 uint_t zfs_fsyncer_key;
5025 extern uint_t rrw_tsd_key;
5028 #define ZFS_DEBUG_STR " (DEBUG mode)"
5030 #define ZFS_DEBUG_STR ""
5038 spa_init(FREAD | FWRITE);
5041 if ((error = zvol_init()) != 0)
5044 if ((error = zfs_attach()) != 0)
5047 tsd_create(&zfs_fsyncer_key, NULL);
5048 tsd_create(&rrw_tsd_key, NULL);
5050 printk(KERN_NOTICE "ZFS: Loaded module v%s%s, "
5051 "ZFS pool version %s, ZFS filesystem version %s\n",
5052 ZFS_META_VERSION, ZFS_DEBUG_STR,
5053 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5062 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s%s"
5063 ", rc = %d\n", ZFS_META_VERSION, ZFS_DEBUG_STR, error);
5076 tsd_destroy(&zfs_fsyncer_key);
5077 tsd_destroy(&rrw_tsd_key);
5079 printk(KERN_NOTICE "ZFS: Unloaded module v%s%s\n",
5080 ZFS_META_VERSION, ZFS_DEBUG_STR);
5086 spl_module_init(_init);
5087 spl_module_exit(_fini);
5089 MODULE_DESCRIPTION("ZFS");
5090 MODULE_AUTHOR(ZFS_META_AUTHOR);
5091 MODULE_LICENSE(ZFS_META_LICENSE);
5092 #endif /* HAVE_SPL */