s3:open.c: Fix a typo

Signed-off-by: Pavel Filipenský <pfilipensky@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

Autobuild-User(master): Pavel Filipensky <pfilipensky@samba.org>
Autobuild-Date(master): Tue Dec 17 11:23:50 UTC 2024 on atb-devel-224

ctdb/config/functions

@@ -1160,11 +1160,8 @@ nfs_callout_init()
 	export CTDB_NFS_CALLOUT_STATE_DIR="${_state_dir}/callout-state"
 
 	# Export, if set, for use by clustered NFS callouts
-	if [ -n "$CTDB_NFS_STATE_FS_TYPE" ]; then
-		export CTDB_NFS_STATE_FS_TYPE
-	fi
-	if [ -n "$CTDB_NFS_STATE_MNT" ]; then
-		export CTDB_NFS_STATE_MNT
+	if [ -n "$CTDB_NFS_SHARED_STATE_DIR" ]; then
+		export CTDB_NFS_SHARED_STATE_DIR
 	fi
 	if [ -n "$CTDB_NFS_EXPORTS_FILE" ]; then
 		export CTDB_NFS_EXPORTS_FILE

ctdb/doc/ctdb-script.options.5.xml

@@ -975,6 +975,19 @@ CTDB_PER_IP_ROUTING_TABLE_ID_HIGH=9000
 	  </listitem>
 	</varlistentry>
 
+	<varlistentry>
+	  <term>
+	    CTDB_NFS_SHARED_STATE_DIR=<parameter>DIRECTORY</parameter>
+	  </term>
+	  <listitem>
+	    <para>
+	      DIRECTORY where clustered NFS shared state will be
+	      located.  DIRECTORY should be in a cluster filesystem
+	      that is shared between the nodes.  No default.
+	    </para>
+	  </listitem>
+	</varlistentry>
+
 	<varlistentry>
 	  <term>
 	    CTDB_NFS_SKIP_SHARE_CHECK=yes|no
@@ -1030,25 +1043,81 @@ CTDB_PER_IP_ROUTING_TABLE_ID_HIGH=9000
 
 	<varlistentry>
 	  <term>
-	    CTDB_NFS_STATE_FS_TYPE=<parameter>TYPE</parameter>
+	    CTDB_STATD_CALLOUT_SHARED_STORAGE=<parameter>LOCATION</parameter>
 	  </term>
 	  <listitem>
 	    <para>
-	      The type of filesystem used for a clustered NFS' shared
-	      state. No default.
-	    </para>
-	  </listitem>
-	</varlistentry>
-
-	<varlistentry>
-	  <term>
-	    CTDB_NFS_STATE_MNT=<parameter>DIR</parameter>
-	  </term>
-	  <listitem>
-	    <para>
-	      The directory where a clustered NFS' shared state will be
-	      located. No default.
+	      LOCATION where NFSv3 statd state will be stored.  Valid
+	      values are:
 	    </para>
+	    <variablelist>
+	      <varlistentry>
+		<term>
+		  persistent_db<optional>:<parameter>TDB</parameter></optional>
+		</term>
+		<listitem>
+		  <para>
+		    Data is queued to local storage and then dequeued
+		    to TDB during monitor events.  This means there is
+		    a window where locking state may be lost.
+		    However, this works around performance limitations
+		    in CTDB's persistent database handling.
+		  </para>
+		  <para>
+		    If :TDB is omitted then TDB defaults to
+		    <filename>ctdb_statd_callout.tdb</filename>.
+		  </para>
+		</listitem>
+	      </varlistentry>
+	      <varlistentry>
+		<term>
+		  shared_dir<optional>:<parameter>DIRECTORY</parameter></optional>
+		</term>
+		<listitem>
+		  <para>
+		    DIRECTORY is a directory in a cluster filesystem
+		    that is shared between the nodes.  If DIRECTORY is
+		    relative (i.e. does not start with '/') then it is
+		    appended to CTDB_NFS_SHARED_STATE_DIR.  If
+		    :DIRECTORY is omitted then DIRECTORY defaults to
+		    <filename>statd</filename>.
+		  </para>
+		  <para>
+		    Using a shared directory may result in performance
+		    and/or stability problems.  rpc.statd is
+		    single-threaded and its HA callout is called
+		    synchronously, causing any latency introduced by
+		    the callout to be cumulative.  Stability issues
+		    are most likely if thousands of clients reclaim
+		    locks after failover and use of the cluster
+		    filesystem introduces too much additional
+		    latency.  Too much latency in in the HA callout
+		    may cause rpc.statd to fail health monitoring.
+		  </para>
+		</listitem>
+	      </varlistentry>
+	      <varlistentry>
+		<term>
+		  none
+		</term>
+		<listitem>
+		  <para>
+		    No cluster-aware handling of NFSv3 statd state is
+		    done.  NFSv3 lock reclaim will not occur and
+		    applications that use locking over NFSv3 are
+		    likely to lose or corrupt data.
+		  </para>
+		  <para>
+		    This should be used with care and only in the case
+		    where no applications are using POSIX locks in
+		    NFSv3 mounts.  It should probably be considered an
+		    option to test the latency of
+		    <filename>statd_callout</filename>, without
+		    including any storage costs.
+		  </para>
+		</listitem>
+	      </varlistentry>
+	    </variablelist>
 	  </listitem>
 	</varlistentry>
 

ctdb/doc/examples/config_migrate.sh

@@ -255,8 +255,7 @@ CTDB_NFS_CHECKS_DIR
 CTDB_NFS_SKIP_SHARE_CHECK
 CTDB_RPCINFO_LOCALHOST
 CTDB_RPCINFO_LOCALHOST6
-CTDB_NFS_STATE_FS_TYPE
-CTDB_NFS_STATE_MNT
+CTDB_NFS_SHARED_STATE_DIR
 # 70.iscsi
 CTDB_START_ISCSI_SCRIPTS
 # 00.ctdb

ctdb/doc/examples/nfs-ganesha-callout

@@ -36,10 +36,40 @@
 # Exit on 1st error
 set -e
 
-# Filesystem type and mount point for the (typically clustered)
-# volume that will contain the NFS-Ganesha state.
-state_fs="${CTDB_NFS_STATE_FS_TYPE:-gpfs}"
-state_dir="${CTDB_NFS_STATE_MNT}" # No sane default.
+die()
+{
+	echo "$1"
+	exit 1
+}
+
+# Shared directory, typically on a cluster filesystem, that will
+# contain the NFS-Ganesha state
+if [ -z "$CTDB_NFS_SHARED_STATE_DIR" ]; then
+	die "$0: CTDB_NFS_SHARED_STATE_DIR is not set"
+fi
+
+if [ ! -d "$CTDB_NFS_SHARED_STATE_DIR" ]; then
+	t="$CTDB_NFS_SHARED_STATE_DIR" # Readability, below
+	die "$0: CTDB_NFS_SHARED_STATE_DIR=${t} not found"
+fi
+
+state_fs=$(findmnt -n --target "$CTDB_NFS_SHARED_STATE_DIR" -o FSTYPE)
+
+case "$state_fs" in
+glusterfs | gpfs)
+	:
+	;;
+fuse.glusterfs)
+	state_fs="glusterfs"
+	;;
+*)
+	d="$CTDB_NFS_SHARED_STATE_DIR"
+	die "$0: filesystem type \"${state_fs}\" is not supported for ${d}"
+	;;
+esac
+
+# Always put NFS-Ganesha state in its own subdirectory
+state_dir="${CTDB_NFS_SHARED_STATE_DIR}/ganesha"
 
 # Location of exports file
 nfs_exports_file="${CTDB_NFS_EXPORTS_FILE:-/etc/ganesha/ganesha.conf}"
@@ -47,7 +77,6 @@ nfs_exports_file="${CTDB_NFS_EXPORTS_FILE:-/etc/ganesha/ganesha.conf}"
 # To change the following, edit the default values below.  Do not set
 # these - they aren't configuration variables, just hooks for testing.
 nfs_service="${CTDB_NFS_SERVICE:-nfs-ganesha}"
-ganesha_rec_subdir=${CTDB_GANESHA_REC_SUBDIR:-.ganesha}
 procfs=${PROCFS_PATH:-/proc}
 
 case "$state_fs" in
@@ -55,13 +84,9 @@ gpfs)
 	GANRECDIR="/var/lib/nfs/ganesha"
 	;;
 glusterfs)
-	if [ -z "${state_dir}" ]; then
-		echo "CTDB_NFS_STATE_MNT not defined for GlusterFS"
-		exit 1
-	fi
 	host=$(hostname)
-	NODESTATEDIR="$state_dir/nfs-ganesha/$host"
-	GANSTATEDIR="$state_dir/nfs-ganesha/.noderefs"
+	NODESTATEDIR="${state_dir}/${host}"
+	GANSTATEDIR="${state_dir}/.noderefs"
 	NODESTATELN="$GANSTATEDIR/$host"
 	;;
 esac
@@ -184,21 +209,10 @@ get_cluster_fs_state()
 
 create_ganesha_recdirs()
 {
-	if ! _mounts=$(mount | grep "$state_fs"); then
-		echo "Failed to find mounts of type $state_fs"
-		exit 1
-	fi
-	if [ -z "$_mounts" ]; then
-		echo "startup $state_fs not ready"
-		exit 0
-	fi
-
 	case "$state_fs" in
 	gpfs)
-		_mntpt=$(echo "$_mounts" | sort | awk 'NR == 1 {print $3}')
-		_link_dst="${_mntpt}/${ganesha_rec_subdir}"
-		mkdir -vp "$_link_dst"
-		check_ln "$_link_dst" "$GANRECDIR"
+		mkdir -vp "$state_dir"
+		check_ln "$state_dir" "$GANRECDIR"
 		;;
 	glusterfs)
 		[ -d /var/lib/nfs.backup ] ||
@@ -279,6 +293,7 @@ nfs_stats()
 		# unhelpful for avoiding an unhealthy service
 		echo "Not implemented" >&2
 		exit 1
+		;;
 	esac
 }
 

ctdb/failover/statd_callout.c

@@ -35,11 +35,16 @@
  * The first line is the mode.  Currently supported modes are:
  *
  *   persistent_db
+ *   shared_dir
+ *   none
  *
- * In this mode, the file contains 2 subsequent lines of text:
+ * In persistent_db and shared_dir modes, the file contains 2
+ * subsequent lines of text:
  *
  *   path: directory where files should be created
  *   ips_file: file containing node's currently assigned public IP addresses
+ *
+ * In none mode, there are no subsequent lines.
  */
 #define CONFIG_FILE CTDB_VARDIR "/scripts/statd_callout.conf"
 
@@ -48,6 +53,8 @@ static const char *progname;
 struct {
 	enum {
 		CTDB_SC_MODE_PERSISTENT_DB,
+		CTDB_SC_MODE_SHARED_DIR,
+		CTDB_SC_MODE_NONE,
 	} mode;
 	union {
 		struct {
@@ -87,10 +94,14 @@ static void free_config(void)
 {
 	switch (config.mode) {
 	case CTDB_SC_MODE_PERSISTENT_DB:
+	case CTDB_SC_MODE_SHARED_DIR:
 		free(config.path);
 		config.path = NULL;
 		free(config.ips_file);
 		config.ips_file = NULL;
+		break;
+	case CTDB_SC_MODE_NONE:
+		break;
 	}
 }
 
@@ -127,6 +138,10 @@ static void read_config(void)
 	}
 	if (strcmp(mode, "persistent_db") == 0) {
 		config.mode = CTDB_SC_MODE_PERSISTENT_DB;
+	} else if (strcmp(mode, "shared_dir") == 0) {
+		config.mode = CTDB_SC_MODE_SHARED_DIR;
+	} else if (strcmp(mode, "none") == 0) {
+		config.mode = CTDB_SC_MODE_NONE;
 	} else {
 		fprintf(stderr,
 			"%s: unknown mode=%s in %s\n",
@@ -140,6 +155,7 @@ static void read_config(void)
 
 	switch (config.mode) {
 	case CTDB_SC_MODE_PERSISTENT_DB:
+	case CTDB_SC_MODE_SHARED_DIR:
 		status = getline_strip(&config.path, &n, f);
 		if (!status) {
 			goto parse_error;
@@ -150,6 +166,8 @@ static void read_config(void)
 			goto parse_error;
 		}
 
+		break;
+	case CTDB_SC_MODE_NONE:
 		break;
 	}
 
@@ -288,6 +306,51 @@ static void del_client_persistent_db(const char *cip)
 	for_each_sip(del_client_persistent_db_line, cip);
 }
 
+static void add_client_shared_dir_line(const char *sip, const char *cip)
+{
+	char path[PATH_MAX];
+	FILE *f;
+
+	make_path(path, sizeof(path), sip, cip);
+
+	f = fopen(path, "w");
+	if (f == NULL) {
+		fprintf(stderr,
+			"%s: unable to open for writing %s\n",
+			progname,
+			path);
+		exit(1);
+	}
+	fclose(f);
+}
+
+static void add_client_shared_dir(const char *cip)
+{
+	for_each_sip(add_client_shared_dir_line, cip);
+}
+
+static void del_client_shared_dir_line(const char *sip, const char *cip)
+{
+	char path[PATH_MAX];
+	int ret;
+
+	make_path(path, sizeof(path), sip, cip);
+
+	ret = unlink(path);
+	if (ret != 0) {
+		fprintf(stderr,
+			"%s: unable to remove %s\n",
+			progname,
+			path);
+		exit(1);
+	}
+}
+
+static void del_client_shared_dir(const char *cip)
+{
+	for_each_sip(del_client_shared_dir_line, cip);
+}
+
 static void usage(void)
 {
 	printf("usage: %s: { add-client | del-client } <client-ip>\n", progname);
@@ -313,6 +376,11 @@ int main(int argc, const char *argv[])
 		case CTDB_SC_MODE_PERSISTENT_DB:
 			add_client_persistent_db(mon_name);
 			break;
+		case CTDB_SC_MODE_SHARED_DIR:
+			add_client_shared_dir(mon_name);
+			break;
+		case CTDB_SC_MODE_NONE:
+			break;
 		}
 	} else if (strcmp(event, "del-client") == 0) {
 		mon_name = argv[2];
@@ -320,6 +388,11 @@ int main(int argc, const char *argv[])
 		case CTDB_SC_MODE_PERSISTENT_DB:
 			del_client_persistent_db(mon_name);
 			break;
+		case CTDB_SC_MODE_SHARED_DIR:
+			del_client_shared_dir(mon_name);
+			break;
+		case CTDB_SC_MODE_NONE:
+			break;
 		}
 	} else {
 		usage();

ctdb/tests/UNIT/eventscripts/scripts/statd-callout.sh

@@ -1,10 +1,28 @@
 setup()
 {
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="$1"
+
 	setup_public_addresses
 	ctdb_set_pnn
 	setup_date "1234567890"
 
 	export FAKE_NFS_HOSTNAME="cluster1"
+
+	case "$CTDB_STATD_CALLOUT_SHARED_STORAGE" in
+	"" | persistent_db)
+		CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db:statd_foo.tdb"
+		;;
+	shared_dir)
+		export CTDB_NFS_SHARED_STATE_DIR="/clusterfs"
+		;;
+	esac
+
+	export CTDB_STATD_CALLOUT_SHARED_STORAGE
+	statd_callout_mode="${CTDB_STATD_CALLOUT_SHARED_STORAGE%%:*}"
+	statd_callout_location="${CTDB_STATD_CALLOUT_SHARED_STORAGE#*:}"
+	if [ "$statd_callout_location" = "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+		statd_callout_location=""
+	fi
 }
 
 ctdb_catdb_format_pairs()
@@ -44,12 +62,64 @@ EOF
 		done |
 		ctdb_catdb_format_pairs | {
 		ok
-		simple_test_command ctdb catdb ctdb.tdb
+		simple_test_command ctdb catdb "$statd_callout_location"
 	} || exit $?
 }
 
+check_shared_dir_statd_state()
+{
+	ctdb_get_my_public_addresses |
+		while read -r _ _sip _; do
+			for _cip; do
+				echo "statd-state@${_sip}@${_cip}"
+			done
+		done |
+		sort | {
+		ok
+		_dir="${CTDB_TEST_TMP_DIR}${statd_callout_location}"
+		mkdir -p "$_dir"
+		(cd "$_dir" && simple_test_command ls)
+	} || exit $?
+}
+
+check_shared_storage_statd_state()
+{
+	case "$statd_callout_mode" in
+	persistent_db)
+		if [ -z "$statd_callout_location" ]; then
+			statd_callout_location="statd_foo.tdb"
+		fi
+		check_ctdb_tdb_statd_state "$@"
+		;;
+	shared_dir)
+		if [ -z "$statd_callout_location" ]; then
+			statd_callout_location="statd"
+		fi
+		case "$statd_callout_location" in
+		/*)
+			:
+			;;
+		*)
+			_t="$CTDB_NFS_SHARED_STATE_DIR"
+			statd_callout_location="${_t}/${statd_callout_location}"
+			;;
+		esac
+		check_shared_dir_statd_state "$@"
+		;;
+	none)
+		:
+		;;
+	esac
+}
+
 check_statd_callout_smnotify()
 {
+	case "$statd_callout_mode" in
+	none)
+		return
+		;;
+	esac
+
 	# The state here doesn't really matter because the date stub
 	# generates a fixed value (as per above setup() function,
 	# which happens to set it to an even value).  In reality,

ctdb/tests/UNIT/eventscripts/statd-callout.001.sh

@@ -2,13 +2,18 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "single add-client"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - single add-client"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
 simple_test_event "add-client" "192.168.123.45"
 simple_test_event "update"
 
-check_ctdb_tdb_statd_state "192.168.123.45"
+check_shared_storage_statd_state "192.168.123.45"

ctdb/tests/UNIT/eventscripts/statd-callout.002.sh

@@ -2,9 +2,14 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "2 x add-client, update"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - 2 x add-client, update"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
@@ -12,4 +17,4 @@ simple_test_event "add-client" "192.168.123.45"
 simple_test_event "add-client" "192.168.123.46"
 simple_test_event "update"
 
-check_ctdb_tdb_statd_state "192.168.123.45" "192.168.123.46"
+check_shared_storage_statd_state "192.168.123.45" "192.168.123.46"

ctdb/tests/UNIT/eventscripts/statd-callout.003.sh

@@ -2,9 +2,14 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "add-client, update, del-client, update"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - add-client, update, del-client, update"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
@@ -14,4 +19,4 @@ simple_test_event "update"
 simple_test_event "del-client" "192.168.123.45"
 simple_test_event "update"
 
-check_ctdb_tdb_statd_state
+check_shared_storage_statd_state

ctdb/tests/UNIT/eventscripts/statd-callout.004.sh

@@ -2,17 +2,22 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "single add-client, notify"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - single add-client, notify"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
 simple_test_event "add-client" "192.168.123.45"
 simple_test_event "update"
 
-check_ctdb_tdb_statd_state "192.168.123.45"
+check_shared_storage_statd_state "192.168.123.45"
 
 check_statd_callout_smnotify "192.168.123.45"
 
-check_ctdb_tdb_statd_state
+check_shared_storage_statd_state

ctdb/tests/UNIT/eventscripts/statd-callout.005.sh

@@ -2,9 +2,14 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "2 x add-client to different nodes, notify on 1"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - 2 x add-client to different nodes, notify on 1"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
@@ -24,4 +29,4 @@ check_statd_callout_smnotify "192.168.123.45"
 
 ctdb_set_pnn 1
 
-check_ctdb_tdb_statd_state "192.168.123.46"
+check_shared_storage_statd_state "192.168.123.46"

ctdb/tests/UNIT/eventscripts/statd-callout.006.sh

@@ -2,9 +2,14 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "2 x add-client to different nodes, notify on both"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - 2 x add-client to different nodes, notify on both"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
@@ -26,4 +31,4 @@ ctdb_set_pnn 1
 
 check_statd_callout_smnotify "192.168.123.46"
 
-check_ctdb_tdb_statd_state
+check_shared_storage_statd_state

ctdb/tests/UNIT/eventscripts/statd-callout.007.sh

@@ -2,9 +2,14 @@
 
 . "${TEST_SCRIPTS_DIR}/unit.sh"
 
-define_test "add-client, del-client, update"
+if [ -z "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	CTDB_STATD_CALLOUT_SHARED_STORAGE="persistent_db"
+fi
+mode="$CTDB_STATD_CALLOUT_SHARED_STORAGE"
 
-setup
+define_test "${mode} - add-client, del-client, update"
+
+setup "$mode"
 
 ok_null
 simple_test_event "startup"
@@ -12,4 +17,4 @@ simple_test_event "add-client" "192.168.123.45"
 simple_test_event "del-client" "192.168.123.45"
 simple_test_event "update"
 
-check_ctdb_tdb_statd_state
+check_shared_storage_statd_state

ctdb/tests/UNIT/eventscripts/statd-callout.101.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.001.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.102.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.002.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.103.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.003.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.104.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.004.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.105.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.005.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.106.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.006.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.107.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="shared_dir"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.007.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.201.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.001.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.202.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.002.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.203.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.003.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.204.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.004.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.205.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.005.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.206.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.006.sh"

ctdb/tests/UNIT/eventscripts/statd-callout.207.sh

@@ -0,0 +1,6 @@
+#!/bin/sh
+
+CTDB_STATD_CALLOUT_SHARED_STORAGE="none"
+
+_dir=$(dirname "$0")
+. "${_dir}/statd-callout.007.sh"

ctdb/tools/statd_callout_helper

@@ -1,6 +1,7 @@
 #!/bin/sh
 
-# statd must be configured to use this script as its high availability call-out.
+# statd must be configured to use statd_callout, CTDB's binary
+# counterpart to this script, as its availability call-out.
 #
 # Modern NFS utils versions use /etc/nfs.conf:
 #
@@ -24,7 +25,7 @@
 # silently if /proc/fs/lockd/nlm_end_grace is not found.
 #
 
-if [ -z "$CTDB_BASE" ] ; then
+if [ -z "$CTDB_BASE" ]; then
 	export CTDB_BASE="/usr/local/etc/ctdb"
 fi
 
@@ -37,6 +38,8 @@ die()
 	exit 1
 }
 
+load_script_options "service" "60.nfs"
+
 ############################################################
 
 ctdb_setup_state_dir "service" "nfs"
@@ -78,8 +81,49 @@ create_add_del_client_dir()
 # shellcheck disable=SC2154
 statd_callout_state_dir="${script_state_dir}/statd_callout"
 
-statd_callout_db="ctdb.tdb"
-statd_callout_queue_dir="${statd_callout_state_dir}/queue"
+# Set default value, if necessary
+: "${CTDB_STATD_CALLOUT_SHARED_STORAGE:=persistent_db}"
+
+statd_callout_mode="${CTDB_STATD_CALLOUT_SHARED_STORAGE%%:*}"
+statd_callout_location="${CTDB_STATD_CALLOUT_SHARED_STORAGE#*:}"
+# If not given then mode determines the default location
+if [ "$statd_callout_location" = "$CTDB_STATD_CALLOUT_SHARED_STORAGE" ]; then
+	statd_callout_location=""
+fi
+
+case "$statd_callout_mode" in
+persistent_db)
+	statd_callout_db="${statd_callout_location:-ctdb_statd_callout.tdb}"
+	statd_callout_queue_dir="${statd_callout_state_dir}/queue"
+	;;
+shared_dir)
+	statd_callout_shared_dir="${statd_callout_location:-statd}"
+	case "$statd_callout_shared_dir" in
+	/*)
+		:
+		;;
+	*)
+		if [ -z "$CTDB_NFS_SHARED_STATE_DIR" ]; then
+			die "CTDB_NFS_SHARED_STATE_DIR is not set"
+		fi
+		t="${CTDB_NFS_SHARED_STATE_DIR}/${statd_callout_shared_dir}"
+		statd_callout_shared_dir="$t"
+		;;
+	esac
+
+	if [ -n "$CTDB_TEST_MODE" ]; then
+		t="${CTDB_TEST_TMP_DIR}${statd_callout_shared_dir}"
+		statd_callout_shared_dir="$t"
+	fi
+	;;
+none)
+	:
+	;;
+*)
+	mode="$statd_callout_mode"
+	die "error: unknown CTDB_STATD_CALLOUT_SHARED_STORAGE mode ${mode}"
+	;;
+esac
 
 ############################################################
 
@@ -139,7 +183,83 @@ send_notifies()
 	done
 }
 
-delete_records()
+############################################################
+
+# Use file/key names of the form statd-state@<server-IP>@<client-IP>
+# to track the last "add-client" or "del-client".  These files contain
+# the key and a value, quoted and ready to pass to "ctdb ptrans".  For
+# add-client the value is the date (for debugging) and for del-client
+# the value is empty (representing a delete).  These get pushed to
+# $statd_callout_db during "update", which will generally be run once each
+# "monitor" cycle.  In this way we avoid scalability problems with
+# flood of persistent transactions after a "notify" when all the
+# clients reclaim their locks.
+
+startup_persistent_db()
+{
+	_config_file="$1"
+
+	create_add_del_client_dir "$statd_callout_queue_dir"
+
+	$CTDB attach "$statd_callout_db" persistent
+
+	cat >"$_config_file" <<EOF
+persistent_db
+${statd_callout_queue_dir}
+${CTDB_MY_PUBLIC_IPS_CACHE}
+EOF
+}
+
+# Used via 'grep -F -f "$persistent_db_grep_filter"' to match database
+# keys currently hosted public IPs
+persistent_db_grep_filter="${statd_callout_state_dir}/.grep_filter"
+
+persistent_db_make_grep_filter()
+{
+	while read -r _ip; do
+		echo "statd-state@${_ip}@"
+	done <"$CTDB_MY_PUBLIC_IPS_CACHE" >"$persistent_db_grep_filter"
+}
+
+update_persistent_db()
+{
+	_files="${statd_callout_state_dir}/.file_list"
+	find "$statd_callout_queue_dir" -name "statd-state@*" >"$_files"
+	if [ ! -s "$_files" ]; then
+		# No files!
+		rm "$_files"
+		exit 0
+	fi
+
+	persistent_db_make_grep_filter
+
+	# Use cat instead of direct grep since POSIX grep does not
+	# have -h
+	_items="${statd_callout_state_dir}/.items"
+	xargs cat <"$_files" | grep -F -f "$persistent_db_grep_filter" >"$_items"
+
+	if [ -s "$_items" ]; then
+		if $CTDB ptrans "$statd_callout_db" <"$_items"; then
+			xargs rm -f <"$_files"
+		fi
+	fi
+
+	rm -f "$_files" "$persistent_db_grep_filter" "$_items"
+}
+
+list_records_persistent_db()
+{
+	persistent_db_make_grep_filter
+
+	$CTDB catdb "$statd_callout_db" |
+		sed -n -e 's|^key([0-9]*) = "\([^"]*\)".*|\1|p' |
+		grep -F -f "$persistent_db_grep_filter" |
+		sed -e 's|statd-state@\([^@]*\)@\(.*\)|\1 \2|'
+
+	rm -f "$persistent_db_grep_filter"
+}
+
+delete_records_persistent_db()
 {
 	while read -r _sip _cip; do
 		_key="statd-state@${_sip}@${_cip}"
@@ -147,30 +267,152 @@ delete_records()
 	done | $CTDB ptrans "$statd_callout_db"
 }
 
+cleanup_persistent_db()
+{
+	# Remove any stale touch files (i.e. for IPs not currently
+	# hosted on this node and created since the last "update").
+	# There's nothing else we can do with them at this stage.
+	_pnn=$(ctdb_get_pnn)
+	_ctdb_all_ips=$($CTDB ip all | tail -n +2)
+	echo "$_ctdb_all_ips" |
+		awk -v pnn="$_pnn" 'pnn != $2 { print $1 }' |
+		while read -r _sip; do
+			rm -f "${statd_callout_queue_dir}/statd-state@${_sip}@"*
+		done
+}
+
 ############################################################
 
-# Keep a file per server-IP/client-IP pair, to keep track of the last
-# "add-client" or "del-client'.  These get pushed to a database during
-# "update", which will generally be run once each "monitor" cycle.  In
-# this way we avoid scalability problems with flood of persistent
-# transactions after a "notify" when all the clients re-take their
-# locks.
+# Use file/key names of the form statd-state@<server-IP>@<client-IP>
+# to track the "add-client" and "del-client".  statd_callout add and
+# removes files directly in $statd_callout_shared_dir.  This may
+# result in performance problems if thousands of clients reclaim locks
+# after failover and the cluster filesystem is unable to handle the
+# load.
 
-startup()
+startup_shared_dir()
 {
-	create_add_del_client_dir "$statd_callout_queue_dir"
+	_config_file="$1"
 
-	$CTDB attach "$statd_callout_db" persistent
+	create_add_del_client_dir "$statd_callout_shared_dir"
 
-	_default="${CTDB_SCRIPT_VARDIR}/statd_callout.conf"
-	_config_file="${CTDB_STATD_CALLOUT_CONFIG_FILE:-"${_default}"}"
 	cat >"$_config_file" <<EOF
-persistent_db
-${statd_callout_queue_dir}
+shared_dir
+${statd_callout_shared_dir}
 ${CTDB_MY_PUBLIC_IPS_CACHE}
 EOF
 }
 
+update_shared_dir()
+{
+	:
+}
+
+list_records_shared_dir()
+{
+	while read -r _ip; do
+		ls "${statd_callout_shared_dir}/statd-state@${_ip}@"*
+	done <"$CTDB_MY_PUBLIC_IPS_CACHE" |
+		while read -r _f; do
+			if [ ! -f "$_f" ]; then
+				continue
+			fi
+			_t="${_f#"${statd_callout_shared_dir}/statd-state@"}"
+			_sip="${_t%@*}"
+			_cip="${_t#*@}"
+			echo "$_sip" "$_cip"
+		done
+}
+
+delete_records_shared_dir()
+{
+	while read -r _sip _cip; do
+		echo "${statd_callout_shared_dir}/statd-state@${_sip}@${_cip}"
+	done | xargs rm -f
+}
+
+cleanup_shared_dir()
+{
+	:
+}
+
+############################################################
+
+# No-op implementation
+
+startup_none()
+{
+	_config_file="$1"
+
+	cat >"$_config_file" <<EOF
+none
+EOF
+}
+
+update_none()
+{
+	:
+}
+
+list_records_none()
+{
+	:
+}
+
+delete_records_none()
+{
+	:
+}
+
+cleanup_none()
+{
+	:
+}
+
+############################################################
+
+# Per-mode initialisation
+startup()
+{
+	_default="${CTDB_SCRIPT_VARDIR}/statd_callout.conf"
+	_config_file="${CTDB_STATD_CALLOUT_CONFIG_FILE:-"${_default}"}"
+
+	mkdir -p "$statd_callout_state_dir"
+
+	"startup_${statd_callout_mode}" "$_config_file"
+}
+
+# Process a record queue in local storage and use it to update cluster
+# storage.  For implementations that update cluster storage directly,
+# this will be a no-op.
+update()
+{
+	"update_${statd_callout_mode}"
+}
+
+# Query cluster storage for entries matching this node's server IPs
+# and Write pairs of:
+#   server-IP client-IP
+# to stdout.
+list_records()
+{
+	"list_records_${statd_callout_mode}" | sort
+}
+
+# Read pairs of:
+#   server-IP client-IP
+# from stdin and delete associated records during notify.
+delete_records()
+{
+	"delete_records_${statd_callout_mode}"
+}
+
+# Do any required cleanup
+cleanup()
+{
+	"cleanup_${statd_callout_mode}"
+}
+
 ############################################################
 
 case "$1" in
@@ -179,25 +421,7 @@ startup)
 	;;
 
 update)
-	cd "$statd_callout_queue_dir" ||
-		die "Failed to change directory to \"${statd_callout_queue_dir}\""
-	files=$(echo statd-state@*)
-	if [ "$files" = "statd-state@*" ]; then
-		# No files!
-		exit 0
-	fi
-	sed_expr=$(awk '{
-		ip = $1; gsub(/\./, "\\.", ip);
-		printf "/statd-state@%s@/p\n", ip }' "$CTDB_MY_PUBLIC_IPS_CACHE")
-	# Intentional multi-word expansion for multiple files
-	# shellcheck disable=SC2086
-	items=$(sed -n "$sed_expr" $files)
-	if [ -n "$items" ]; then
-		if echo "$items" | $CTDB ptrans "$statd_callout_db"; then
-			# shellcheck disable=SC2086
-			rm $files
-		fi
-	fi
+	update
 	;;
 
 notify)
@@ -206,12 +430,6 @@ notify)
 	# conflicting locks through other nodes before all locks have been
 	# reclaimed)
 
-	# we need these settings to make sure that no tcp connections survive
-	# across a very fast failover/failback
-	#echo 10 > /proc/sys/net/ipv4/tcp_fin_timeout
-	#echo 0 > /proc/sys/net/ipv4/tcp_max_tw_buckets
-	#echo 0 > /proc/sys/net/ipv4/tcp_max_orphans
-
 	# Delete the notification list for statd, we don't want it to
 	# ping any clients
 	dir=$(find_statd_sm_dir)
@@ -227,53 +445,19 @@ notify)
 	sleep 2
 	"$CTDB_NFS_CALLOUT" "start" "nlockmgr" >/dev/null 2>&1
 
-	# we now need to send out additional statd notifications to ensure
-	# that clients understand that the lockmanager has restarted.
-	# we have three cases:
-	# 1, clients that ignore the ip address the stat notification came from
-	#    and ONLY care about the 'name' in the notify packet.
-	#    these clients ONLY work with lock failover IFF that name
-	#    can be resolved into an ipaddress that matches the one used
-	#    to mount the share.  (==linux clients)
-	#    This is handled when starting lockmanager above,  but those
-	#    packets are sent from the "wrong" ip address, something linux
-	#    clients are ok with, buth other clients will barf at.
-	# 2, Some clients only accept statd packets IFF they come from the
-	#    'correct' ip address.
-	#    Send out the notification using the 'correct' ip address and also
-	#    specify the 'correct' hostname in the statd packet.
-	#    Some clients require both the correct source address and also the
-	#    correct name. (these clients also ONLY work if the ip addresses
-	#    used to map the share can be resolved into the name returned in
-	#    the notify packet.)
-	#
-	# For all IPs we serve, collect info and push to the config database
+	statd_state="${statd_callout_state_dir}/.statd_state"
+	list_records >"$statd_state"
 
-	# Construct a sed expression to take catdb output and produce pairs of:
-	#   server-IP client-IP
-	# but only for the server-IPs that are hosted on this node.
-	sed_expr=$(awk '{
-		       ip = $1; gsub(/\./, "\\.", ip);
-		       printf "s/^key.*=.*statd-state@\\(%s\\)@\\([^\"]*\\).*/\\1 \\2/p\n", ip }' \
-			       "$CTDB_MY_PUBLIC_IPS_CACHE")
+	if [ ! -s "$statd_state" ]; then
+		rm -f "$statd_state"
+		exit 0
+	fi
 
-	statd_state=$($CTDB catdb "$statd_callout_db" |
-		sed -n "$sed_expr" |
-		sort)
-	[ -n "$statd_state" ] || exit 0
+	delete_records <"$statd_state"
+	send_notifies <"$statd_state"
 
-	echo "$statd_state" | send_notifies
-	echo "$statd_state" | delete_records
+	rm -f "$statd_state"
 
-	# Remove any stale touch files (i.e. for IPs not currently
-	# hosted on this node and created since the last "update").
-	# There's nothing else we can do with them at this stage.
-	pnn=$(ctdb_get_pnn)
-	$CTDB ip all |
-		tail -n +2 |
-		awk -v pnn="$pnn" 'pnn != $2 { print $1 }' |
-		while read -r sip; do
-			rm -f "${statd_callout_queue_dir}/statd-state@${sip}@"*
-		done
+	cleanup
 	;;
 esac

docs-xml/manpages/net.8.xml

@@ -1548,12 +1548,33 @@ to show in the result.
 <title>ADS KEYTAB <replaceable>CREATE</replaceable></title>
 
 <para>
-Creates a new keytab file if one doesn't exist with default entries. Default
-entries are kerberos principals created from the machinename of the
-client, the UPN (if it exists) and any Windows SPN(s) associated with the
-computer AD account for the client. If a keytab file already exists then only
-missing kerberos principals from the default entries are added. No changes
-are made to the computer AD account.
+Since Samba 4.21.0, keytab file is created as specified in <smbconfoption
+name="sync machine password to keytab"/>. The keytab is created only for
+<smbconfoption name="kerberos method">secrets only</smbconfoption> and
+<smbconfoption name="kerberos method">secrets and keytab</smbconfoption>. With
+the smb.conf default values for <smbconfoption name="kerberos method"> secrets
+only</smbconfoption> and <smbconfoption name="sync machine password to keytab"/>
+(default is empty) the keytab is not generated at all.  Keytab with a default
+name and SPNs synced from AD is created for <smbconfoption name="kerberos
+method">secrets and keytab</smbconfoption> if <smbconfoption name="sync machine
+password to keytab"/> is missing.
+</para>
+<para>
+Till Samba 4.20.0, two more entries were created by default: the machinename of
+the client (ending with '$') and the UPN (host/domain@REALM).  If these two
+entries are still needed, each must be specified in an own keytab file.
+Example below will generate three keytab files that contain SPNs synced from
+AD, host UPN and machine$ SPN:
+</para>
+<programlisting>
+<smbconfoption name="sync machine password to keytab">
+/etc/krb5.keytab0:sync_spns:machine_password,
+/etc/krb5.keytab1:spns=host/smb.com@SMB.COM:machine_password,
+/etc/krb5.keytab2:account_name:machine_password
+</smbconfoption>
+</programlisting>
+<para>
+No changes are made to the computer AD account.
 </para>
 </refsect2>
 

docs-xml/manpages/vfs_crossrename.8.xml

@@ -62,7 +62,10 @@
 		</varlistentry>
 	</variablelist>
 
-	<para>This module is stackable.</para>
+	<para> This module is not fully stackable. It can be combined with other
+	modules, but should be the last module in the <command>vfs objects</command>
+	list. It directly access the files in the OS filesystem.
+	</para>
 
 </refsect1>
 

selftest/target/Samba3.pm

@@ -2783,6 +2783,9 @@ sub provision($$)
 	my $recycle_shrdir="$shrdir/recycle";
 	push(@dirs,$recycle_shrdir);
 
+	my $recycle_shrdir2="$shrdir/recycle2";
+	push(@dirs,$recycle_shrdir2);
+
 	my $fakedircreatetimes_shrdir="$shrdir/fakedircreatetimes";
 	push(@dirs,$fakedircreatetimes_shrdir);
 
@@ -3718,6 +3721,15 @@ sub provision($$)
 	recycle : exclude = *.tmp
 	recycle : directory_mode = 755
 
+[recycle2]
+	copy = tmp
+	path = $recycle_shrdir2
+	vfs objects = recycle crossrename
+	recycle : repository = .trash
+	recycle : exclude = *.tmp
+	recycle : directory_mode = 755
+	wide links = yes
+
 [fakedircreatetimes]
 	copy = tmp
 	path = $fakedircreatetimes_shrdir

source3/modules/vfs_crossrename.c

@@ -54,10 +54,12 @@ static NTSTATUS copy_reg(vfs_handle_struct *handle,
 			 struct files_struct *dstfsp,
 			 const struct smb_filename *dest)
 {
-	NTSTATUS status;
-	struct smb_filename *full_fname_src = NULL;
-	struct smb_filename *full_fname_dst = NULL;
+	NTSTATUS status = NT_STATUS_OK;
 	int ret;
+	off_t off;
+	int ifd = -1;
+	int ofd = -1;
+	struct timespec ts[2];
 
 	if (!VALID_STAT(source->st)) {
 		status = NT_STATUS_OBJECT_PATH_NOT_FOUND;
@@ -79,64 +81,105 @@ static NTSTATUS copy_reg(vfs_handle_struct *handle,
 		goto out;
 	}
 
-	full_fname_src = full_path_from_dirfsp_atname(talloc_tos(),
-						      srcfsp,
-						      source);
-	if (full_fname_src == NULL) {
-		status = NT_STATUS_NO_MEMORY;
-		goto out;
-	}
-	full_fname_dst = full_path_from_dirfsp_atname(talloc_tos(),
-						      dstfsp,
-						      dest);
-	if (full_fname_dst == NULL) {
-		status = NT_STATUS_NO_MEMORY;
-		goto out;
-	}
-
 	ret = SMB_VFS_NEXT_UNLINKAT(handle,
 				    dstfsp,
 				    dest,
 				    0);
-	if (ret == -1) {
+	if (ret == -1 && errno != ENOENT) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+
+	ifd = openat(fsp_get_pathref_fd(srcfsp),
+		     source->base_name,
+		     O_RDONLY,
+		     0);
+	if (ifd < 0) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+
+	ofd = openat(fsp_get_pathref_fd(dstfsp),
+		     dest->base_name,
+		     O_WRONLY | O_CREAT | O_TRUNC | O_NOFOLLOW,
+		     0600);
+	if (ofd < 0) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+
+	off = transfer_file(ifd, ofd, source->st.st_ex_size);
+	if (off == -1) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+
+	ret = fchown(ofd, source->st.st_ex_uid, source->st.st_ex_gid);
+	if (ret == -1 && errno != EPERM) {
 		status = map_nt_error_from_unix(errno);
 		goto out;
 	}
 
 	/*
-	 * copy_internals() takes attribute values from the NTrename call.
-	 *
-	 * From MS-CIFS:
-	 *
-	 * "If the attribute is 0x0000, then only normal files are renamed.
-	 * If the system file or hidden attributes are specified, then the
-	 * rename is inclusive of both special types."
+	 * fchown turns off set[ug]id bits for non-root,
+	 * so do the chmod last.
 	 */
-	status = copy_internals(talloc_tos(),
-				handle->conn,
-				NULL,
-				srcfsp, /* src_dirfsp */
-				full_fname_src,
-				dstfsp, /* dst_dirfsp */
-				full_fname_dst,
-				FILE_ATTRIBUTE_HIDDEN | FILE_ATTRIBUTE_SYSTEM);
-	if (!NT_STATUS_IS_OK(status)) {
-		goto out;
-	}
-
-	ret = SMB_VFS_NEXT_UNLINKAT(handle,
-				    srcfsp,
-				    source,
-				    0);
-	if (ret == -1) {
+	ret = fchmod(ofd, source->st.st_ex_mode & 07777);
+	if (ret == -1 && errno != EPERM) {
 		status = map_nt_error_from_unix(errno);
 		goto out;
 	}
 
-  out:
+	/* Try to copy the old file's modtime and access time.  */
+	ts[0] = source->st.st_ex_atime;
+	ts[1] = source->st.st_ex_mtime;
+	ret = futimens(ofd, ts);
+	if (ret == -1) {
+		DBG_DEBUG("Updating the time stamp on destinaton '%s' failed "
+			  "with '%s'. Rename operation can continue.\n",
+			  dest->base_name,
+			  strerror(errno));
+	}
+
+	ret = close(ifd);
+	if (ret == -1) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+	ifd = -1;
+
+	ret = close(ofd);
+	if (ret == -1) {
+		status = map_nt_error_from_unix(errno);
+		goto out;
+	}
+	ofd = -1;
+
+	ret = SMB_VFS_NEXT_UNLINKAT(handle, srcfsp, source, 0);
+	if (ret == -1) {
+		status = map_nt_error_from_unix(errno);
+	}
+
+out:
+	if (ifd != -1) {
+		ret = close(ifd);
+		if (ret == -1) {
+			DBG_DEBUG("Failed to close %s (%d): %s.\n",
+				  source->base_name,
+				  ifd,
+				  strerror(errno));
+		}
+	}
+	if (ofd != -1) {
+		ret = close(ofd);
+		if (ret == -1) {
+			DBG_DEBUG("Failed to close %s (%d): %s.\n",
+				  dest->base_name,
+				  ofd,
+				  strerror(errno));
+		}
+	}
 
-	TALLOC_FREE(full_fname_src);
-	TALLOC_FREE(full_fname_dst);
 	return status;
 }
 
@@ -170,6 +213,7 @@ static int crossrename_renameat(vfs_handle_struct *handle,
 					   smb_fname_src,
 					   dstfsp,
 					   smb_fname_dst);
+		result = 0;
 		if (!NT_STATUS_IS_OK(status)) {
 			errno = map_errno_from_nt_status(status);
 			result = -1;

source3/script/tests/test_recycle.sh

@@ -29,7 +29,8 @@ export SAMBA_DEPRECATED_SUPPRESS
 
 # Define the test environment/filenames.
 #
-share_test_dir="$LOCAL_PATH"
+share_test_dir="$LOCAL_PATH/recycle"
+share_test_dir2="$LOCAL_PATH/recycle2"
 
 #
 # Cleanup function.
@@ -43,6 +44,13 @@ do_cleanup()
 		rm -f testfile2.tmp
 		rm -rf .trash
 	)
+	(
+		#subshell.
+		cd "$share_test_dir2" || return
+		rm -f testfile3
+		rm -f testfile4.tmp
+		rm -rf .trash
+	)
 }
 
 #
@@ -50,6 +58,25 @@ do_cleanup()
 #
 do_cleanup
 
+# Setup .trash on a different filesystem to test crossrename
+# /tmp or /dev/shm should provide tmpfs
+#
+for T in /tmp /dev/shm
+do
+	if df --portability --print-type $T 2>/dev/null | grep -q tmpfs; then
+		TRASHDIR=$T
+		break
+	fi
+done
+
+if [ -z $TRASHDIR ]; then
+	echo "No tmpfs filesystem found."
+	exit 1
+fi
+
+TRASHDIR=$(mktemp -d /$TRASHDIR/.trash_XXXXXX)
+chmod 0755 $TRASHDIR
+ln -s $TRASHDIR $share_test_dir2/.trash
 
 test_recycle()
 {
@@ -90,12 +117,61 @@ quit
 	return 0
 }
 
+test_recycle_crossrename()
+{
+	tmpfile=$PREFIX/smbclient_interactive_prompt_commands
+	echo "
+put $tmpfile testfile3
+put $tmpfile testfile4.tmp
+del testfile3
+del testfile4.tmp
+quit
+" > $tmpfile
+	cmd='CLI_FORCE_INTERACTIVE=yes $SMBCLIENT -U$USERNAME%$PASSWORD //$SERVER/recycle2 -I$SERVER_IP $ADDARGS < $tmpfile 2>&1'
+	eval echo "$cmd"
+	out=$(eval "$cmd")
+	ret=$?
+	rm -f "$tmpfile"
+
+	if [ $ret != 0 ]; then
+		printf "%s\n" "$out"
+		printf "failed recycle smbclient run with error %s\n" "$ret"
+		return 1
+	fi
+
+	test -e "$share_test_dir2/.trash/testfile3" || {
+		printf ".trash/testfile3 expected to exist but does NOT exist\n"
+		return 1
+	}
+	test -e "$share_test_dir2/.trash/testfile4.tmp" && {
+		printf ".trash/testfile4.tmp not expected to exist but DOES exist\n"
+		return 1
+	}
+	deviceid1=`stat -c '%d' "$share_test_dir2/"`
+	deviceid2=`stat -c '%d' "$share_test_dir2/.trash/"`
+	test "$deviceid1=" != "$deviceid2" || {
+		printf ".trash/ should be on a different filesystem!\n"
+		return 1
+	}
+	perm_want=755
+	perm_is=`stat -c '%a' "$share_test_dir2/.trash/"`
+	test "$perm_is" = "$perm_want" || {
+		printf ".trash/ permission should be $perm_want but is $perm_is\n"
+		return 1
+	}
+	return 0
+}
+
 panic_count_0=$(grep -c PANIC $SMBD_TEST_LOG)
 
 testit "recycle" \
 	test_recycle ||
 	failed=$((failed + 1))
 
+testit "recycle_crossrename" \
+	test_recycle_crossrename ||
+	failed=$((failed + 1))
+
 panic_count_1=$(grep -c PANIC $SMBD_TEST_LOG)
 
 testit "check_panic" test $panic_count_0 -eq $panic_count_1 || failed=$(expr $failed + 1)
@@ -103,5 +179,7 @@ testit "check_panic" test $panic_count_0 -eq $panic_count_1 || failed=$(expr $fa
 #
 # Cleanup.
 do_cleanup
+# Cleanup above only deletes a symlink, delete also /tmp/.trash_XXXXXX dir
+rm -rf "$TRASHDIR"
 
 testok "$0" "$failed"

source3/selftest/tests.py

@@ -784,7 +784,7 @@ for env in ["fileserver"]:
     plantestsuite("samba3.blackbox.force_create_mode", env, [os.path.join(samba3srcdir, "script/tests/test_force_create_mode.sh"), '$SERVER', '$DOMAIN', '$USERNAME', '$PASSWORD', '$PREFIX', env, smbclient3])
     plantestsuite("samba3.blackbox.dropbox", env, [os.path.join(samba3srcdir, "script/tests/test_dropbox.sh"), '$SERVER', '$DOMAIN', 'gooduser', '$PASSWORD', '$PREFIX', env, smbclient3])
     plantestsuite("samba3.blackbox.offline", env, [os.path.join(samba3srcdir, "script/tests/test_offline.sh"), '$SERVER', '$SERVER_IP', '$DOMAIN', '$USERNAME', '$PASSWORD', '$LOCAL_PATH/offline', smbclient3])
-    plantestsuite("samba3.blackbox.recycle", env, [os.path.join(samba3srcdir, "script/tests/test_recycle.sh"), '$SERVER', '$SERVER_IP', '$USERNAME', '$PASSWORD', '$LOCAL_PATH/recycle', '$PREFIX', smbclient3])
+    plantestsuite("samba3.blackbox.recycle", env, [os.path.join(samba3srcdir, "script/tests/test_recycle.sh"), '$SERVER', '$SERVER_IP', '$USERNAME', '$PASSWORD', '$LOCAL_PATH', '$PREFIX', smbclient3])
     plantestsuite("samba3.blackbox.fakedircreatetimes", env, [os.path.join(samba3srcdir, "script/tests/test_fakedircreatetimes.sh"), '$SERVER', '$SERVER_IP', '$USERNAME', '$PASSWORD', '$LOCAL_PATH/fakedircreatetimes', '$PREFIX', smbclient3])
     plantestsuite("samba3.blackbox.shadow_copy2.NT1", env + "_smb1_done", [os.path.join(samba3srcdir, "script/tests/test_shadow_copy.sh"), '$SERVER', '$SERVER_IP', '$DOMAIN', '$USERNAME', '$PASSWORD', '$LOCAL_PATH/shadow', smbclient3, '-m', 'NT1'])
     plantestsuite("samba3.blackbox.shadow_copy2.SMB3", env, [os.path.join(samba3srcdir, "script/tests/test_shadow_copy.sh"), '$SERVER', '$SERVER_IP', '$DOMAIN', '$USERNAME', '$PASSWORD', '$LOCAL_PATH/shadow', smbclient3, '-m', 'SMB3'])

source3/smbd/open.c

@@ -4749,7 +4749,7 @@ mkdir_first:
 			       &rhow);
 	if (ret == -1 && errno == EINVAL) {
 		/*
-		 * This is the strategie we use without having
+		 * This is the strategy we use without having
 		 * renameat2(RENAME_NOREPLACE):
 		 *
 		 * renameat() is able to replace a directory if the source is