+#endif /* HAVE_ZPL */
+
+static int
+zfs_space_delta_cb(dmu_object_type_t bonustype, void *data,
+ uint64_t *userp, uint64_t *groupp)
+{
+ znode_phys_t *znp = data;
+ int error = 0;
+
+ /*
+ * Is it a valid type of object to track?
+ */
+ if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
+ return (ENOENT);
+
+ /*
+ * If we have a NULL data pointer
+ * then assume the id's aren't changing and
+ * return EEXIST to the dmu to let it know to
+ * use the same ids
+ */
+ if (data == NULL)
+ return (EEXIST);
+
+ if (bonustype == DMU_OT_ZNODE) {
+ *userp = znp->zp_uid;
+ *groupp = znp->zp_gid;
+ } else {
+ int hdrsize;
+
+ ASSERT(bonustype == DMU_OT_SA);
+ hdrsize = sa_hdrsize(data);
+
+ if (hdrsize != 0) {
+ *userp = *((uint64_t *)((uintptr_t)data + hdrsize +
+ SA_UID_OFFSET));
+ *groupp = *((uint64_t *)((uintptr_t)data + hdrsize +
+ SA_GID_OFFSET));
+ } else {
+ /*
+ * This should only happen for newly created
+ * files that haven't had the znode data filled
+ * in yet.
+ */
+ *userp = 0;
+ *groupp = 0;
+ }
+ }
+ return (error);
+}
+
+#ifdef HAVE_ZPL
+static void
+fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr,
+ char *domainbuf, int buflen, uid_t *ridp)
+{
+ uint64_t fuid;
+ const char *domain;
+
+ fuid = strtonum(fuidstr, NULL);
+
+ domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid));
+ if (domain)
+ (void) strlcpy(domainbuf, domain, buflen);
+ else
+ domainbuf[0] = '\0';
+ *ridp = FUID_RID(fuid);
+}
+
+static uint64_t
+zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type)
+{
+ switch (type) {
+ case ZFS_PROP_USERUSED:
+ return (DMU_USERUSED_OBJECT);
+ case ZFS_PROP_GROUPUSED:
+ return (DMU_GROUPUSED_OBJECT);
+ case ZFS_PROP_USERQUOTA:
+ return (zfsvfs->z_userquota_obj);
+ case ZFS_PROP_GROUPQUOTA:
+ return (zfsvfs->z_groupquota_obj);
+ }
+ return (0);
+}
+
+int
+zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+ uint64_t *cookiep, void *vbuf, uint64_t *bufsizep)
+{
+ int error;
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ zfs_useracct_t *buf = vbuf;
+ uint64_t obj;
+
+ if (!dmu_objset_userspace_present(zfsvfs->z_os))
+ return (ENOTSUP);
+
+ obj = zfs_userquota_prop_to_obj(zfsvfs, type);
+ if (obj == 0) {
+ *bufsizep = 0;
+ return (0);
+ }
+
+ for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep);
+ (error = zap_cursor_retrieve(&zc, &za)) == 0;
+ zap_cursor_advance(&zc)) {
+ if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) >
+ *bufsizep)
+ break;
+
+ fuidstr_to_sid(zfsvfs, za.za_name,
+ buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid);
+
+ buf->zu_space = za.za_first_integer;
+ buf++;
+ }
+ if (error == ENOENT)
+ error = 0;
+
+ ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep);
+ *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf;
+ *cookiep = zap_cursor_serialize(&zc);
+ zap_cursor_fini(&zc);
+ return (error);
+}
+
+/*
+ * buf must be big enough (eg, 32 bytes)
+ */
+static int
+id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid,
+ char *buf, boolean_t addok)
+{
+ uint64_t fuid;
+ int domainid = 0;
+
+ if (domain && domain[0]) {
+ domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok);
+ if (domainid == -1)
+ return (ENOENT);
+ }
+ fuid = FUID_ENCODE(domainid, rid);
+ (void) sprintf(buf, "%llx", (longlong_t)fuid);
+ return (0);
+}
+
+int
+zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+ const char *domain, uint64_t rid, uint64_t *valp)
+{
+ char buf[32];
+ int err;
+ uint64_t obj;
+
+ *valp = 0;
+
+ if (!dmu_objset_userspace_present(zfsvfs->z_os))
+ return (ENOTSUP);
+
+ obj = zfs_userquota_prop_to_obj(zfsvfs, type);
+ if (obj == 0)
+ return (0);
+
+ err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE);
+ if (err)
+ return (err);
+
+ err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp);
+ if (err == ENOENT)
+ err = 0;
+ return (err);
+}
+
+int
+zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+ const char *domain, uint64_t rid, uint64_t quota)
+{
+ char buf[32];
+ int err;
+ dmu_tx_t *tx;
+ uint64_t *objp;
+ boolean_t fuid_dirtied;
+
+ if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
+ return (EINVAL);
+
+ if (zfsvfs->z_version < ZPL_VERSION_USERSPACE)
+ return (ENOTSUP);
+
+ objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj :
+ &zfsvfs->z_groupquota_obj;
+
+ err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE);
+ if (err)
+ return (err);
+ fuid_dirtied = zfsvfs->z_fuid_dirty;
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
+ if (*objp == 0) {
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
+ zfs_userquota_prop_prefixes[type]);
+ }
+ if (fuid_dirtied)
+ zfs_fuid_txhold(zfsvfs, tx);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+
+ mutex_enter(&zfsvfs->z_lock);
+ if (*objp == 0) {
+ *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA,
+ DMU_OT_NONE, 0, tx);
+ VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
+ }
+ mutex_exit(&zfsvfs->z_lock);
+
+ if (quota == 0) {
+ err = zap_remove(zfsvfs->z_os, *objp, buf, tx);
+ if (err == ENOENT)
+ err = 0;
+ } else {
+ err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx);
+ }
+ ASSERT(err == 0);
+ if (fuid_dirtied)
+ zfs_fuid_sync(zfsvfs, tx);
+ dmu_tx_commit(tx);
+ return (err);
+}
+
+boolean_t
+zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid)
+{
+ char buf[32];
+ uint64_t used, quota, usedobj, quotaobj;
+ int err;
+
+ usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
+ quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;
+
+ if (quotaobj == 0 || zfsvfs->z_replay)
+ return (B_FALSE);
+
+ (void) sprintf(buf, "%llx", (longlong_t)fuid);
+ err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a);
+ if (err != 0)
+ return (B_FALSE);
+
+ err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used);
+ if (err != 0)
+ return (B_FALSE);
+ return (used >= quota);
+}
+
+boolean_t
+zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup)
+{
+ uint64_t fuid;
+ uint64_t quotaobj;
+
+ quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;
+
+ fuid = isgroup ? zp->z_gid : zp->z_uid;
+
+ if (quotaobj == 0 || zfsvfs->z_replay)
+ return (B_FALSE);
+
+ return (zfs_fuid_overquota(zfsvfs, isgroup, fuid));
+}
+
+int
+zfsvfs_create(const char *osname, zfsvfs_t **zfvp)
+{
+ objset_t *os;
+ zfsvfs_t *zfsvfs;
+ uint64_t zval;
+ int i, error;
+ uint64_t sa_obj;
+
+ zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
+
+ /*
+ * We claim to always be readonly so we can open snapshots;
+ * other ZPL code will prevent us from writing to snapshots.
+ */
+ error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os);
+ if (error) {
+ kmem_free(zfsvfs, sizeof (zfsvfs_t));
+ return (error);
+ }
+
+ /*
+ * Initialize the zfs-specific filesystem structure.
+ * Should probably make this a kmem cache, shuffle fields,
+ * and just bzero up to z_hold_mtx[].
+ */
+ zfsvfs->z_vfs = NULL;
+ zfsvfs->z_parent = zfsvfs;
+ zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
+ zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
+ zfsvfs->z_os = os;
+
+ error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
+ if (error) {
+ goto out;
+ } else if (zfsvfs->z_version >
+ zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
+ (void) printf("Can't mount a version %lld file system "
+ "on a version %lld pool\n. Pool must be upgraded to mount "
+ "this file system.", (u_longlong_t)zfsvfs->z_version,
+ (u_longlong_t)spa_version(dmu_objset_spa(os)));
+ error = ENOTSUP;
+ goto out;
+ }
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
+ goto out;
+ zfsvfs->z_norm = (int)zval;
+
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
+ goto out;
+ zfsvfs->z_utf8 = (zval != 0);
+
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
+ goto out;
+ zfsvfs->z_case = (uint_t)zval;
+
+ /*
+ * Fold case on file systems that are always or sometimes case
+ * insensitive.
+ */
+ if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
+ zfsvfs->z_case == ZFS_CASE_MIXED)
+ zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
+
+ zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
+ zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os);
+
+ if (zfsvfs->z_use_sa) {
+ /* should either have both of these objects or none */
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
+ &sa_obj);
+ if (error)
+ return (error);
+ } else {
+ /*
+ * Pre SA versions file systems should never touch
+ * either the attribute registration or layout objects.
+ */
+ sa_obj = 0;
+ }
+
+ error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
+ &zfsvfs->z_attr_table);
+ if (error)
+ goto out;
+
+ if (zfsvfs->z_version >= ZPL_VERSION_SA)
+ sa_register_update_callback(os, zfs_sa_upgrade);
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
+ &zfsvfs->z_root);
+ if (error)
+ goto out;
+ ASSERT(zfsvfs->z_root != 0);
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
+ &zfsvfs->z_unlinkedobj);
+ if (error)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
+ 8, 1, &zfsvfs->z_userquota_obj);
+ if (error && error != ENOENT)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
+ 8, 1, &zfsvfs->z_groupquota_obj);
+ if (error && error != ENOENT)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
+ &zfsvfs->z_fuid_obj);
+ if (error && error != ENOENT)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
+ &zfsvfs->z_shares_dir);
+ if (error && error != ENOENT)
+ goto out;
+
+ mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
+ offsetof(znode_t, z_link_node));
+ rrw_init(&zfsvfs->z_teardown_lock);
+ rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
+ rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
+
+ *zfvp = zfsvfs;
+ return (0);
+
+out:
+ dmu_objset_disown(os, zfsvfs);
+ *zfvp = NULL;
+ kmem_free(zfsvfs, sizeof (zfsvfs_t));
+ return (error);
+}