Move the world out of /zfs/ and seperate out module build tree
[zfs.git] / module / zfs / zfs_fuid.c
diff --git a/module/zfs/zfs_fuid.c b/module/zfs/zfs_fuid.c
new file mode 100644 (file)
index 0000000..7cb5052
--- /dev/null
@@ -0,0 +1,704 @@
+/*
+ * 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.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/sunddi.h>
+#include <sys/dmu.h>
+#include <sys/avl.h>
+#include <sys/zap.h>
+#include <sys/refcount.h>
+#include <sys/nvpair.h>
+#ifdef _KERNEL
+#include <sys/kidmap.h>
+#include <sys/sid.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/zfs_znode.h>
+#endif
+#include <sys/zfs_fuid.h>
+
+/*
+ * FUID Domain table(s).
+ *
+ * The FUID table is stored as a packed nvlist of an array
+ * of nvlists which contain an index, domain string and offset
+ *
+ * During file system initialization the nvlist(s) are read and
+ * two AVL trees are created.  One tree is keyed by the index number
+ * and the other by the domain string.  Nodes are never removed from
+ * trees, but new entries may be added.  If a new entry is added then the
+ * on-disk packed nvlist will also be updated.
+ */
+
+#define        FUID_IDX        "fuid_idx"
+#define        FUID_DOMAIN     "fuid_domain"
+#define        FUID_OFFSET     "fuid_offset"
+#define        FUID_NVP_ARRAY  "fuid_nvlist"
+
+typedef struct fuid_domain {
+       avl_node_t      f_domnode;
+       avl_node_t      f_idxnode;
+       ksiddomain_t    *f_ksid;
+       uint64_t        f_idx;
+} fuid_domain_t;
+
+static char *nulldomain = "";
+
+/*
+ * Compare two indexes.
+ */
+static int
+idx_compare(const void *arg1, const void *arg2)
+{
+       const fuid_domain_t *node1 = arg1;
+       const fuid_domain_t *node2 = arg2;
+
+       if (node1->f_idx < node2->f_idx)
+               return (-1);
+       else if (node1->f_idx > node2->f_idx)
+               return (1);
+       return (0);
+}
+
+/*
+ * Compare two domain strings.
+ */
+static int
+domain_compare(const void *arg1, const void *arg2)
+{
+       const fuid_domain_t *node1 = arg1;
+       const fuid_domain_t *node2 = arg2;
+       int val;
+
+       val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
+       if (val == 0)
+               return (0);
+       return (val > 0 ? 1 : -1);
+}
+
+/*
+ * load initial fuid domain and idx trees.  This function is used by
+ * both the kernel and zdb.
+ */
+uint64_t
+zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
+    avl_tree_t *domain_tree)
+{
+       dmu_buf_t *db;
+       uint64_t fuid_size;
+
+       avl_create(idx_tree, idx_compare,
+           sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
+       avl_create(domain_tree, domain_compare,
+           sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
+
+       VERIFY(0 == dmu_bonus_hold(os, fuid_obj, FTAG, &db));
+       fuid_size = *(uint64_t *)db->db_data;
+       dmu_buf_rele(db, FTAG);
+
+       if (fuid_size)  {
+               nvlist_t **fuidnvp;
+               nvlist_t *nvp = NULL;
+               uint_t count;
+               char *packed;
+               int i;
+
+               packed = kmem_alloc(fuid_size, KM_SLEEP);
+               VERIFY(dmu_read(os, fuid_obj, 0, fuid_size, packed) == 0);
+               VERIFY(nvlist_unpack(packed, fuid_size,
+                   &nvp, 0) == 0);
+               VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
+                   &fuidnvp, &count) == 0);
+
+               for (i = 0; i != count; i++) {
+                       fuid_domain_t *domnode;
+                       char *domain;
+                       uint64_t idx;
+
+                       VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
+                           &domain) == 0);
+                       VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
+                           &idx) == 0);
+
+                       domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
+
+                       domnode->f_idx = idx;
+                       domnode->f_ksid = ksid_lookupdomain(domain);
+                       avl_add(idx_tree, domnode);
+                       avl_add(domain_tree, domnode);
+               }
+               nvlist_free(nvp);
+               kmem_free(packed, fuid_size);
+       }
+       return (fuid_size);
+}
+
+void
+zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
+{
+       fuid_domain_t *domnode;
+       void *cookie;
+
+       cookie = NULL;
+       while (domnode = avl_destroy_nodes(domain_tree, &cookie))
+               ksiddomain_rele(domnode->f_ksid);
+
+       avl_destroy(domain_tree);
+       cookie = NULL;
+       while (domnode = avl_destroy_nodes(idx_tree, &cookie))
+               kmem_free(domnode, sizeof (fuid_domain_t));
+       avl_destroy(idx_tree);
+}
+
+char *
+zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
+{
+       fuid_domain_t searchnode, *findnode;
+       avl_index_t loc;
+
+       searchnode.f_idx = idx;
+
+       findnode = avl_find(idx_tree, &searchnode, &loc);
+
+       return (findnode ? findnode->f_ksid->kd_name : nulldomain);
+}
+
+#ifdef _KERNEL
+/*
+ * Load the fuid table(s) into memory.
+ */
+static void
+zfs_fuid_init(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
+{
+       int error = 0;
+
+       rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
+
+       if (zfsvfs->z_fuid_loaded) {
+               rw_exit(&zfsvfs->z_fuid_lock);
+               return;
+       }
+
+       if (zfsvfs->z_fuid_obj == 0) {
+
+               /* first make sure we need to allocate object */
+
+               error = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
+                   ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
+               if (error == ENOENT && tx != NULL) {
+                       zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
+                           DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
+                           sizeof (uint64_t), tx);
+                       VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
+                           ZFS_FUID_TABLES, sizeof (uint64_t), 1,
+                           &zfsvfs->z_fuid_obj, tx) == 0);
+               }
+       }
+
+       if (zfsvfs->z_fuid_obj != 0) {
+               zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
+                   zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
+                   &zfsvfs->z_fuid_domain);
+               zfsvfs->z_fuid_loaded = B_TRUE;
+       }
+
+       rw_exit(&zfsvfs->z_fuid_lock);
+}
+
+/*
+ * Query domain table for a given domain.
+ *
+ * If domain isn't found it is added to AVL trees and
+ * the results are pushed out to disk.
+ */
+int
+zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
+    dmu_tx_t *tx)
+{
+       fuid_domain_t searchnode, *findnode;
+       avl_index_t loc;
+       krw_t rw = RW_READER;
+
+       /*
+        * If the dummy "nobody" domain then return an index of 0
+        * to cause the created FUID to be a standard POSIX id
+        * for the user nobody.
+        */
+       if (domain[0] == '\0') {
+               *retdomain = nulldomain;
+               return (0);
+       }
+
+       searchnode.f_ksid = ksid_lookupdomain(domain);
+       if (retdomain) {
+               *retdomain = searchnode.f_ksid->kd_name;
+       }
+       if (!zfsvfs->z_fuid_loaded)
+               zfs_fuid_init(zfsvfs, tx);
+
+retry:
+       rw_enter(&zfsvfs->z_fuid_lock, rw);
+       findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
+
+       if (findnode) {
+               rw_exit(&zfsvfs->z_fuid_lock);
+               ksiddomain_rele(searchnode.f_ksid);
+               return (findnode->f_idx);
+       } else {
+               fuid_domain_t *domnode;
+               nvlist_t *nvp;
+               nvlist_t **fuids;
+               uint64_t retidx;
+               size_t nvsize = 0;
+               char *packed;
+               dmu_buf_t *db;
+               int i = 0;
+
+               if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
+                       rw_exit(&zfsvfs->z_fuid_lock);
+                       rw = RW_WRITER;
+                       goto retry;
+               }
+
+               domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
+               domnode->f_ksid = searchnode.f_ksid;
+
+               retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
+
+               avl_add(&zfsvfs->z_fuid_domain, domnode);
+               avl_add(&zfsvfs->z_fuid_idx, domnode);
+               /*
+                * Now resync the on-disk nvlist.
+                */
+               VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+               domnode = avl_first(&zfsvfs->z_fuid_domain);
+               fuids = kmem_alloc(retidx * sizeof (void *), KM_SLEEP);
+               while (domnode) {
+                       VERIFY(nvlist_alloc(&fuids[i],
+                           NV_UNIQUE_NAME, KM_SLEEP) == 0);
+                       VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
+                           domnode->f_idx) == 0);
+                       VERIFY(nvlist_add_uint64(fuids[i],
+                           FUID_OFFSET, 0) == 0);
+                       VERIFY(nvlist_add_string(fuids[i++], FUID_DOMAIN,
+                           domnode->f_ksid->kd_name) == 0);
+                       domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode);
+               }
+               VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
+                   fuids, retidx) == 0);
+               for (i = 0; i != retidx; i++)
+                       nvlist_free(fuids[i]);
+               kmem_free(fuids, retidx * sizeof (void *));
+               VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
+               packed = kmem_alloc(nvsize, KM_SLEEP);
+               VERIFY(nvlist_pack(nvp, &packed, &nvsize,
+                   NV_ENCODE_XDR, KM_SLEEP) == 0);
+               nvlist_free(nvp);
+               zfsvfs->z_fuid_size = nvsize;
+               dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
+                   zfsvfs->z_fuid_size, packed, tx);
+               kmem_free(packed, zfsvfs->z_fuid_size);
+               VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
+                   FTAG, &db));
+               dmu_buf_will_dirty(db, tx);
+               *(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
+               dmu_buf_rele(db, FTAG);
+
+               rw_exit(&zfsvfs->z_fuid_lock);
+               return (retidx);
+       }
+}
+
+/*
+ * Query domain table by index, returning domain string
+ *
+ * Returns a pointer from an avl node of the domain string.
+ *
+ */
+static char *
+zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
+{
+       char *domain;
+
+       if (idx == 0 || !zfsvfs->z_use_fuids)
+               return (NULL);
+
+       if (!zfsvfs->z_fuid_loaded)
+               zfs_fuid_init(zfsvfs, NULL);
+
+       rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
+
+       if (zfsvfs->z_fuid_obj)
+               domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
+       else
+               domain = nulldomain;
+       rw_exit(&zfsvfs->z_fuid_lock);
+
+       ASSERT(domain);
+       return (domain);
+}
+
+void
+zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
+{
+       *uidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_phys->zp_uid,
+           cr, ZFS_OWNER);
+       *gidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_phys->zp_gid,
+           cr, ZFS_GROUP);
+}
+
+uid_t
+zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
+    cred_t *cr, zfs_fuid_type_t type)
+{
+       uint32_t index = FUID_INDEX(fuid);
+       char *domain;
+       uid_t id;
+
+       if (index == 0)
+               return (fuid);
+
+       domain = zfs_fuid_find_by_idx(zfsvfs, index);
+       ASSERT(domain != NULL);
+
+       if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
+               (void) kidmap_getuidbysid(crgetzone(cr), domain,
+                   FUID_RID(fuid), &id);
+       } else {
+               (void) kidmap_getgidbysid(crgetzone(cr), domain,
+                   FUID_RID(fuid), &id);
+       }
+       return (id);
+}
+
+/*
+ * Add a FUID node to the list of fuid's being created for this
+ * ACL
+ *
+ * If ACL has multiple domains, then keep only one copy of each unique
+ * domain.
+ */
+static void
+zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
+    uint64_t idx, uint64_t id, zfs_fuid_type_t type)
+{
+       zfs_fuid_t *fuid;
+       zfs_fuid_domain_t *fuid_domain;
+       zfs_fuid_info_t *fuidp;
+       uint64_t fuididx;
+       boolean_t found = B_FALSE;
+
+       if (*fuidpp == NULL)
+               *fuidpp = zfs_fuid_info_alloc();
+
+       fuidp = *fuidpp;
+       /*
+        * First find fuid domain index in linked list
+        *
+        * If one isn't found then create an entry.
+        */
+
+       for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
+           fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
+           fuid_domain), fuididx++) {
+               if (idx == fuid_domain->z_domidx) {
+                       found = B_TRUE;
+                       break;
+               }
+       }
+
+       if (!found) {
+               fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
+               fuid_domain->z_domain = domain;
+               fuid_domain->z_domidx = idx;
+               list_insert_tail(&fuidp->z_domains, fuid_domain);
+               fuidp->z_domain_str_sz += strlen(domain) + 1;
+               fuidp->z_domain_cnt++;
+       }
+
+       if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
+               /*
+                * Now allocate fuid entry and add it on the end of the list
+                */
+
+               fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
+               fuid->z_id = id;
+               fuid->z_domidx = idx;
+               fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
+
+               list_insert_tail(&fuidp->z_fuids, fuid);
+               fuidp->z_fuid_cnt++;
+       } else {
+               if (type == ZFS_OWNER)
+                       fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
+               else
+                       fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
+       }
+}
+
+/*
+ * Create a file system FUID, based on information in the users cred
+ */
+uint64_t
+zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
+    dmu_tx_t *tx, cred_t *cr, zfs_fuid_info_t **fuidp)
+{
+       uint64_t        idx;
+       ksid_t          *ksid;
+       uint32_t        rid;
+       char            *kdomain;
+       const char      *domain;
+       uid_t           id;
+
+       VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
+
+       ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
+       if (ksid) {
+               id = ksid_getid(ksid);
+       } else {
+               if (type == ZFS_OWNER)
+                       id = crgetuid(cr);
+               else
+                       id = crgetgid(cr);
+       }
+
+       if (!zfsvfs->z_use_fuids || (!IS_EPHEMERAL(id)))
+               return ((uint64_t)id);
+
+       rid = ksid_getrid(ksid);
+       domain = ksid_getdomain(ksid);
+
+       idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+
+       zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
+
+       return (FUID_ENCODE(idx, rid));
+}
+
+/*
+ * Create a file system FUID for an ACL ace
+ * or a chown/chgrp of the file.
+ * This is similar to zfs_fuid_create_cred, except that
+ * we can't find the domain + rid information in the
+ * cred.  Instead we have to query Winchester for the
+ * domain and rid.
+ *
+ * During replay operations the domain+rid information is
+ * found in the zfs_fuid_info_t that the replay code has
+ * attached to the zfsvfs of the file system.
+ */
+uint64_t
+zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
+    zfs_fuid_type_t type, dmu_tx_t *tx, zfs_fuid_info_t **fuidpp)
+{
+       const char *domain;
+       char *kdomain;
+       uint32_t fuid_idx = FUID_INDEX(id);
+       uint32_t rid;
+       idmap_stat status;
+       uint64_t idx;
+       boolean_t is_replay = (zfsvfs->z_assign >= TXG_INITIAL);
+       zfs_fuid_t *zfuid = NULL;
+       zfs_fuid_info_t *fuidp;
+
+       /*
+        * If POSIX ID, or entry is already a FUID then
+        * just return the id
+        *
+        * We may also be handed an already FUID'ized id via
+        * chmod.
+        */
+
+       if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
+               return (id);
+
+       if (is_replay) {
+               fuidp = zfsvfs->z_fuid_replay;
+
+               /*
+                * If we are passed an ephemeral id, but no
+                * fuid_info was logged then return NOBODY.
+                * This is most likely a result of idmap service
+                * not being available.
+                */
+               if (fuidp == NULL)
+                       return (UID_NOBODY);
+
+               switch (type) {
+               case ZFS_ACE_USER:
+               case ZFS_ACE_GROUP:
+                       zfuid = list_head(&fuidp->z_fuids);
+                       rid = FUID_RID(zfuid->z_logfuid);
+                       idx = FUID_INDEX(zfuid->z_logfuid);
+                       break;
+               case ZFS_OWNER:
+                       rid = FUID_RID(fuidp->z_fuid_owner);
+                       idx = FUID_INDEX(fuidp->z_fuid_owner);
+                       break;
+               case ZFS_GROUP:
+                       rid = FUID_RID(fuidp->z_fuid_group);
+                       idx = FUID_INDEX(fuidp->z_fuid_group);
+                       break;
+               };
+               domain = fuidp->z_domain_table[idx -1];
+       } else {
+               if (type == ZFS_OWNER || type == ZFS_ACE_USER)
+                       status = kidmap_getsidbyuid(crgetzone(cr), id,
+                           &domain, &rid);
+               else
+                       status = kidmap_getsidbygid(crgetzone(cr), id,
+                           &domain, &rid);
+
+               if (status != 0) {
+                       /*
+                        * When returning nobody we will need to
+                        * make a dummy fuid table entry for logging
+                        * purposes.
+                        */
+                       rid = UID_NOBODY;
+                       domain = nulldomain;
+               }
+       }
+
+       idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+
+       if (!is_replay)
+               zfs_fuid_node_add(fuidpp, kdomain, rid, idx, id, type);
+       else if (zfuid != NULL) {
+               list_remove(&fuidp->z_fuids, zfuid);
+               kmem_free(zfuid, sizeof (zfs_fuid_t));
+       }
+       return (FUID_ENCODE(idx, rid));
+}
+
+void
+zfs_fuid_destroy(zfsvfs_t *zfsvfs)
+{
+       rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
+       if (!zfsvfs->z_fuid_loaded) {
+               rw_exit(&zfsvfs->z_fuid_lock);
+               return;
+       }
+       zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
+       rw_exit(&zfsvfs->z_fuid_lock);
+}
+
+/*
+ * Allocate zfs_fuid_info for tracking FUIDs created during
+ * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
+ */
+zfs_fuid_info_t *
+zfs_fuid_info_alloc(void)
+{
+       zfs_fuid_info_t *fuidp;
+
+       fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
+       list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
+           offsetof(zfs_fuid_domain_t, z_next));
+       list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
+           offsetof(zfs_fuid_t, z_next));
+       return (fuidp);
+}
+
+/*
+ * Release all memory associated with zfs_fuid_info_t
+ */
+void
+zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
+{
+       zfs_fuid_t *zfuid;
+       zfs_fuid_domain_t *zdomain;
+
+       while ((zfuid = list_head(&fuidp->z_fuids)) != NULL) {
+               list_remove(&fuidp->z_fuids, zfuid);
+               kmem_free(zfuid, sizeof (zfs_fuid_t));
+       }
+
+       if (fuidp->z_domain_table != NULL)
+               kmem_free(fuidp->z_domain_table,
+                   (sizeof (char **)) * fuidp->z_domain_cnt);
+
+       while ((zdomain = list_head(&fuidp->z_domains)) != NULL) {
+               list_remove(&fuidp->z_domains, zdomain);
+               kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
+       }
+
+       kmem_free(fuidp, sizeof (zfs_fuid_info_t));
+}
+
+/*
+ * Check to see if id is a groupmember.  If cred
+ * has ksid info then sidlist is checked first
+ * and if still not found then POSIX groups are checked
+ *
+ * Will use a straight FUID compare when possible.
+ */
+boolean_t
+zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
+{
+       ksid_t          *ksid = crgetsid(cr, KSID_GROUP);
+       uid_t           gid;
+
+       if (ksid) {
+               int             i;
+               ksid_t          *ksid_groups;
+               ksidlist_t      *ksidlist = crgetsidlist(cr);
+               uint32_t        idx = FUID_INDEX(id);
+               uint32_t        rid = FUID_RID(id);
+
+               ASSERT(ksidlist);
+               ksid_groups = ksidlist->ksl_sids;
+
+               for (i = 0; i != ksidlist->ksl_nsid; i++) {
+                       if (idx == 0) {
+                               if (id != IDMAP_WK_CREATOR_GROUP_GID &&
+                                   id == ksid_groups[i].ks_id) {
+                                       return (B_TRUE);
+                               }
+                       } else {
+                               char *domain;
+
+                               domain = zfs_fuid_find_by_idx(zfsvfs, idx);
+                               ASSERT(domain != NULL);
+
+                               if (strcmp(domain,
+                                   IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
+                                       return (B_FALSE);
+
+                               if ((strcmp(domain,
+                                   ksid_groups[i].ks_domain->kd_name) == 0) &&
+                                   rid == ksid_groups[i].ks_rid)
+                                       return (B_TRUE);
+                       }
+               }
+       }
+
+       /*
+        * Not found in ksidlist, check posix groups
+        */
+       gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
+       return (groupmember(gid, cr));
+}
+#endif