Move the world out of /zfs/ and seperate out module build tree
[zfs.git] / module / zfs / dsl_deleg.c
diff --git a/module/zfs/dsl_deleg.c b/module/zfs/dsl_deleg.c
new file mode 100644 (file)
index 0000000..da5d157
--- /dev/null
@@ -0,0 +1,735 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * DSL permissions are stored in a two level zap attribute
+ * mechanism.   The first level identifies the "class" of
+ * entry.  The class is identified by the first 2 letters of
+ * the attribute.  The second letter "l" or "d" identifies whether
+ * it is a local or descendent permission.  The first letter
+ * identifies the type of entry.
+ *
+ * ul$<id>    identifies permissions granted locally for this userid.
+ * ud$<id>    identifies permissions granted on descendent datasets for
+ *            this userid.
+ * Ul$<id>    identifies permission sets granted locally for this userid.
+ * Ud$<id>    identifies permission sets granted on descendent datasets for
+ *            this userid.
+ * gl$<id>    identifies permissions granted locally for this groupid.
+ * gd$<id>    identifies permissions granted on descendent datasets for
+ *            this groupid.
+ * Gl$<id>    identifies permission sets granted locally for this groupid.
+ * Gd$<id>    identifies permission sets granted on descendent datasets for
+ *            this groupid.
+ * el$        identifies permissions granted locally for everyone.
+ * ed$        identifies permissions granted on descendent datasets
+ *            for everyone.
+ * El$        identifies permission sets granted locally for everyone.
+ * Ed$        identifies permission sets granted to descendent datasets for
+ *            everyone.
+ * c-$        identifies permission to create at dataset creation time.
+ * C-$        identifies permission sets to grant locally at dataset creation
+ *            time.
+ * s-$@<name> permissions defined in specified set @<name>
+ * S-$@<name> Sets defined in named set @<name>
+ *
+ * Each of the above entities points to another zap attribute that contains one
+ * attribute for each allowed permission, such as create, destroy,...
+ * All of the "upper" case class types will specify permission set names
+ * rather than permissions.
+ *
+ * Basically it looks something like this:
+ * ul$12 -> ZAP OBJ -> permissions...
+ *
+ * The ZAP OBJ is referred to as the jump object.
+ */
+
+#pragma ident  "%Z%%M% %I%     %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_tx.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dsl_deleg.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/zio_checksum.h> /* for the default checksum value */
+#include <sys/zap.h>
+#include <sys/fs/zfs.h>
+#include <sys/cred.h>
+#include <sys/sunddi.h>
+
+#include "zfs_deleg.h"
+
+/*
+ * Validate that user is allowed to delegate specified permissions.
+ *
+ * In order to delegate "create" you must have "create"
+ * and "allow".
+ */
+int
+dsl_deleg_can_allow(char *ddname, nvlist_t *nvp, cred_t *cr)
+{
+       nvpair_t *whopair = NULL;
+       int error;
+
+       if ((error = dsl_deleg_access(ddname, ZFS_DELEG_PERM_ALLOW, cr)) != 0)
+               return (error);
+
+       while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+               nvlist_t *perms;
+               nvpair_t *permpair = NULL;
+
+               VERIFY(nvpair_value_nvlist(whopair, &perms) == 0);
+
+               while (permpair = nvlist_next_nvpair(perms, permpair)) {
+                       const char *perm = nvpair_name(permpair);
+
+                       if (strcmp(perm, ZFS_DELEG_PERM_ALLOW) == 0)
+                               return (EPERM);
+
+                       if ((error = dsl_deleg_access(ddname, perm, cr)) != 0)
+                               return (error);
+               }
+       }
+       return (0);
+}
+
+/*
+ * Validate that user is allowed to unallow specified permissions.  They
+ * must have the 'allow' permission, and even then can only unallow
+ * perms for their uid.
+ */
+int
+dsl_deleg_can_unallow(char *ddname, nvlist_t *nvp, cred_t *cr)
+{
+       nvpair_t *whopair = NULL;
+       int error;
+       char idstr[32];
+
+       if ((error = dsl_deleg_access(ddname, ZFS_DELEG_PERM_ALLOW, cr)) != 0)
+               return (error);
+
+       (void) snprintf(idstr, sizeof (idstr), "%lld",
+           (longlong_t)crgetuid(cr));
+
+       while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+               zfs_deleg_who_type_t type = nvpair_name(whopair)[0];
+
+               if (type != ZFS_DELEG_USER &&
+                   type != ZFS_DELEG_USER_SETS)
+                       return (EPERM);
+
+               if (strcmp(idstr, &nvpair_name(whopair)[3]) != 0)
+                       return (EPERM);
+       }
+       return (0);
+}
+
+static void
+dsl_deleg_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+       dsl_dir_t *dd = arg1;
+       nvlist_t *nvp = arg2;
+       objset_t *mos = dd->dd_pool->dp_meta_objset;
+       nvpair_t *whopair = NULL;
+       uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+       if (zapobj == 0) {
+               dmu_buf_will_dirty(dd->dd_dbuf, tx);
+               zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
+                   DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+       }
+
+       while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+               const char *whokey = nvpair_name(whopair);
+               nvlist_t *perms;
+               nvpair_t *permpair = NULL;
+               uint64_t jumpobj;
+
+               VERIFY(nvpair_value_nvlist(whopair, &perms) == 0);
+
+               if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0) {
+                       jumpobj = zap_create(mos, DMU_OT_DSL_PERMS,
+                           DMU_OT_NONE, 0, tx);
+                       VERIFY(zap_update(mos, zapobj,
+                           whokey, 8, 1, &jumpobj, tx) == 0);
+               }
+
+               while (permpair = nvlist_next_nvpair(perms, permpair)) {
+                       const char *perm = nvpair_name(permpair);
+                       uint64_t n = 0;
+
+                       VERIFY(zap_update(mos, jumpobj,
+                           perm, 8, 1, &n, tx) == 0);
+                       spa_history_internal_log(LOG_DS_PERM_UPDATE,
+                           dd->dd_pool->dp_spa, tx, cr,
+                           "%s %s dataset = %llu", whokey, perm,
+                           dd->dd_phys->dd_head_dataset_obj);
+               }
+       }
+}
+
+static void
+dsl_deleg_unset_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+       dsl_dir_t *dd = arg1;
+       nvlist_t *nvp = arg2;
+       objset_t *mos = dd->dd_pool->dp_meta_objset;
+       nvpair_t *whopair = NULL;
+       uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+       if (zapobj == 0)
+               return;
+
+       while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+               const char *whokey = nvpair_name(whopair);
+               nvlist_t *perms;
+               nvpair_t *permpair = NULL;
+               uint64_t jumpobj;
+
+               if (nvpair_value_nvlist(whopair, &perms) != 0) {
+                       if (zap_lookup(mos, zapobj, whokey, 8,
+                           1, &jumpobj) == 0) {
+                               (void) zap_remove(mos, zapobj, whokey, tx);
+                               VERIFY(0 == zap_destroy(mos, jumpobj, tx));
+                       }
+                       spa_history_internal_log(LOG_DS_PERM_WHO_REMOVE,
+                           dd->dd_pool->dp_spa, tx, cr,
+                           "%s dataset = %llu", whokey,
+                           dd->dd_phys->dd_head_dataset_obj);
+                       continue;
+               }
+
+               if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0)
+                       continue;
+
+               while (permpair = nvlist_next_nvpair(perms, permpair)) {
+                       const char *perm = nvpair_name(permpair);
+                       uint64_t n = 0;
+
+                       (void) zap_remove(mos, jumpobj, perm, tx);
+                       if (zap_count(mos, jumpobj, &n) == 0 && n == 0) {
+                               (void) zap_remove(mos, zapobj,
+                                   whokey, tx);
+                               VERIFY(0 == zap_destroy(mos,
+                                   jumpobj, tx));
+                       }
+                       spa_history_internal_log(LOG_DS_PERM_REMOVE,
+                           dd->dd_pool->dp_spa, tx, cr,
+                           "%s %s dataset = %llu", whokey, perm,
+                           dd->dd_phys->dd_head_dataset_obj);
+               }
+       }
+}
+
+int
+dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset)
+{
+       dsl_dir_t *dd;
+       int error;
+       nvpair_t *whopair = NULL;
+       int blocks_modified = 0;
+
+       error = dsl_dir_open(ddname, FTAG, &dd, NULL);
+       if (error)
+               return (error);
+
+       if (spa_version(dmu_objset_spa(dd->dd_pool->dp_meta_objset)) <
+           SPA_VERSION_DELEGATED_PERMS) {
+               dsl_dir_close(dd, FTAG);
+               return (ENOTSUP);
+       }
+
+       while (whopair = nvlist_next_nvpair(nvp, whopair))
+               blocks_modified++;
+
+       error = dsl_sync_task_do(dd->dd_pool, NULL,
+           unset ? dsl_deleg_unset_sync : dsl_deleg_set_sync,
+           dd, nvp, blocks_modified);
+       dsl_dir_close(dd, FTAG);
+
+       return (error);
+}
+
+/*
+ * Find all 'allow' permissions from a given point and then continue
+ * traversing up to the root.
+ *
+ * This function constructs an nvlist of nvlists.
+ * each setpoint is an nvlist composed of an nvlist of an nvlist
+ * of the individual * users/groups/everyone/create
+ * permissions.
+ *
+ * The nvlist will look like this.
+ *
+ * { source fsname -> { whokeys { permissions,...}, ...}}
+ *
+ * The fsname nvpairs will be arranged in a bottom up order.  For example,
+ * if we have the following structure a/b/c then the nvpairs for the fsnames
+ * will be ordered a/b/c, a/b, a.
+ */
+int
+dsl_deleg_get(const char *ddname, nvlist_t **nvp)
+{
+       dsl_dir_t *dd, *startdd;
+       dsl_pool_t *dp;
+       int error;
+       objset_t *mos;
+
+       error = dsl_dir_open(ddname, FTAG, &startdd, NULL);
+       if (error)
+               return (error);
+
+       dp = startdd->dd_pool;
+       mos = dp->dp_meta_objset;
+
+       VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+       rw_enter(&dp->dp_config_rwlock, RW_READER);
+       for (dd = startdd; dd != NULL; dd = dd->dd_parent) {
+               zap_cursor_t basezc;
+               zap_attribute_t baseza;
+               nvlist_t *sp_nvp;
+               uint64_t n;
+               char source[MAXNAMELEN];
+
+               if (dd->dd_phys->dd_deleg_zapobj &&
+                   (zap_count(mos, dd->dd_phys->dd_deleg_zapobj,
+                   &n) == 0) && n) {
+                       VERIFY(nvlist_alloc(&sp_nvp,
+                           NV_UNIQUE_NAME, KM_SLEEP) == 0);
+               } else {
+                       continue;
+               }
+
+               for (zap_cursor_init(&basezc, mos,
+                   dd->dd_phys->dd_deleg_zapobj);
+                   zap_cursor_retrieve(&basezc, &baseza) == 0;
+                   zap_cursor_advance(&basezc)) {
+                       zap_cursor_t zc;
+                       zap_attribute_t za;
+                       nvlist_t *perms_nvp;
+
+                       ASSERT(baseza.za_integer_length == 8);
+                       ASSERT(baseza.za_num_integers == 1);
+
+                       VERIFY(nvlist_alloc(&perms_nvp,
+                           NV_UNIQUE_NAME, KM_SLEEP) == 0);
+                       for (zap_cursor_init(&zc, mos, baseza.za_first_integer);
+                           zap_cursor_retrieve(&zc, &za) == 0;
+                           zap_cursor_advance(&zc)) {
+                               VERIFY(nvlist_add_boolean(perms_nvp,
+                                   za.za_name) == 0);
+                       }
+                       zap_cursor_fini(&zc);
+                       VERIFY(nvlist_add_nvlist(sp_nvp, baseza.za_name,
+                           perms_nvp) == 0);
+                       nvlist_free(perms_nvp);
+               }
+
+               zap_cursor_fini(&basezc);
+
+               dsl_dir_name(dd, source);
+               VERIFY(nvlist_add_nvlist(*nvp, source, sp_nvp) == 0);
+               nvlist_free(sp_nvp);
+       }
+       rw_exit(&dp->dp_config_rwlock);
+
+       dsl_dir_close(startdd, FTAG);
+       return (0);
+}
+
+/*
+ * Routines for dsl_deleg_access() -- access checking.
+ */
+typedef struct perm_set {
+       avl_node_t      p_node;
+       boolean_t       p_matched;
+       char            p_setname[ZFS_MAX_DELEG_NAME];
+} perm_set_t;
+
+static int
+perm_set_compare(const void *arg1, const void *arg2)
+{
+       const perm_set_t *node1 = arg1;
+       const perm_set_t *node2 = arg2;
+       int val;
+
+       val = strcmp(node1->p_setname, node2->p_setname);
+       if (val == 0)
+               return (0);
+       return (val > 0 ? 1 : -1);
+}
+
+/*
+ * Determine whether a specified permission exists.
+ *
+ * First the base attribute has to be retrieved.  i.e. ul$12
+ * Once the base object has been retrieved the actual permission
+ * is lookup up in the zap object the base object points to.
+ *
+ * Return 0 if permission exists, ENOENT if there is no whokey, EPERM if
+ * there is no perm in that jumpobj.
+ */
+static int
+dsl_check_access(objset_t *mos, uint64_t zapobj,
+    char type, char checkflag, void *valp, const char *perm)
+{
+       int error;
+       uint64_t jumpobj, zero;
+       char whokey[ZFS_MAX_DELEG_NAME];
+
+       zfs_deleg_whokey(whokey, type, checkflag, valp);
+       error = zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj);
+       if (error == 0) {
+               error = zap_lookup(mos, jumpobj, perm, 8, 1, &zero);
+               if (error == ENOENT)
+                       error = EPERM;
+       }
+       return (error);
+}
+
+/*
+ * check a specified user/group for a requested permission
+ */
+static int
+dsl_check_user_access(objset_t *mos, uint64_t zapobj, const char *perm,
+    int checkflag, cred_t *cr)
+{
+       const   gid_t *gids;
+       int     ngids;
+       int     i;
+       uint64_t id;
+
+       /* check for user */
+       id = crgetuid(cr);
+       if (dsl_check_access(mos, zapobj,
+           ZFS_DELEG_USER, checkflag, &id, perm) == 0)
+               return (0);
+
+       /* check for users primary group */
+       id = crgetgid(cr);
+       if (dsl_check_access(mos, zapobj,
+           ZFS_DELEG_GROUP, checkflag, &id, perm) == 0)
+               return (0);
+
+       /* check for everyone entry */
+       id = -1;
+       if (dsl_check_access(mos, zapobj,
+           ZFS_DELEG_EVERYONE, checkflag, &id, perm) == 0)
+               return (0);
+
+       /* check each supplemental group user is a member of */
+       ngids = crgetngroups(cr);
+       gids = crgetgroups(cr);
+       for (i = 0; i != ngids; i++) {
+               id = gids[i];
+               if (dsl_check_access(mos, zapobj,
+                   ZFS_DELEG_GROUP, checkflag, &id, perm) == 0)
+                       return (0);
+       }
+
+       return (EPERM);
+}
+
+/*
+ * Iterate over the sets specified in the specified zapobj
+ * and load them into the permsets avl tree.
+ */
+static int
+dsl_load_sets(objset_t *mos, uint64_t zapobj,
+    char type, char checkflag, void *valp, avl_tree_t *avl)
+{
+       zap_cursor_t zc;
+       zap_attribute_t za;
+       perm_set_t *permnode;
+       avl_index_t idx;
+       uint64_t jumpobj;
+       int error;
+       char whokey[ZFS_MAX_DELEG_NAME];
+
+       zfs_deleg_whokey(whokey, type, checkflag, valp);
+
+       error = zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj);
+       if (error != 0)
+               return (error);
+
+       for (zap_cursor_init(&zc, mos, jumpobj);
+           zap_cursor_retrieve(&zc, &za) == 0;
+           zap_cursor_advance(&zc)) {
+               permnode = kmem_alloc(sizeof (perm_set_t), KM_SLEEP);
+               (void) strlcpy(permnode->p_setname, za.za_name,
+                   sizeof (permnode->p_setname));
+               permnode->p_matched = B_FALSE;
+
+               if (avl_find(avl, permnode, &idx) == NULL) {
+                       avl_insert(avl, permnode, idx);
+               } else {
+                       kmem_free(permnode, sizeof (perm_set_t));
+               }
+       }
+       zap_cursor_fini(&zc);
+       return (0);
+}
+
+/*
+ * Load all permissions user based on cred belongs to.
+ */
+static void
+dsl_load_user_sets(objset_t *mos, uint64_t zapobj, avl_tree_t *avl,
+    char checkflag, cred_t *cr)
+{
+       const   gid_t *gids;
+       int     ngids, i;
+       uint64_t id;
+
+       id = crgetuid(cr);
+       (void) dsl_load_sets(mos, zapobj,
+           ZFS_DELEG_USER_SETS, checkflag, &id, avl);
+
+       id = crgetgid(cr);
+       (void) dsl_load_sets(mos, zapobj,
+           ZFS_DELEG_GROUP_SETS, checkflag, &id, avl);
+
+       (void) dsl_load_sets(mos, zapobj,
+           ZFS_DELEG_EVERYONE_SETS, checkflag, NULL, avl);
+
+       ngids = crgetngroups(cr);
+       gids = crgetgroups(cr);
+       for (i = 0; i != ngids; i++) {
+               id = gids[i];
+               (void) dsl_load_sets(mos, zapobj,
+                   ZFS_DELEG_GROUP_SETS, checkflag, &id, avl);
+       }
+}
+
+/*
+ * Check if user has requested permission.
+ */
+int
+dsl_deleg_access(const char *dsname, const char *perm, cred_t *cr)
+{
+       dsl_dataset_t *ds;
+       dsl_dir_t *dd;
+       dsl_pool_t *dp;
+       void *cookie;
+       int     error;
+       char    checkflag = ZFS_DELEG_LOCAL;
+       objset_t *mos;
+       avl_tree_t permsets;
+       perm_set_t *setnode;
+
+       error = dsl_dataset_hold(dsname, FTAG, &ds);
+       if (error)
+               return (error);
+
+       dp = ds->ds_dir->dd_pool;
+       mos = dp->dp_meta_objset;
+
+       if (dsl_delegation_on(mos) == B_FALSE) {
+               dsl_dataset_rele(ds, FTAG);
+               return (ECANCELED);
+       }
+
+       if (spa_version(dmu_objset_spa(dp->dp_meta_objset)) <
+           SPA_VERSION_DELEGATED_PERMS) {
+               dsl_dataset_rele(ds, FTAG);
+               return (EPERM);
+       }
+
+       avl_create(&permsets, perm_set_compare, sizeof (perm_set_t),
+           offsetof(perm_set_t, p_node));
+
+       rw_enter(&dp->dp_config_rwlock, RW_READER);
+       for (dd = ds->ds_dir; dd != NULL; dd = dd->dd_parent,
+           checkflag = ZFS_DELEG_DESCENDENT) {
+               uint64_t zapobj;
+               boolean_t expanded;
+
+               /*
+                * If not in global zone then make sure
+                * the zoned property is set
+                */
+               if (!INGLOBALZONE(curproc)) {
+                       uint64_t zoned;
+
+                       if (dsl_prop_get_dd(dd,
+                           zfs_prop_to_name(ZFS_PROP_ZONED),
+                           8, 1, &zoned, NULL) != 0)
+                               break;
+                       if (!zoned)
+                               break;
+               }
+               zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+               if (zapobj == 0)
+                       continue;
+
+               dsl_load_user_sets(mos, zapobj, &permsets, checkflag, cr);
+again:
+               expanded = B_FALSE;
+               for (setnode = avl_first(&permsets); setnode;
+                   setnode = AVL_NEXT(&permsets, setnode)) {
+                       if (setnode->p_matched == B_TRUE)
+                               continue;
+
+                       /* See if this set directly grants this permission */
+                       error = dsl_check_access(mos, zapobj,
+                           ZFS_DELEG_NAMED_SET, 0, setnode->p_setname, perm);
+                       if (error == 0)
+                               goto success;
+                       if (error == EPERM)
+                               setnode->p_matched = B_TRUE;
+
+                       /* See if this set includes other sets */
+                       error = dsl_load_sets(mos, zapobj,
+                           ZFS_DELEG_NAMED_SET_SETS, 0,
+                           setnode->p_setname, &permsets);
+                       if (error == 0)
+                               setnode->p_matched = expanded = B_TRUE;
+               }
+               /*
+                * If we expanded any sets, that will define more sets,
+                * which we need to check.
+                */
+               if (expanded)
+                       goto again;
+
+               error = dsl_check_user_access(mos, zapobj, perm, checkflag, cr);
+               if (error == 0)
+                       goto success;
+       }
+       error = EPERM;
+success:
+       rw_exit(&dp->dp_config_rwlock);
+       dsl_dataset_rele(ds, FTAG);
+
+       cookie = NULL;
+       while ((setnode = avl_destroy_nodes(&permsets, &cookie)) != NULL)
+               kmem_free(setnode, sizeof (perm_set_t));
+
+       return (error);
+}
+
+/*
+ * Other routines.
+ */
+
+static void
+copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj,
+    boolean_t dosets, uint64_t uid, dmu_tx_t *tx)
+{
+       objset_t *mos = dd->dd_pool->dp_meta_objset;
+       uint64_t jumpobj, pjumpobj;
+       uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+       zap_cursor_t zc;
+       zap_attribute_t za;
+       char whokey[ZFS_MAX_DELEG_NAME];
+
+       zfs_deleg_whokey(whokey,
+           dosets ? ZFS_DELEG_CREATE_SETS : ZFS_DELEG_CREATE,
+           ZFS_DELEG_LOCAL, NULL);
+       if (zap_lookup(mos, pzapobj, whokey, 8, 1, &pjumpobj) != 0)
+               return;
+
+       if (zapobj == 0) {
+               dmu_buf_will_dirty(dd->dd_dbuf, tx);
+               zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
+                   DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+       }
+
+       zfs_deleg_whokey(whokey,
+           dosets ? ZFS_DELEG_USER_SETS : ZFS_DELEG_USER,
+           ZFS_DELEG_LOCAL, &uid);
+       if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == ENOENT) {
+               jumpobj = zap_create(mos, DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+               VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0);
+       }
+
+       for (zap_cursor_init(&zc, mos, pjumpobj);
+           zap_cursor_retrieve(&zc, &za) == 0;
+           zap_cursor_advance(&zc)) {
+               uint64_t zero = 0;
+               ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
+
+               VERIFY(zap_update(mos, jumpobj, za.za_name,
+                   8, 1, &zero, tx) == 0);
+       }
+       zap_cursor_fini(&zc);
+}
+
+/*
+ * set all create time permission on new dataset.
+ */
+void
+dsl_deleg_set_create_perms(dsl_dir_t *sdd, dmu_tx_t *tx, cred_t *cr)
+{
+       dsl_dir_t *dd;
+       uint64_t uid = crgetuid(cr);
+
+       if (spa_version(dmu_objset_spa(sdd->dd_pool->dp_meta_objset)) <
+           SPA_VERSION_DELEGATED_PERMS)
+               return;
+
+       for (dd = sdd->dd_parent; dd != NULL; dd = dd->dd_parent) {
+               uint64_t pzapobj = dd->dd_phys->dd_deleg_zapobj;
+
+               if (pzapobj == 0)
+                       continue;
+
+               copy_create_perms(sdd, pzapobj, B_FALSE, uid, tx);
+               copy_create_perms(sdd, pzapobj, B_TRUE, uid, tx);
+       }
+}
+
+int
+dsl_deleg_destroy(objset_t *mos, uint64_t zapobj, dmu_tx_t *tx)
+{
+       zap_cursor_t zc;
+       zap_attribute_t za;
+
+       if (zapobj == 0)
+               return (0);
+
+       for (zap_cursor_init(&zc, mos, zapobj);
+           zap_cursor_retrieve(&zc, &za) == 0;
+           zap_cursor_advance(&zc)) {
+               ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
+               VERIFY(0 == zap_destroy(mos, za.za_first_integer, tx));
+       }
+       zap_cursor_fini(&zc);
+       VERIFY(0 == zap_destroy(mos, zapobj, tx));
+       return (0);
+}
+
+boolean_t
+dsl_delegation_on(objset_t *os)
+{
+       return (os->os->os_spa->spa_delegation);
+}