Merge branch 'ctdb-merge' into v3-2-test

(This used to be commit 175662b5bbb5c8ecd81a60539f5846a938c26a3a)

33 files changed, 3033 insertions, 276 deletions

diff --git a/source3/Makefile.in b/source3/Makefile.in
index 46f733c0bc..265d14547c 100644
--- a/source3/Makefile.in
+++ b/source3/Makefile.in
@@ -308,7 +308,7 @@ LIB_WITH_PROTO_OBJ = $(VERSION_OBJ) lib/charcnv.o lib/debug.o lib/fault.o \
 	  lib/util_str.o lib/clobber.o lib/util_sid.o lib/util_uuid.o \
 	  lib/util_unistr.o lib/util_file.o lib/data_blob.o \
 	  lib/util.o lib/util_sock.o lib/sock_exec.o lib/util_sec.o \
-	  lib/substitute.o lib/fsusage.o \
+	  lib/substitute.o lib/fsusage.o lib/dbwrap_util.o \
 	  lib/ms_fnmatch.o lib/select.o lib/errmap_unix.o \
 	  lib/tallocmsg.o lib/dmallocmsg.o libsmb/smb_signing.o \
 	  lib/md5.o lib/hmacmd5.o lib/arc4.o lib/iconv.o \
@@ -503,6 +503,7 @@ VFS_READONLY_OBJ = modules/vfs_readonly.o modules/getdate.o
 VFS_CAP_OBJ = modules/vfs_cap.o
 VFS_EXPAND_MSDFS_OBJ = modules/vfs_expand_msdfs.o
 VFS_SHADOW_COPY_OBJ = modules/vfs_shadow_copy.o
+VFS_SHADOW_COPY2_OBJ = modules/vfs_shadow_copy2.o
 VFS_AFSACL_OBJ = modules/vfs_afsacl.o
 VFS_XATTR_TDB_OBJ = modules/vfs_xattr_tdb.o librpc/gen_ndr/ndr_xattr.o
 VFS_POSIXACL_OBJ = modules/vfs_posixacl.o
@@ -520,6 +521,7 @@ VFS_COMMIT_OBJ = modules/vfs_commit.o
 VFS_GPFS_OBJ = modules/vfs_gpfs.o modules/gpfs.o modules/nfs4_acls.o
 VFS_NOTIFY_FAM_OBJ = modules/vfs_notify_fam.o
 VFS_READAHEAD_OBJ = modules/vfs_readahead.o
+VFS_TSMSM_OBJ = modules/vfs_tsmsm.o
 VFS_FILEID_OBJ = modules/vfs_fileid.o
 VFS_SYNCOPS_OBJ = modules/vfs_syncops.o
 
@@ -1611,6 +1613,10 @@ bin/ad.@SHLIBEXT@: $(BINARY_PREREQS) winbindd/idmap_ad.o
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) winbindd/idmap_ad.o
 
+bin/tdb2.@SHLIBEXT@: $(BINARY_PREREQS) winbindd/idmap_tdb2.o
+	@echo "Building plugin $@"
+	@$(SHLD_MODULE) winbindd/idmap_tdb2.o
+
 bin/ldap.@SHLIBEXT@: $(BINARY_PREREQS) winbindd/idmap_ldap.o
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) winbindd/idmap_ldap.o
@@ -1670,6 +1676,10 @@ bin/shadow_copy.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_SHADOW_COPY_OBJ)
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) $(VFS_SHADOW_COPY_OBJ)
 
+bin/shadow_copy2.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_SHADOW_COPY2_OBJ)
+	@echo "Building plugin $@"
+	@$(SHLD_MODULE) $(VFS_SHADOW_COPY2_OBJ)
+
 bin/syncops.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_SYNCOPS_OBJ)
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) $(VFS_SYNCOPS_OBJ)
@@ -1750,6 +1760,10 @@ bin/readahead.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_READAHEAD_OBJ)
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) $(VFS_READAHEAD_OBJ)
 
+bin/tsmsm.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_TSMSM_OBJ)
+	@echo "Building plugin $@"
+	@$(SHLD_MODULE) $(VFS_TSMSM_OBJ)
+
 bin/fileid.@SHLIBEXT@: $(BINARY_PREREQS) $(VFS_FILEID_OBJ)
 	@echo "Building plugin $@"
 	@$(SHLD_MODULE) $(VFS_FILEID_OBJ)
diff --git a/source3/configure.in b/source3/configure.in
index cd04b598c0..f65eb3c204 100644
--- a/source3/configure.in
+++ b/source3/configure.in
@@ -707,7 +707,7 @@ dnl These have to be built static:
 default_static_modules="pdb_smbpasswd pdb_tdbsam rpc_lsa rpc_samr rpc_winreg rpc_initshutdown rpc_lsa_ds rpc_wkssvc rpc_svcctl2 rpc_ntsvcs rpc_net rpc_netdfs rpc_srvsvc2 rpc_spoolss rpc_eventlog2 auth_sam auth_unix auth_winbind auth_server auth_domain auth_builtin vfs_default nss_info_template"
 
 dnl These are preferably build shared, and static if dlopen() is not available
-default_shared_modules="vfs_recycle vfs_audit vfs_extd_audit vfs_full_audit vfs_netatalk vfs_fake_perms vfs_default_quota vfs_readonly vfs_cap vfs_expand_msdfs vfs_shadow_copy charset_CP850 charset_CP437 auth_script vfs_readahead vfs_syncops vfs_xattr_tdb"
+default_shared_modules="vfs_recycle vfs_audit vfs_extd_audit vfs_full_audit vfs_netatalk vfs_fake_perms vfs_default_quota vfs_readonly vfs_cap vfs_expand_msdfs vfs_shadow_copy vfs_shadow_copy2 charset_CP850 charset_CP437 auth_script vfs_readahead vfs_syncops vfs_xattr_tdb"
 
 if test "x$developer" = xyes; then
    default_static_modules="$default_static_modules rpc_rpcecho"
@@ -2751,6 +2751,12 @@ AC_SUBST(SMB_FAM_LIBS)
 
 SMB_CHECK_DMAPI([], AC_MSG_NOTICE(DMAPI support not present) )
 
+# Add TSM SM VFS module only if there are both GPFS and DMAPI support
+# Theoretically it should work with AIX JFS2 too but this needs testing
+if test x"$samba_cv_HAVE_GPFS" = x"yes" && test x"$samba_dmapi_libs" != x"" ; then
+    default_shared_modules="$default_shared_modules vfs_tsmsm"
+fi
+
 AC_CACHE_CHECK([for kernel share modes],samba_cv_HAVE_KERNEL_SHARE_MODES,[
 AC_TRY_RUN([
 #include <sys/types.h>
@@ -6452,6 +6458,7 @@ SMB_SUBSYSTEM(RPC,smbd/server.o)
 
 SMB_MODULE(idmap_ldap, winbindd/idmap_ldap.o, "bin/ldap.$SHLIBEXT", IDMAP)
 SMB_MODULE(idmap_tdb, winbindd/idmap_tdb.o, "bin/tdb.$SHLIBEXT", IDMAP)
+SMB_MODULE(idmap_tdb2, winbindd/idmap_tdb2.o, "bin/tdb2.$SHLIBEXT", IDMAP)
 SMB_MODULE(idmap_passdb, winbindd/idmap_passdb.o, "bin/passdb.$SHLIBEXT", IDMAP)
 SMB_MODULE(idmap_nss, winbindd/idmap_nss.o, "bin/nss.$SHLIBEXT", IDMAP)
 SMB_MODULE(idmap_rid, winbindd/idmap_rid.o, "bin/rid.$SHLIBEXT", IDMAP)
@@ -6488,6 +6495,7 @@ SMB_MODULE(vfs_readonly, \$(VFS_READONLY_OBJ), "bin/readonly.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_cap, \$(VFS_CAP_OBJ), "bin/cap.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_expand_msdfs, \$(VFS_EXPAND_MSDFS_OBJ), "bin/expand_msdfs.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_shadow_copy, \$(VFS_SHADOW_COPY_OBJ), "bin/shadow_copy.$SHLIBEXT", VFS)
+SMB_MODULE(vfs_shadow_copy2, \$(VFS_SHADOW_COPY2_OBJ), "bin/shadow_copy2.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_afsacl, \$(VFS_AFSACL_OBJ), "bin/afsacl.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_xattr_tdb, \$(VFS_XATTR_TDB_OBJ), "bin/xattr_tdb.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_posixacl, \$(VFS_POSIXACL_OBJ), "bin/posixacl.$SHLIBEXT", VFS)
@@ -6503,6 +6511,7 @@ SMB_MODULE(vfs_prealloc, \$(VFS_PREALLOC_OBJ), "bin/prealloc.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_commit, \$(VFS_COMMIT_OBJ), "bin/commit.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_gpfs, \$(VFS_GPFS_OBJ), "bin/gpfs.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_readahead, \$(VFS_READAHEAD_OBJ), "bin/readahead.$SHLIBEXT", VFS)
+SMB_MODULE(vfs_tsmsm, \$(VFS_TSMSM_OBJ), "bin/tsmsm.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_fileid, \$(VFS_FILEID_OBJ), "bin/fileid.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_syncops, \$(VFS_SYNCOPS_OBJ), "bin/syncops.$SHLIBEXT", VFS)
 SMB_MODULE(vfs_zfsacl, \$(VFS_ZFSACL_OBJ), "bin/zfsacl.$SHLIBEXT", VFS)
diff --git a/source3/include/ctdbd_conn.h b/source3/include/ctdbd_conn.h
index 425cc65a00..6e1b2f737a 100644
--- a/source3/include/ctdbd_conn.h
+++ b/source3/include/ctdbd_conn.h
@@ -17,6 +17,9 @@
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
+#ifndef _CTDBD_CONN_H
+#define _CTDBD_CONN_H
+
 struct ctdbd_connection;
 
 NTSTATUS ctdbd_init_connection(TALLOC_CTX *mem_ctx,
@@ -62,3 +65,6 @@ NTSTATUS ctdbd_register_ips(struct ctdbd_connection *conn,
 
 NTSTATUS ctdbd_register_reconfigure(struct ctdbd_connection *conn);
 
+NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data);
+
+#endif /* _CTDBD_CONN_H */
diff --git a/source3/include/dbwrap.h b/source3/include/dbwrap.h
index 3bb378c841..4eb174fef1 100644
--- a/source3/include/dbwrap.h
+++ b/source3/include/dbwrap.h
@@ -43,6 +43,7 @@ struct db_context {
 			     void *private_data);
 	int (*get_seqnum)(struct db_context *db);
 	void *private_data;
+	bool persistent;
 };
 
 struct db_context *db_open(TALLOC_CTX *mem_ctx,
diff --git a/source3/include/vfs.h b/source3/include/vfs.h
index 0be3886227..be3cd91520 100644
--- a/source3/include/vfs.h
+++ b/source3/include/vfs.h
@@ -70,6 +70,7 @@
  * timestamp resolition. JRA. */
 /* Changed to version21 to add chflags operation -- jpeach */
 /* Changed to version22 to add lchown operation -- jra */
+/* Additional change: add operations for offline files and remote storage volume abstraction -- ab*/
 /* Leave at 22 - not yet released. But change set_nt_acl to return an NTSTATUS. jra. */
 /* Leave at 22 - not yet released. Add file_id_create operation. --metze */
 /* Leave at 22 - not yet released. Change all BOOL parameters (int) to bool. jra. */
@@ -257,6 +258,12 @@ typedef enum _vfs_op_type {
 	SMB_VFS_OP_AIO_ERROR,
 	SMB_VFS_OP_AIO_FSYNC,
 	SMB_VFS_OP_AIO_SUSPEND,
+        SMB_VFS_OP_AIO_FORCE,
+
+	/* offline operations */
+	SMB_VFS_OP_IS_OFFLINE,
+	SMB_VFS_OP_SET_OFFLINE,
+	SMB_VFS_OP_IS_REMOTESTORAGE,
 
 	/* This should always be last enum value */
 	
@@ -405,6 +412,12 @@ struct vfs_ops {
 		int (*aio_error_fn)(struct vfs_handle_struct *handle, struct files_struct *fsp, SMB_STRUCT_AIOCB *aiocb);
 		int (*aio_fsync)(struct vfs_handle_struct *handle, struct files_struct *fsp, int op, SMB_STRUCT_AIOCB *aiocb);
 		int (*aio_suspend)(struct vfs_handle_struct *handle, struct files_struct *fsp, const SMB_STRUCT_AIOCB * const aiocb[], int n, const struct timespec *timeout);
+		bool (*aio_force)(struct vfs_handle_struct *handle, struct files_struct *fsp);
+		
+		/* offline operations */
+		int (*is_offline)(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path, SMB_STRUCT_STAT *sbuf, bool *offline);
+		int (*set_offline)(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path);
+		bool (*is_remotestorage)(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path);
 
 	} ops;
 
@@ -526,6 +539,12 @@ struct vfs_ops {
 		struct vfs_handle_struct *aio_error;
 		struct vfs_handle_struct *aio_fsync;
 		struct vfs_handle_struct *aio_suspend;
+		struct vfs_handle_struct *aio_force;
+		
+		/* offline operations */
+		struct vfs_handle_struct *is_offline;
+		struct vfs_handle_struct *set_offline;
+		struct vfs_handle_struct *is_remotestorage;
 	} handles;
 };
 
diff --git a/source3/include/vfs_macros.h b/source3/include/vfs_macros.h
index 9232e94a42..dbba781377 100644
--- a/source3/include/vfs_macros.h
+++ b/source3/include/vfs_macros.h
@@ -138,6 +138,12 @@
 #define SMB_VFS_AIO_ERROR(fsp,aiocb) ((fsp)->conn->vfs.ops.aio_error_fn((fsp)->conn->vfs.handles.aio_error,(fsp),(aiocb)))
 #define SMB_VFS_AIO_FSYNC(fsp,op,aiocb) ((fsp)->conn->vfs.ops.aio_fsync((fsp)->conn->vfs.handles.aio_fsync,(fsp),(op),(aiocb)))
 #define SMB_VFS_AIO_SUSPEND(fsp,aiocb,n,ts) ((fsp)->conn->vfs.ops.aio_suspend((fsp)->conn->vfs.handles.aio_suspend,(fsp),(aiocb),(n),(ts)))
+#define SMB_VFS_AIO_FORCE(fsp) ((fsp)->conn->vfs.ops.aio_force((fsp)->conn->vfs.handles.aio_force,(fsp)))
+
+/* Offline operations */
+#define SMB_VFS_IS_OFFLINE(conn,path,sbuf,offline) ((conn)->vfs.ops.is_offline((conn)->vfs.handles.is_offline,(conn),(path),(sbuf),(offline)))
+#define SMB_VFS_SET_OFFLINE(conn,path) ((conn)->vfs.ops.set_offline((conn)->vfs.handles.set_offline,(conn),(path)))
+#define SMB_VFS_IS_REMOTESTORAGE(conn,path) ((conn)->vfs.ops.is_remotestorage((conn)->vfs.handles.is_remotestorage,(conn),(path)))
 
 /*******************************************************************
  Don't access conn->vfs_opaque.ops directly!!!
@@ -257,6 +263,12 @@
 #define SMB_VFS_OPAQUE_AIO_ERROR(fsp,aiocb) ((fsp)->conn->vfs_opaque.ops.aio_error_fn((fsp)->conn->vfs_opaque.handles.aio_error,(fsp),(aiocb)))
 #define SMB_VFS_OPAQUE_AIO_FSYNC(fsp,op,aiocb) ((fsp)->conn->vfs_opaque.ops.aio_fsync((fsp)->conn->vfs_opaque.handles.aio_fsync,(fsp),(op),(aiocb)))
 #define SMB_VFS_OPAQUE_AIO_SUSPEND(fsp,aiocb,n,ts) ((fsp)->conn->vfs_opaque.ops.aio_suspend((fsp)->conn->vfs_opaque.handles.aio_suspend,(fsp),(aiocb),(n),(ts)))
+#define SMB_VFS_OPAQUE_AIO_FORCE(fsp) ((fsp)->conn->vfs_opaque.ops.aio_force((fsp)->conn->vfs_opaque.handles.aio_force,(fsp)))
+
+/* Offline operations */
+#define SMB_VFS_OPAQUE_IS_OFFLINE(conn,path,sbuf,offline) ((conn)->vfs_opaque.ops.is_offline((conn)->vfs_opaque.handles.is_offline,(conn),(path),(sbuf),(offline)))
+#define SMB_VFS_OPAQUE_SET_OFFLINE(conn,path) ((conn)->vfs_opaque.ops.set_offline((conn)->vfs_opaque.handles.set_offline,(conn),(path)))
+#define SMB_VFS_OPAQUE_IS_REMOTESTORAGE(conn,path) ((conn)->vfs_opaque.ops.is_remotestorage((conn)->vfs_opaque.handles.is_remotestorage,(conn),(path)))
 
 /*******************************************************************
  Don't access handle->vfs_next.ops.* directly!!!
@@ -377,5 +389,11 @@
 #define SMB_VFS_NEXT_AIO_ERROR(handle,fsp,aiocb) ((handle)->vfs_next.ops.aio_error_fn((handle)->vfs_next.handles.aio_error,(fsp),(aiocb)))
 #define SMB_VFS_NEXT_AIO_FSYNC(handle,fsp,op,aiocb) ((handle)->vfs_next.ops.aio_fsync((handle)->vfs_next.handles.aio_fsync,(fsp),(op),(aiocb)))
 #define SMB_VFS_NEXT_AIO_SUSPEND(handle,fsp,aiocb,n,ts) ((handle)->vfs_next.ops.aio_suspend((handle)->vfs_next.handles.aio_suspend,(fsp),(aiocb),(n),(ts)))
+#define SMB_VFS_NEXT_AIO_FORCE(handle,fsp) ((handle)->vfs_next.ops.aio_force((handle)->vfs_next.handles.aio_force,(fsp)))
+
+/* Offline operations */
+#define SMB_VFS_NEXT_IS_OFFLINE(conn,path,sbuf,offline) ((conn)->vfs_next.ops.is_offline((conn)->vfs_next.handles.is_offline,(conn),(path),(sbuf),(offline)))
+#define SMB_VFS_NEXT_SET_OFFLINE(conn,path) ((conn)->vfs_next.ops.set_offline((conn)->vfs_next.handles.set_offline,(conn),(path)))
+#define SMB_VFS_NEXT_IS_REMOTESTORAGE(conn,path) ((conn)->vfs_next.ops.is_remotestorage((conn)->vfs_next.handles.is_remotestorage,(conn),(path)))
 
 #endif /* _VFS_MACROS_H */
diff --git a/source3/lib/conn_tdb.c b/source3/lib/conn_tdb.c
index dd0a354a85..22d85c873d 100644
--- a/source3/lib/conn_tdb.c
+++ b/source3/lib/conn_tdb.c
@@ -34,7 +34,7 @@ static struct db_context *connections_db_ctx(bool rw)
 	}
 	else {
 		db_ctx = db_open(NULL, lock_path("connections.tdb"), 0,
-				 TDB_DEFAULT, O_RDONLY, 0);
+				 TDB_CLEAR_IF_FIRST|TDB_DEFAULT, O_RDONLY, 0);
 	}
 
 	return db_ctx;
diff --git a/source3/lib/ctdbd_conn.c b/source3/lib/ctdbd_conn.c
index 899bbcfcce..18e9879601 100644
--- a/source3/lib/ctdbd_conn.c
+++ b/source3/lib/ctdbd_conn.c
@@ -1203,6 +1203,42 @@ NTSTATUS ctdbd_register_reconfigure(struct ctdbd_connection *conn)
 	return register_with_ctdbd(conn, CTDB_SRVID_RECONFIGURE);
 }
 
+/*
+  persstent store. Used when we update a record in a persistent database
+ */
+NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data)
+{
+       int cstatus=0;
+       struct ctdb_rec_data *rec;
+       TDB_DATA recdata;
+       size_t length;
+       NTSTATUS status;
+
+       length = offsetof(struct ctdb_rec_data, data) + key.dsize + data.dsize;
+
+       rec = (struct ctdb_rec_data *)talloc_size(conn, length);
+       NT_STATUS_HAVE_NO_MEMORY(rec);
+
+       rec->length = length;
+       rec->reqid  = db_id;
+       rec->keylen = key.dsize;
+       rec->datalen= data.dsize;
+       memcpy(&rec->data[0], key.dptr, key.dsize);
+       memcpy(&rec->data[key.dsize], data.dptr, data.dsize);
+
+       recdata.dptr  = (uint8_t *)rec;
+       recdata.dsize = length;
+
+       status = ctdbd_control(conn, CTDB_CURRENT_NODE, 
+                              CTDB_CONTROL_PERSISTENT_STORE, 
+                              0, recdata, NULL, NULL, &cstatus);
+       if (cstatus != 0) {
+               return NT_STATUS_INTERNAL_DB_CORRUPTION;
+       }
+       return status;
+}
+
+
 #else
 
 NTSTATUS ctdbd_init_connection(TALLOC_CTX *mem_ctx,
diff --git a/source3/lib/dbwrap.c b/source3/lib/dbwrap.c
index 4e16d18682..001424a6c0 100644
--- a/source3/lib/dbwrap.c
+++ b/source3/lib/dbwrap.c
@@ -20,7 +20,9 @@
 */
 
 #include "includes.h"
-
+#ifdef CLUSTER_SUPPORT
+#include "ctdb_private.h"
+#endif
 /*
  * Fall back using fetch_locked if no genuine fetch operation is provided
  */
@@ -46,10 +48,16 @@ struct db_context *db_open(TALLOC_CTX *mem_ctx,
 			   int open_flags, mode_t mode)
 {
 	struct db_context *result = NULL;
+#ifdef CLUSTER_SUPPORT
+	const char *sockname = lp_ctdbd_socket();
+#endif
 
 #ifdef CLUSTER_SUPPORT
+	if(!sockname || !*sockname) {
+		sockname = CTDB_PATH;
+	}
 
-	if (lp_clustering()) {
+	if (lp_clustering() && socket_exist(sockname)) {
 		const char *partname;
 		/* ctdb only wants the file part of the name */
 		partname = strrchr(name, '/');
diff --git a/source3/lib/dbwrap_ctdb.c b/source3/lib/dbwrap_ctdb.c
index 90d0b260c5..f497f871d2 100644
--- a/source3/lib/dbwrap_ctdb.c
+++ b/source3/lib/dbwrap_ctdb.c
@@ -18,16 +18,14 @@
 */
 
 #include "includes.h"
-
 #ifdef CLUSTER_SUPPORT
-
 #include "ctdb.h"
 #include "ctdb_private.h"
+#include "ctdbd_conn.h"
 
 struct db_ctdb_ctx {
 	struct tdb_wrap *wtdb;
 	uint32 db_id;
-	struct ctdbd_connection *conn;
 };
 
 struct db_ctdb_rec {
@@ -35,8 +33,6 @@ struct db_ctdb_rec {
 	struct ctdb_ltdb_header header;
 };
 
-static struct ctdbd_connection *db_ctdbd_conn(struct db_ctdb_ctx *ctx);
-
 static NTSTATUS db_ctdb_store(struct db_record *rec, TDB_DATA data, int flag)
 {
 	struct db_ctdb_rec *crec = talloc_get_type_abort(
@@ -60,6 +56,42 @@ static NTSTATUS db_ctdb_store(struct db_record *rec, TDB_DATA data, int flag)
 	return (ret == 0) ? NT_STATUS_OK : NT_STATUS_INTERNAL_DB_CORRUPTION;
 }
 
+
+/* for persistent databases the store is a bit different. We have to
+   ask the ctdb daemon to push the record to all nodes after the
+   store */
+static NTSTATUS db_ctdb_store_persistent(struct db_record *rec, TDB_DATA data, int flag)
+{
+	struct db_ctdb_rec *crec = talloc_get_type_abort(
+		rec->private_data, struct db_ctdb_rec);
+	TDB_DATA cdata;
+	int ret;
+	NTSTATUS status;
+
+	cdata.dsize = sizeof(crec->header) + data.dsize;
+
+	if (!(cdata.dptr = SMB_MALLOC_ARRAY(uint8, cdata.dsize))) {
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	crec->header.rsn++;
+
+	memcpy(cdata.dptr, &crec->header, sizeof(crec->header));
+	memcpy(cdata.dptr + sizeof(crec->header), data.dptr, data.dsize);
+
+	ret = tdb_store(crec->ctdb_ctx->wtdb->tdb, rec->key, cdata, TDB_REPLACE);
+	status = (ret == 0) ? NT_STATUS_OK : NT_STATUS_INTERNAL_DB_CORRUPTION;
+	
+	/* now tell ctdbd to update this record on all other nodes */
+	if (NT_STATUS_IS_OK(status)) {
+		status = ctdbd_persistent_store(messaging_ctdbd_connection(), crec->ctdb_ctx->db_id, rec->key, cdata);
+	}
+
+	SAFE_FREE(cdata.dptr);
+
+	return status;
+}
+
 static NTSTATUS db_ctdb_delete(struct db_record *rec)
 {
 	struct db_ctdb_rec *crec = talloc_get_type_abort(
@@ -110,6 +142,7 @@ static struct db_record *db_ctdb_fetch_locked(struct db_context *db,
 	struct db_ctdb_rec *crec;
 	NTSTATUS status;
 	TDB_DATA ctdb_data;
+	int migrate_attempts = 0;
 
 	if (!(result = talloc(mem_ctx, struct db_record))) {
 		DEBUG(0, ("talloc failed\n"));
@@ -153,7 +186,11 @@ again:
 		return NULL;
 	}
 
-	result->store = db_ctdb_store;
+	if (db->persistent) {
+		result->store = db_ctdb_store_persistent;
+	} else {
+		result->store = db_ctdb_store;
+	}
 	result->delete_rec = db_ctdb_delete;
 	talloc_set_destructor(result, db_ctdb_record_destr);
 
@@ -175,12 +212,14 @@ again:
 		tdb_chainunlock(ctx->wtdb->tdb, key);
 		talloc_set_destructor(result, NULL);
 
+		migrate_attempts += 1;
+
 		DEBUG(10, ("ctdb_data.dptr = %p, dmaster = %u (%u)\n",
 			   ctdb_data.dptr, ctdb_data.dptr ?
 			   ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster : -1,
 			   get_my_vnn()));
 
-		status = ctdbd_migrate(db_ctdbd_conn(ctx), ctx->db_id, key);
+		status = ctdbd_migrate(messaging_ctdbd_connection(),ctx->db_id, key);
 		if (!NT_STATUS_IS_OK(status)) {
 			DEBUG(5, ("ctdb_migrate failed: %s\n",
 				  nt_errstr(status)));
@@ -191,6 +230,11 @@ again:
 		goto again;
 	}
 
+	if (migrate_attempts > 10) {
+		DEBUG(0, ("db_ctdb_fetch_locked needed %d attempts\n",
+			  migrate_attempts));
+	}
+
 	memcpy(&crec->header, ctdb_data.dptr, sizeof(crec->header));
 
 	result->value.dsize = ctdb_data.dsize - sizeof(crec->header);
@@ -226,10 +270,12 @@ static int db_ctdb_fetch(struct db_context *db, TALLOC_CTX *mem_ctx,
 	/*
 	 * See if we have a valid record and we are the dmaster. If so, we can
 	 * take the shortcut and just return it.
+	 * we bypass the dmaster check for persistent databases
 	 */
 	if ((ctdb_data.dptr != NULL) &&
 	    (ctdb_data.dsize >= sizeof(struct ctdb_ltdb_header)) &&
-	    ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster == get_my_vnn()) {
+	    (db->persistent ||
+	     ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster == get_my_vnn())) {
 		/* we are the dmaster - avoid the ctdb protocol op */
 
 		data->dsize = ctdb_data.dsize - sizeof(struct ctdb_ltdb_header);
@@ -254,8 +300,7 @@ static int db_ctdb_fetch(struct db_context *db, TALLOC_CTX *mem_ctx,
 	SAFE_FREE(ctdb_data.dptr);
 
 	/* we weren't able to get it locally - ask ctdb to fetch it for us */
-	status = ctdbd_fetch(db_ctdbd_conn(ctx), ctx->db_id, key, mem_ctx,
-			     data);
+	status = ctdbd_fetch(messaging_ctdbd_connection(),ctx->db_id, key, mem_ctx, data);
 	if (!NT_STATUS_IS_OK(status)) {
 		DEBUG(5, ("ctdbd_fetch failed: %s\n", nt_errstr(status)));
 		return -1;
@@ -283,6 +328,22 @@ static void traverse_callback(TDB_DATA key, TDB_DATA data, void *private_data)
 	talloc_free(tmp_ctx);
 }
 
+static int traverse_persistent_callback(TDB_CONTEXT *tdb, TDB_DATA kbuf, TDB_DATA dbuf,
+					void *private_data)
+{
+	struct traverse_state *state = (struct traverse_state *)private_data;
+	struct db_record *rec;
+	TALLOC_CTX *tmp_ctx = talloc_new(state->db);
+	int ret = 0;
+	/* we have to give them a locked record to prevent races */
+	rec = db_ctdb_fetch_locked(state->db, tmp_ctx, kbuf);
+	if (rec && rec->value.dsize > 0) {
+		ret = state->fn(rec, state->private_data);
+	}
+	talloc_free(tmp_ctx);
+	return ret;
+}
+
 static int db_ctdb_traverse(struct db_context *db,
 			    int (*fn)(struct db_record *rec,
 				      void *private_data),
@@ -296,6 +357,13 @@ static int db_ctdb_traverse(struct db_context *db,
 	state.fn = fn;
 	state.private_data = private_data;
 
+	if (db->persistent) {
+		/* for persistent databases we don't need to do a ctdb traverse,
+		   we can do a faster local traverse */
+		return tdb_traverse(ctx->wtdb->tdb, traverse_persistent_callback, &state);
+	}
+
+
 	ctdbd_traverse(ctx->db_id, traverse_callback, &state);
 	return 0;
 }
@@ -322,6 +390,27 @@ static void traverse_read_callback(TDB_DATA key, TDB_DATA data, void *private_da
 	state->fn(&rec, state->private_data);
 }
 
+static int traverse_persistent_callback_read(TDB_CONTEXT *tdb, TDB_DATA kbuf, TDB_DATA dbuf,
+					void *private_data)
+{
+	struct traverse_state *state = (struct traverse_state *)private_data;
+	struct db_record rec;
+	rec.key = kbuf;
+	rec.value = dbuf;
+	rec.store = db_ctdb_store_deny;
+	rec.delete_rec = db_ctdb_delete_deny;
+	rec.private_data = state->db;
+
+	if (rec.value.dsize <= sizeof(struct ctdb_ltdb_header)) {
+		/* a deleted record */
+		return 0;
+	}
+	rec.value.dsize -= sizeof(struct ctdb_ltdb_header);
+	rec.value.dptr += sizeof(struct ctdb_ltdb_header);
+
+	return state->fn(&rec, state->private_data);
+}
+
 static int db_ctdb_traverse_read(struct db_context *db,
 				 int (*fn)(struct db_record *rec,
 					   void *private_data),
@@ -335,6 +424,12 @@ static int db_ctdb_traverse_read(struct db_context *db,
 	state.fn = fn;
 	state.private_data = private_data;
 
+	if (db->persistent) {
+		/* for persistent databases we don't need to do a ctdb traverse,
+		   we can do a faster local traverse */
+		return tdb_traverse_read(ctx->wtdb->tdb, traverse_persistent_callback_read, &state);
+	}
+
 	ctdbd_traverse(ctx->db_id, traverse_read_callback, &state);
 	return 0;
 }
@@ -346,41 +441,6 @@ static int db_ctdb_get_seqnum(struct db_context *db)
 	return tdb_get_seqnum(ctx->wtdb->tdb);
 }
 
-/*
- * Get the ctdbd connection for a database. If possible, re-use the messaging
- * ctdbd connection
- */
-static struct ctdbd_connection *db_ctdbd_conn(struct db_ctdb_ctx *ctx)
-{
-	struct ctdbd_connection *result;
-
-	result = messaging_ctdbd_connection();
-
-	if (result != NULL) {
-
-		if (ctx->conn == NULL) {
-			/*
-			 * Someone has initialized messaging since we
-			 * initialized our own connection, we don't need it
-			 * anymore.
-			 */
-			TALLOC_FREE(ctx->conn);
-		}
-
-		return result;
-	}
-
-	if (ctx->conn == NULL) {
-		NTSTATUS status;
-		status = ctdbd_init_connection(ctx, &ctx->conn);
-		if (!NT_STATUS_IS_OK(status)) {
-			return NULL;
-		}
-		set_my_vnn(ctdbd_vnn(ctx->conn));
-	}
-
-	return ctx->conn;
-}
 
 struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
 				const char *name,
@@ -390,7 +450,6 @@ struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
 	struct db_context *result;
 	struct db_ctdb_ctx *db_ctdb;
 	char *db_path;
-	NTSTATUS status;
 
 	if (!lp_clustering()) {
 		DEBUG(10, ("Clustering disabled -- no ctdb\n"));
@@ -409,20 +468,15 @@ struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
 		return NULL;
 	}
 
-	db_ctdb->conn = NULL;
-
-	status = ctdbd_db_attach(db_ctdbd_conn(db_ctdb), name,
-				 &db_ctdb->db_id, tdb_flags);
-
-	if (!NT_STATUS_IS_OK(status)) {
-		DEBUG(0, ("ctdbd_db_attach failed for %s: %s\n", name,
-			  nt_errstr(status)));
+	if (!NT_STATUS_IS_OK(ctdbd_db_attach(messaging_ctdbd_connection(),name, &db_ctdb->db_id, tdb_flags))) {
+		DEBUG(0, ("ctdbd_db_attach failed for %s\n", name));
 		TALLOC_FREE(result);
 		return NULL;
 	}
 
-	db_path = ctdbd_dbpath(db_ctdbd_conn(db_ctdb), db_ctdb,
-			       db_ctdb->db_id);
+	db_path = ctdbd_dbpath(messaging_ctdbd_connection(), db_ctdb, db_ctdb->db_id);
+
+	result->persistent = ((tdb_flags & TDB_CLEAR_IF_FIRST) == 0);
 
 	/* only pass through specific flags */
 	tdb_flags &= TDB_SEQNUM;
@@ -447,16 +501,4 @@ struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
 
 	return result;
 }
-
-#else
-
-struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
-				const char *name,
-				int hash_size, int tdb_flags,
-				int open_flags, mode_t mode)
-{
-	DEBUG(0, ("no clustering compiled in\n"));
-	return NULL;
-}
-
 #endif
diff --git a/source3/lib/dbwrap_file.c b/source3/lib/dbwrap_file.c
index 5b41f5941b..e3779de1e4 100644
--- a/source3/lib/dbwrap_file.c
+++ b/source3/lib/dbwrap_file.c
@@ -17,10 +17,6 @@
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
-/*
- * Be aware that this is just sample code that has not seen too much testing
- */
-
 #include "includes.h"
 
 struct db_file_ctx {
@@ -367,6 +363,7 @@ struct db_context *db_open_file(TALLOC_CTX *mem_ctx,
 	result->fetch_locked = db_file_fetch_locked;
 	result->traverse = db_file_traverse;
 	result->traverse_read = db_file_traverse;
+	result->persistent = ((tdb_flags & TDB_CLEAR_IF_FIRST) == 0);
 
 	ctx->locked_record = NULL;
 	if (!(ctx->dirname = talloc_strdup(ctx, name))) {
diff --git a/source3/lib/dbwrap_tdb.c b/source3/lib/dbwrap_tdb.c
index 710e45de6b..e87ceb428f 100644
--- a/source3/lib/dbwrap_tdb.c
+++ b/source3/lib/dbwrap_tdb.c
@@ -31,6 +31,11 @@ static int db_tdb_record_destr(struct db_record* data)
 	struct db_tdb_ctx *ctx =
 		talloc_get_type_abort(data->private_data, struct db_tdb_ctx);
 
+	/* This hex_encode() call allocates memory on data context. By way how current 
+	   __talloc_free() code works, it is OK to allocate in the destructor as 
+	   the children of data will be freed after call to the destructor and this 
+	   new 'child' will be caught and freed correctly.
+	 */
 	DEBUG(10, (DEBUGLEVEL > 10
 		   ? "Unlocking key %s\n" : "Unlocking key %.20s\n",
 		   hex_encode(data, (unsigned char *)data->key.dptr,
@@ -88,8 +93,9 @@ static struct db_record *db_tdb_fetch_locked(struct db_context *db,
 	struct tdb_fetch_locked_state state;
 	int res;
 
-	if (DEBUGLEVEL >= 10) {
-		char *keystr = hex_encode(NULL, key.dptr, key.dsize);
+	/* Do not accidently allocate/deallocate w/o need when debug level is lower than needed */
+	if(DEBUGLEVEL >= 10) {
+		char *keystr = hex_encode(NULL, (unsigned char*)key.dptr, key.dsize);
 		DEBUG(10, (DEBUGLEVEL > 10
 			   ? "Locking key %s\n" : "Locking key %.20s\n",
 			   keystr));
@@ -191,15 +197,9 @@ static NTSTATUS db_tdb_delete(struct db_record *rec)
 {
 	struct db_tdb_ctx *ctx = talloc_get_type_abort(rec->private_data,
 						       struct db_tdb_ctx);
-	int res;
-	
-	res = tdb_delete(ctx->wtdb->tdb, rec->key);
 
-	if (res == 0) {
-		return NT_STATUS_OK;
-	}
-
-	return map_nt_error_from_tdb(tdb_error(ctx->wtdb->tdb));
+	return (tdb_delete(ctx->wtdb->tdb, rec->key) == 0) ?
+		NT_STATUS_OK : NT_STATUS_UNSUCCESSFUL;
 }
 
 struct db_tdb_traverse_ctx {
@@ -318,6 +318,7 @@ struct db_context *db_open_tdb(TALLOC_CTX *mem_ctx,
 	result->traverse = db_tdb_traverse;
 	result->traverse_read = db_tdb_traverse_read;
 	result->get_seqnum = db_tdb_get_seqnum;
+	result->persistent = ((tdb_flags & TDB_CLEAR_IF_FIRST) == 0);
 	return result;
 
  fail:
diff --git a/source3/lib/dbwrap_util.c b/source3/lib/dbwrap_util.c
new file mode 100644
index 0000000000..002d572019
--- /dev/null
+++ b/source3/lib/dbwrap_util.c
@@ -0,0 +1,90 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Utility functions for the dbwrap API
+   Copyright (C) Volker Lendecke 2007
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include "includes.h"
+
+int32_t dbwrap_fetch_int32(struct db_context *db, const char *keystr)
+{
+	TDB_DATA dbuf;
+	int32 ret;
+
+	if (db->fetch(db, NULL, string_term_tdb_data(keystr), &dbuf) != 0) {
+		return -1;
+	}
+
+	if ((dbuf.dptr == NULL) || (dbuf.dsize != sizeof(int32_t))) {
+		TALLOC_FREE(dbuf.dptr);
+		return -1;
+	}
+
+	ret = IVAL(dbuf.dptr, 0);
+	TALLOC_FREE(dbuf.dptr);
+	return ret;
+}
+
+int dbwrap_store_int32(struct db_context *db, const char *keystr, int32_t v)
+{
+	struct db_record *rec;
+	int32 v_store;
+	NTSTATUS status;
+
+	rec = db->fetch_locked(db, NULL, string_term_tdb_data(keystr));
+	if (rec == NULL) {
+		return -1;
+	}
+
+	SIVAL(&v_store, 0, v);
+
+	status = rec->store(rec, make_tdb_data((const uint8 *)&v_store,
+					       sizeof(v_store)),
+			    TDB_REPLACE);
+	TALLOC_FREE(rec);
+	return NT_STATUS_IS_OK(status) ? 0 : -1;
+}
+
+uint32_t dbwrap_change_uint32_atomic(struct db_context *db, const char *keystr,
+				     uint32_t *oldval, uint32_t change_val)
+{
+	struct db_record *rec;
+	uint32 val = -1;
+	TDB_DATA data;
+
+	if (!(rec = db->fetch_locked(db, NULL,
+				     string_term_tdb_data(keystr)))) {
+		return -1;
+	}
+
+	if ((rec->value.dptr != NULL)
+	    && (rec->value.dsize == sizeof(val))) {
+		val = IVAL(rec->value.dptr, 0);
+	}
+
+	val += change_val;
+
+	data.dsize = sizeof(val);
+	data.dptr = (uint8 *)&val;
+
+	rec->store(rec, data, TDB_REPLACE);
+
+	TALLOC_FREE(rec);
+
+	return 0;
+}
+
diff --git a/source3/lib/messages_ctdbd.c b/source3/lib/messages_ctdbd.c
index 6e9b934a75..f1a02e6af9 100644
--- a/source3/lib/messages_ctdbd.c
+++ b/source3/lib/messages_ctdbd.c
@@ -22,6 +22,10 @@
 #ifdef CLUSTER_SUPPORT
 
 #include "librpc/gen_ndr/messaging.h"
+#include "ctdb.h"
+#include "ctdb_private.h"
+#include "ctdbd_conn.h"
+
 
 struct messaging_ctdbd_context {
 	struct ctdbd_connection *conn;
diff --git a/source3/lib/util.c b/source3/lib/util.c
index 0653fc9d3f..bc3eaa8d5e 100644
--- a/source3/lib/util.c
+++ b/source3/lib/util.c
@@ -506,6 +506,19 @@ bool file_exist(const char *fname,SMB_STRUCT_STAT *sbuf)
 }
 
 /*******************************************************************
+ Check if a unix domain socket exists - call vfs_file_exist for samba files.
+********************************************************************/
+
+bool socket_exist(const char *fname)
+{
+	SMB_STRUCT_STAT st;
+	if (sys_stat(fname,&st) != 0) 
+		return(False);
+
+	return S_ISSOCK(st.st_mode);
+}
+
+/*******************************************************************
  Check a files mod time.
 ********************************************************************/
 
diff --git a/source3/locking/brlock.c b/source3/locking/brlock.c
index eb42d081fe..4191871bb1 100644
--- a/source3/locking/brlock.c
+++ b/source3/locking/brlock.c
@@ -41,11 +41,11 @@ static struct db_context *brlock_db;
 
 static void print_lock_struct(unsigned int i, struct lock_struct *pls)
 {
-	DEBUG(10,("[%u]: smbpid = %u, tid = %u, pid = %s, ",
+	DEBUG(10,("[%u]: smbpid = %u, tid = %u, pid = %u, ",
 			i,
 			(unsigned int)pls->context.smbpid,
 			(unsigned int)pls->context.tid,
-			procid_str_static(&pls->context.pid) ));
+			(unsigned int)procid_to_pid(&pls->context.pid) ));
 	
 	DEBUG(10,("start = %.0f, size = %.0f, fnum = %d, %s %s\n",
 		(double)pls->start,
@@ -263,10 +263,9 @@ void brl_init(bool read_only)
 	if (brlock_db) {
 		return;
 	}
-	brlock_db = db_open(NULL, lock_path("brlock.tdb"), 0,
-			    TDB_DEFAULT
-			    |TDB_VOLATILE
-			    |(read_only?0x0:TDB_CLEAR_IF_FIRST),
+	brlock_db = db_open(NULL, lock_path("brlock.tdb"),
+			    lp_open_files_db_hash_size(),
+			    TDB_DEFAULT | TDB_CLEAR_IF_FIRST,
 			    read_only?O_RDONLY:(O_RDWR|O_CREAT), 0644 );
 	if (!brlock_db) {
 		DEBUG(0,("Failed to open byte range locking database %s\n",
@@ -1495,14 +1494,16 @@ static int traverse_fn(struct db_record *rec, void *state)
 		}
 	}
 
-	for ( i=0; i<num_locks; i++) {
-		cb->fn(*key,
-		       locks[i].context.pid,
-		       locks[i].lock_type,
-		       locks[i].lock_flav,
-		       locks[i].start,
-		       locks[i].size,
-		       cb->private_data);
+	if (cb->fn) {
+		for ( i=0; i<num_locks; i++) {
+			cb->fn(*key,
+				locks[i].context.pid,
+				locks[i].lock_type,
+				locks[i].lock_flav,
+				locks[i].start,
+				locks[i].size,
+				cb->private_data);
+		}
 	}
 
 	SAFE_FREE(locks);
diff --git a/source3/modules/gpfs.c b/source3/modules/gpfs.c
index 300e90fa69..590dbac26f 100644
--- a/source3/modules/gpfs.c
+++ b/source3/modules/gpfs.c
@@ -22,9 +22,11 @@
 #ifdef HAVE_GPFS
 
 #include "gpfs_gpl.h"
+#include "vfs_gpfs.h"
 
 static void *libgpfs_handle = NULL;
 static bool gpfs_share_modes;
+static bool gpfs_leases;
 
 static int (*gpfs_set_share_fn)(int fd, unsigned int allow, unsigned int deny);
 static int (*gpfs_set_lease_fn)(int fd, unsigned int leaseType);
@@ -42,7 +44,7 @@ bool set_gpfs_sharemode(files_struct *fsp, uint32 access_mask,
 	if (!gpfs_share_modes) {
 		return True;
 	}
-
+	
 	if (gpfs_set_share_fn == NULL) {
 		return False;
 	}
@@ -88,7 +90,7 @@ int set_gpfs_lease(int fd, int leasetype)
 {
 	int gpfs_type = GPFS_LEASE_NONE;
 
-	if (!gpfs_share_modes) {
+	if (!gpfs_leases) {
 		return True;
 	}
 
@@ -103,6 +105,13 @@ int set_gpfs_lease(int fd, int leasetype)
 	if (leasetype == F_WRLCK) {
 		gpfs_type = GPFS_LEASE_WRITE;
 	}
+	
+	/* we unconditionally set CAP_LEASE, rather than looking for
+	   -1/EACCES as there is a bug in some versions of
+	   libgpfs_gpl.so which results in a leaked fd on /dev/ss0
+	   each time we try this with the wrong capabilities set
+	*/
+	linux_set_lease_capability();
 	return gpfs_set_lease_fn(fd, gpfs_type);
 }
 
@@ -172,11 +181,8 @@ void init_gpfs(void)
 		goto failed;
 	}
 
-	if (lp_parm_bool(-1, "gpfs", "sharemodes", True)) {
-		gpfs_share_modes = True;
-	} else {
-		gpfs_share_modes = False;
-	}
+	gpfs_share_modes = lp_parm_bool(-1, "gpfs", "sharemodes", True);
+	gpfs_leases      = lp_parm_bool(-1, "gpfs", "leases", True);
 
 	return;
 
diff --git a/source3/modules/nfs4_acls.c b/source3/modules/nfs4_acls.c
index 52d3983fff..0c3d010dcd 100644
--- a/source3/modules/nfs4_acls.c
+++ b/source3/modules/nfs4_acls.c
@@ -20,6 +20,9 @@
 #include "includes.h"
 #include "nfs4_acls.h"
 
+#undef DBGC_CLASS
+#define DBGC_CLASS DBGC_ACLS
+
 #define SMBACL4_PARAM_TYPE_NAME "nfs4"
 
 #define SMB_ACE4_INT_MAGIC 0x76F8A967
@@ -352,6 +355,7 @@ typedef struct _smbacl4_vfs_params {
 	enum smbacl4_mode_enum mode;
 	bool do_chown;
 	enum smbacl4_acedup_enum acedup;
+	struct db_context *sid_mapping_table;
 } smbacl4_vfs_params;
 
 /*
@@ -451,8 +455,65 @@ static SMB_ACE4PROP_T *smbacl4_find_equal_special(
 	return NULL;
 }
 
-static int smbacl4_fill_ace4(
+static bool nfs4_map_sid(smbacl4_vfs_params *params, const DOM_SID *src,
+			 DOM_SID *dst)
+{
+	static struct db_context *mapping_db = NULL;
+	TDB_DATA data;
+	
+	if (mapping_db == NULL) {
+		const char *dbname = lp_parm_const_string(
+			-1, SMBACL4_PARAM_TYPE_NAME, "sidmap", NULL);
+		
+		if (dbname == NULL) {
+			DEBUG(10, ("%s:sidmap not defined\n",
+				   SMBACL4_PARAM_TYPE_NAME));
+			return False;
+		}
+		
+		become_root();
+		mapping_db = db_open(NULL, dbname, 0, TDB_DEFAULT,
+				     O_RDONLY, 0600);
+		unbecome_root();
+		
+		if (mapping_db == NULL) {
+			DEBUG(1, ("could not open sidmap: %s\n",
+				  strerror(errno)));
+			return False;
+		}
+	}
+	
+	if (mapping_db->fetch(mapping_db, NULL,
+			      string_term_tdb_data(sid_string_tos(src)),
+			      &data) == -1) {
+		DEBUG(10, ("could not find mapping for SID %s\n",
+			   sid_string_dbg(src)));
+		return False;
+	}
+	
+	if ((data.dptr == NULL) || (data.dsize <= 0)
+	    || (data.dptr[data.dsize-1] != '\0')) {
+		DEBUG(5, ("invalid mapping for SID %s\n",
+			  sid_string_dbg(src)));
+		TALLOC_FREE(data.dptr);
+		return False;
+	}
+	
+	if (!string_to_sid(dst, (char *)data.dptr)) {
+		DEBUG(1, ("invalid mapping %s for SID %s\n",
+			  (char *)data.dptr, sid_string_dbg(src)));
+		TALLOC_FREE(data.dptr);
+		return False;
+	}
+
+	TALLOC_FREE(data.dptr);
+	
+	return True;
+}
+
+static bool smbacl4_fill_ace4(
 	TALLOC_CTX *mem_ctx,
+	const char *filename,
 	smbacl4_vfs_params *params,
 	uid_t ownerUID,
 	gid_t ownerGID,
@@ -460,11 +521,6 @@ static int smbacl4_fill_ace4(
 	SMB_ACE4PROP_T *ace_v4 /* output */
 )
 {
-	const char *dom, *name;
-	enum lsa_SidType type;
-	uid_t uid;
-	gid_t gid;
-
 	DEBUG(10, ("got ace for %s\n", sid_string_dbg(&ace_nt->trustee)));
 
 	memset(ace_v4, 0, sizeof(SMB_ACE4PROP_T));
@@ -485,18 +541,46 @@ static int smbacl4_fill_ace4(
 		ace_v4->who.special_id = SMB_ACE4_WHO_EVERYONE;
 		ace_v4->flags |= SMB_ACE4_ID_SPECIAL;
 	} else {
-		if (!lookup_sid(mem_ctx, &ace_nt->trustee, &dom, &name, &type)) {
-			DEBUG(8, ("Could not find %s' type\n",
-				  sid_string_dbg(&ace_nt->trustee)));
-			errno = EINVAL;
-			return -1;
+		const char *dom, *name;
+		enum lsa_SidType type;
+		uid_t uid;
+		gid_t gid;
+		DOM_SID sid;
+		
+		sid_copy(&sid, &ace_nt->trustee);
+		
+		if (!lookup_sid(mem_ctx, &sid, &dom, &name, &type)) {
+			
+			DOM_SID mapped;
+			
+			if (!nfs4_map_sid(params, &sid, &mapped)) {
+				DEBUG(1, ("nfs4_acls.c: file [%s]: SID %s "
+					  "unknown\n", filename, sid_string_dbg(&sid)));
+				errno = EINVAL;
+				return False;
+			}
+			
+			DEBUG(2, ("nfs4_acls.c: file [%s]: mapped SID %s "
+				  "to %s\n", filename, sid_string_dbg(&sid), sid_string_dbg(&mapped)));
+			
+			if (!lookup_sid(mem_ctx, &mapped, &dom,
+					&name, &type)) {
+				DEBUG(1, ("nfs4_acls.c: file [%s]: SID %s "
+					  "mapped from %s is unknown\n",
+					  filename, sid_string_dbg(&mapped), sid_string_dbg(&sid)));
+				errno = EINVAL;
+				return False;
+			}
+			
+			sid_copy(&sid, &mapped);
 		}
-
+		
 		if (type == SID_NAME_USER) {
-			if (!sid_to_uid(&ace_nt->trustee, &uid)) {
-				DEBUG(2, ("Could not convert %s to uid\n",
-					  sid_string_dbg(&ace_nt->trustee)));
-				return -1;
+			if (!sid_to_uid(&sid, &uid)) {
+				DEBUG(1, ("nfs4_acls.c: file [%s]: could not "
+					  "convert %s to uid\n", filename,
+					  sid_string_dbg(&sid)));
+				return False;
 			}
 
 			if (params->mode==e_special && uid==ownerUID) {
@@ -506,11 +590,13 @@ static int smbacl4_fill_ace4(
 				ace_v4->who.uid = uid;
 			}
 		} else { /* else group? - TODO check it... */
-			if (!sid_to_gid(&ace_nt->trustee, &gid)) {
-				DEBUG(2, ("Could not convert %s to gid\n",
-					  sid_string_dbg(&ace_nt->trustee)));
-				return -1;
+			if (!sid_to_gid(&sid, &gid)) {
+				DEBUG(1, ("nfs4_acls.c: file [%s]: could not "
+					  "convert %s to gid\n", filename,
+					  sid_string_dbg(&sid)));
+				return False;
 			}
+				
 			ace_v4->aceFlags |= SMB_ACE4_IDENTIFIER_GROUP;
 
 			if (params->mode==e_special && gid==ownerGID) {
@@ -522,7 +608,7 @@ static int smbacl4_fill_ace4(
 		}
 	}
 
-	return 0; /* OK */
+	return True; /* OK */
 }
 
 static int smbacl4_MergeIgnoreReject(
@@ -560,6 +646,7 @@ static int smbacl4_MergeIgnoreReject(
 }
 
 static SMB4ACL_T *smbacl4_win2nfs4(
+	const char *filename,
 	SEC_ACL *dacl,
 	smbacl4_vfs_params *pparams,
 	uid_t ownerUID,
@@ -580,9 +667,14 @@ static SMB4ACL_T *smbacl4_win2nfs4(
 		SMB_ACE4PROP_T	ace_v4;
 		bool	addNewACE = True;
 
-		if (smbacl4_fill_ace4(mem_ctx, pparams, ownerUID, ownerGID,
-			dacl->aces + i, &ace_v4))
-			return NULL;
+		if (!smbacl4_fill_ace4(mem_ctx, filename, pparams,
+				       ownerUID, ownerGID,
+				       dacl->aces + i, &ace_v4)) {
+			DEBUG(3, ("Could not fill ace for file %s, SID %s\n",
+				  filename,
+				  sid_string_dbg(&((dacl->aces+i)->trustee))));
+			continue;
+		}
 
 		if (pparams->acedup!=e_dontcare) {
 			if (smbacl4_MergeIgnoreReject(pparams->acedup, acl,
@@ -607,6 +699,7 @@ NTSTATUS smb_set_nt_acl_nfs4(files_struct *fsp,
 	bool	result;
 
 	SMB_STRUCT_STAT sbuf;
+	bool need_chown = False;
 	uid_t newUID = (uid_t)-1;
 	gid_t newGID = (gid_t)-1;
 
@@ -635,25 +728,33 @@ NTSTATUS smb_set_nt_acl_nfs4(files_struct *fsp,
 			return status;
 		}
 		if (((newUID != (uid_t)-1) && (sbuf.st_uid != newUID)) ||
-				((newGID != (gid_t)-1) && (sbuf.st_gid != newGID))) {
-			if(try_chown(fsp->conn, fsp->fsp_name, newUID, newGID)) {
-				DEBUG(3,("chown %s, %u, %u failed. Error = %s.\n",
-					fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID, strerror(errno) ));
-				if (errno == EPERM) {
-					return NT_STATUS_INVALID_OWNER;
+		    ((newGID != (gid_t)-1) && (sbuf.st_gid != newGID))) {
+			need_chown = True;
+		}
+		if (need_chown) {
+			if ((newUID == (uid_t)-1 || newUID == current_user.ut.uid)) {
+				if(try_chown(fsp->conn, fsp->fsp_name, newUID, newGID)) {
+					DEBUG(3,("chown %s, %u, %u failed. Error = %s.\n",
+						 fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID, 
+						 strerror(errno)));
+					return map_nt_error_from_unix(errno);
 				}
-				return map_nt_error_from_unix(errno);
+
+				DEBUG(10,("chown %s, %u, %u succeeded.\n",
+					  fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID));
+				if (smbacl4_GetFileOwner(fsp->conn, fsp->fsp_name, &sbuf))
+					return map_nt_error_from_unix(errno);
+				need_chown = False;
+			} else { /* chown is needed, but _after_ changing acl */
+				sbuf.st_uid = newUID; /* OWNER@ in case of e_special */
+				sbuf.st_gid = newGID; /* GROUP@ in case of e_special */
 			}
-			DEBUG(10,("chown %s, %u, %u succeeded.\n",
-				fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID));
-			if (smbacl4_fGetFileOwner(fsp, &sbuf))
-				return map_nt_error_from_unix(errno);
 		}
 	}
 
 	if ((security_info_sent & DACL_SECURITY_INFORMATION)!=0 && psd->dacl!=NULL)
 	{
-		acl = smbacl4_win2nfs4(psd->dacl, &params, sbuf.st_uid, sbuf.st_gid);
+		acl = smbacl4_win2nfs4(fsp->fsp_name, psd->dacl, &params, sbuf.st_uid, sbuf.st_gid);
 		if (!acl)
 			return map_nt_error_from_unix(errno);
 
@@ -668,6 +769,20 @@ NTSTATUS smb_set_nt_acl_nfs4(files_struct *fsp,
 	} else
 		DEBUG(10, ("no dacl found; security_info_sent = 0x%x\n", security_info_sent));
 
+	/* Any chown pending? */
+	if (need_chown) {
+		DEBUG(3,("chown#2 %s. uid = %u, gid = %u.\n",
+			 fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID));
+		if (try_chown(fsp->conn, fsp->fsp_name, newUID, newGID)) {
+			DEBUG(2,("chown#2 %s, %u, %u failed. Error = %s.\n",
+				 fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID,
+				 strerror(errno)));
+			return map_nt_error_from_unix(errno);
+		}
+		DEBUG(10,("chown#2 %s, %u, %u succeeded.\n",
+			  fsp->fsp_name, (unsigned int)newUID, (unsigned int)newGID));
+	}
+
 	DEBUG(10, ("smb_set_nt_acl_nfs4 succeeded\n"));
 	return NT_STATUS_OK;
 }
diff --git a/source3/modules/vfs_default.c b/source3/modules/vfs_default.c
index e21136ccee..3e78c69a5e 100644
--- a/source3/modules/vfs_default.c
+++ b/source3/modules/vfs_default.c
@@ -1225,6 +1225,38 @@ static int vfswrap_aio_suspend(struct vfs_handle_struct *handle, struct files_st
 	return sys_aio_suspend(aiocb, n, timeout);
 }
 
+static int vfswrap_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
+{
+	return False;
+}
+
+static int vfswrap_is_offline(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path, SMB_STRUCT_STAT *sbuf, bool *offline)
+{
+	if (ISDOT(path) || ISDOTDOT(path)) {
+		return 1;
+	}
+	
+	if (!lp_dmapi_support(SNUM(conn)) || !dmapi_have_session()) {
+		return -1;
+	}
+	
+	*offline = (dmapi_file_flags(path) & FILE_ATTRIBUTE_OFFLINE) != 0;
+	return 0;
+}
+
+static int vfswrap_set_offline(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path)
+{
+	/* We don't know how to set offline bit by default, needs to be overriden in the vfs modules */
+	return -1;
+}
+
+static bool vfswrap_is_remotestorage(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path)
+{
+	/* We don't know how to detect that volume is remote storage. VFS modules should redefine it. */
+	return False;
+}
+
+
 static vfs_op_tuple vfs_default_ops[] = {
 
 	/* Disk operations */
@@ -1442,6 +1474,16 @@ static vfs_op_tuple vfs_default_ops[] = {
 	{SMB_VFS_OP(vfswrap_aio_suspend),SMB_VFS_OP_AIO_SUSPEND,
 	 SMB_VFS_LAYER_OPAQUE},
 
+	{SMB_VFS_OP(vfswrap_aio_force), SMB_VFS_OP_AIO_FORCE,
+	 SMB_VFS_LAYER_OPAQUE},
+	
+	{SMB_VFS_OP(vfswrap_is_offline),SMB_VFS_OP_IS_OFFLINE,
+	 SMB_VFS_LAYER_OPAQUE},
+	{SMB_VFS_OP(vfswrap_set_offline),SMB_VFS_OP_SET_OFFLINE,
+	 SMB_VFS_LAYER_OPAQUE},
+	{SMB_VFS_OP(vfswrap_is_remotestorage),SMB_VFS_OP_IS_REMOTESTORAGE,
+	 SMB_VFS_LAYER_OPAQUE},
+	
 	/* Finish VFS operations definition */
 
 	{SMB_VFS_OP(NULL),		SMB_VFS_OP_NOOP,
diff --git a/source3/modules/vfs_gpfs.c b/source3/modules/vfs_gpfs.c
index bcf61f3bc7..d10906dfb1 100644
--- a/source3/modules/vfs_gpfs.c
+++ b/source3/modules/vfs_gpfs.c
@@ -30,7 +30,7 @@
 
 #include <gpfs_gpl.h>
 #include "nfs4_acls.h"
-
+#include "vfs_gpfs.h"
 
 static int vfs_gpfs_kernel_flock(vfs_handle_struct *handle, files_struct *fsp, 
 				 uint32 share_mode)
@@ -153,7 +153,7 @@ static int gpfs_get_nfs4_acl(const char *fname, SMB4ACL_T **ppacl)
 	DEBUG(10, ("gpfs_get_nfs4_acl invoked for %s\n", fname));
 
 	/* First get the real acl length */
-	gacl = gpfs_getacl_alloc(fname, GPFS_ACL_TYPE_NFS4);
+	gacl = gpfs_getacl_alloc(fname, 0);
 	if (gacl == NULL) {
 		DEBUG(9, ("gpfs_getacl failed for %s with %s\n",
 			   fname, strerror(errno)));
@@ -208,10 +208,10 @@ static int gpfs_get_nfs4_acl(const char *fname, SMB4ACL_T **ppacl)
 		if (i > 0 && gace->aceType == SMB_ACE4_ACCESS_DENIED_ACE_TYPE) {
 			struct gpfs_ace_v4 *prev = &gacl->ace_v4[i-1];
 			if (prev->aceType == SMB_ACE4_ACCESS_ALLOWED_ACE_TYPE &&
-					prev->aceFlags == gace->aceFlags &&
-					prev->aceIFlags == gace->aceIFlags &&
-					(gace->aceMask & prev->aceMask) == 0 &&
-					gace->aceWho == prev->aceWho) {
+			    prev->aceFlags == gace->aceFlags &&
+			    prev->aceIFlags == gace->aceIFlags &&
+			    (gace->aceMask & prev->aceMask) == 0 &&
+			    gace->aceWho == prev->aceWho) {
 				/* its redundent - skip it */
 				continue;
 			}                                                
@@ -256,7 +256,7 @@ static NTSTATUS gpfsacl_get_nt_acl(vfs_handle_struct *handle,
 	int	result;
 
 	*ppdesc = NULL;
-	result = gpfs_get_nfs4_acl(fsp->fsp_name, &pacl);
+	result = gpfs_get_nfs4_acl(name, &pacl);
 
 	if (result == 0)
 		return smb_get_nt_acl_nfs4(handle->conn, name, security_info, ppdesc, pacl);
@@ -301,8 +301,31 @@ static bool gpfsacl_process_smbacl(files_struct *fsp, SMB4ACL_T *smbacl)
 		gace->aceType = aceprop->aceType;
 		gace->aceFlags = aceprop->aceFlags;
 		gace->aceMask = aceprop->aceMask;
+		
+		/*
+		 * GPFS can't distinguish between WRITE and APPEND on
+		 * files, so one being set without the other is an
+		 * error. Sorry for the many ()'s :-)
+		 */
+		
+		if (!fsp->is_directory
+		    &&
+		    ((((gace->aceMask & ACE4_MASK_WRITE) == 0)
+		      && ((gace->aceMask & ACE4_MASK_APPEND) != 0))
+		     ||
+		     (((gace->aceMask & ACE4_MASK_WRITE) != 0)
+		      && ((gace->aceMask & ACE4_MASK_APPEND) == 0)))
+		    &&
+		    lp_parm_bool(fsp->conn->params->service, "gpfs",
+				 "merge_writeappend", True)) {
+			DEBUG(2, ("vfs_gpfs.c: file [%s]: ACE contains "
+				  "WRITE^APPEND, setting WRITE|APPEND\n",
+				  fsp->fsp_name));
+			gace->aceMask |= ACE4_MASK_WRITE|ACE4_MASK_APPEND;
+		}
+		
 		gace->aceIFlags = (aceprop->flags&SMB_ACE4_ID_SPECIAL) ? ACE4_IFLAG_SPECIAL_ID : 0;
-
+		
 		if (aceprop->flags&SMB_ACE4_ID_SPECIAL)
 		{
 			switch(aceprop->who.special_id)
@@ -347,7 +370,7 @@ static NTSTATUS gpfsacl_set_nt_acl_internal(files_struct *fsp, uint32 security_i
 	struct gpfs_acl *acl;
 	NTSTATUS result = NT_STATUS_ACCESS_DENIED;
 
-	acl = gpfs_getacl_alloc(fsp->fsp_name, GPFS_ACL_TYPE_ACCESS);
+	acl = gpfs_getacl_alloc(fsp->fsp_name, 0);
 	if (acl == NULL)
 		return result;
 
@@ -628,75 +651,225 @@ int gpfsacl_sys_acl_delete_def_file(vfs_handle_struct *handle,
 	return -1;
 }
 
+/*
+ * Assumed: mode bits are shiftable and standard
+ * Output: the new aceMask field for an smb nfs4 ace
+ */
+static uint32 gpfsacl_mask_filter(uint32 aceType, uint32 aceMask, uint32 rwx)
+{
+	const uint32 posix_nfs4map[3] = {
+                SMB_ACE4_EXECUTE, /* execute */
+		SMB_ACE4_WRITE_DATA | SMB_ACE4_APPEND_DATA, /* write; GPFS specific */
+                SMB_ACE4_READ_DATA /* read */
+	};
+	int     i;
+	uint32_t        posix_mask = 0x01;
+	uint32_t        posix_bit;
+	uint32_t        nfs4_bits;
+	
+	for(i=0; i<3; i++) {
+		nfs4_bits = posix_nfs4map[i];
+		posix_bit = rwx & posix_mask;
+		
+		if (aceType==SMB_ACE4_ACCESS_ALLOWED_ACE_TYPE) {
+			if (posix_bit)
+				aceMask |= nfs4_bits;
+			else
+				aceMask &= ~nfs4_bits;
+		} else {
+			/* add deny bits when suitable */
+			if (!posix_bit)
+				aceMask |= nfs4_bits;
+			else
+				aceMask &= ~nfs4_bits;
+		} /* other ace types are unexpected */
+		
+		posix_mask <<= 1;
+	}
+	
+	return aceMask;
+}
+
+static int gpfsacl_emu_chmod(const char *path, mode_t mode)
+{
+	SMB4ACL_T *pacl = NULL;
+	int     result;
+	bool    haveAllowEntry[SMB_ACE4_WHO_EVERYONE + 1] = {False, False, False, False};
+	int     i;
+	files_struct    fake_fsp; /* TODO: rationalize parametrization */
+	SMB4ACE_T       *smbace;
+	
+	DEBUG(10, ("gpfsacl_emu_chmod invoked for %s mode %o\n", path, mode));
+	
+	result = gpfs_get_nfs4_acl(path, &pacl);
+	if (result)
+		return result;
+	
+	if (mode & ~(S_IRWXU | S_IRWXG | S_IRWXO)) {
+		DEBUG(2, ("WARNING: cutting extra mode bits %o on %s\n", mode, path));
+	}
+	
+	for (smbace=smb_first_ace4(pacl); smbace!=NULL; smbace = smb_next_ace4(smbace)) {
+		SMB_ACE4PROP_T  *ace = smb_get_ace4(smbace);
+		uint32_t        specid = ace->who.special_id;
+		
+		if (ace->flags&SMB_ACE4_ID_SPECIAL &&
+		    ace->aceType<=SMB_ACE4_ACCESS_DENIED_ACE_TYPE &&
+		    specid <= SMB_ACE4_WHO_EVERYONE) {
+			
+			uint32_t newMask;
+			
+			if (ace->aceType==SMB_ACE4_ACCESS_ALLOWED_ACE_TYPE)
+				haveAllowEntry[specid] = True;
+			
+			/* mode >> 6 for @owner, mode >> 3 for @group,
+			 * mode >> 0 for @everyone */
+			newMask = gpfsacl_mask_filter(ace->aceType, ace->aceMask,
+						      mode >> ((SMB_ACE4_WHO_EVERYONE - specid) * 3));
+			if (ace->aceMask!=newMask) {
+				DEBUG(10, ("ace changed for %s (%o -> %o) id=%d\n",
+					   path, ace->aceMask, newMask, specid));
+			}
+			ace->aceMask = newMask;
+		}
+	}
+
+	/* make sure we have at least ALLOW entries
+	 * for all the 3 special ids (@EVERYONE, @OWNER, @GROUP)
+	 * - if necessary
+	 */
+	for(i = SMB_ACE4_WHO_OWNER; i<=SMB_ACE4_WHO_EVERYONE; i++) {
+		SMB_ACE4PROP_T  ace;
+		
+		if (haveAllowEntry[i]==True)
+			continue;
+		
+		memset(&ace, 0, sizeof(SMB_ACE4PROP_T));
+		ace.aceType = SMB_ACE4_ACCESS_ALLOWED_ACE_TYPE;
+		ace.flags |= SMB_ACE4_ID_SPECIAL;
+		ace.who.special_id = i;
+		
+		if (i==SMB_ACE4_WHO_GROUP) /* not sure it's necessary... */
+			ace.aceFlags |= SMB_ACE4_IDENTIFIER_GROUP;
+		
+		ace.aceMask = gpfsacl_mask_filter(ace.aceType, ace.aceMask,
+						  mode >> ((SMB_ACE4_WHO_EVERYONE - i) * 3));
+		
+		/* don't add unnecessary aces */
+		if (!ace.aceMask)
+			continue;
+		
+		/* we add it to the END - as windows expects allow aces */
+		smb_add_ace4(pacl, &ace);
+		DEBUG(10, ("Added ALLOW ace for %s, mode=%o, id=%d, aceMask=%x\n",
+			   path, mode, i, ace.aceMask));
+	}
+	
+	/* don't add complementary DENY ACEs here */
+	memset(&fake_fsp, 0, sizeof(struct files_struct));
+	fake_fsp.fsp_name = (char *)path; /* no file_new is needed here */
+	
+	/* put the acl */
+	if (gpfsacl_process_smbacl(&fake_fsp, pacl) == False)
+		return -1;
+	return 0; /* ok for [f]chmod */
+}
+
 static int vfs_gpfs_chmod(vfs_handle_struct *handle, const char *path, mode_t mode)
 {
 		 SMB_STRUCT_STAT st;
+		 int rc;
+		 
 		 if (SMB_VFS_NEXT_STAT(handle, path, &st) != 0) {
-		 		 return -1;
+			 return -1;
 		 }
+		 
 		 /* avoid chmod() if possible, to preserve acls */
 		 if ((st.st_mode & ~S_IFMT) == mode) {
-		 		 return 0;
+			 return 0;
 		 }
-		 return SMB_VFS_NEXT_CHMOD(handle, path, mode);
+
+		 rc = gpfsacl_emu_chmod(path, mode);
+		 if (rc == 1)
+			 return SMB_VFS_NEXT_CHMOD(handle, path, mode);
+		 return rc;
 }
 
 static int vfs_gpfs_fchmod(vfs_handle_struct *handle, files_struct *fsp, mode_t mode)
 {
 		 SMB_STRUCT_STAT st;
+		 int rc;
+		 
 		 if (SMB_VFS_NEXT_FSTAT(handle, fsp, &st) != 0) {
-		 		 return -1;
+			 return -1;
 		 }
+
 		 /* avoid chmod() if possible, to preserve acls */
 		 if ((st.st_mode & ~S_IFMT) == mode) {
-		 		 return 0;
+			 return 0;
 		 }
-		 return SMB_VFS_NEXT_FCHMOD(handle, fsp, mode);
+
+		 rc = gpfsacl_emu_chmod(fsp->fsp_name, mode);
+		 if (rc == 1)
+			 return SMB_VFS_NEXT_FCHMOD(handle, fsp, mode);
+		 return rc;
 }
 
 /* VFS operations structure */
 
 static vfs_op_tuple gpfs_op_tuples[] = {
-
-		{ SMB_VFS_OP(vfs_gpfs_kernel_flock), SMB_VFS_OP_KERNEL_FLOCK,
-				SMB_VFS_LAYER_OPAQUE },
-
-        { SMB_VFS_OP(vfs_gpfs_setlease), SMB_VFS_OP_LINUX_SETLEASE,
-        		SMB_VFS_LAYER_OPAQUE },
-
-        { SMB_VFS_OP(gpfsacl_fget_nt_acl), SMB_VFS_OP_FGET_NT_ACL,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_get_nt_acl), SMB_VFS_OP_GET_NT_ACL,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_fset_nt_acl), SMB_VFS_OP_FSET_NT_ACL,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_set_nt_acl), SMB_VFS_OP_SET_NT_ACL,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_sys_acl_get_file), SMB_VFS_OP_SYS_ACL_GET_FILE,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_sys_acl_get_fd), SMB_VFS_OP_SYS_ACL_GET_FD,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_sys_acl_set_file), SMB_VFS_OP_SYS_ACL_SET_FILE,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(gpfsacl_sys_acl_set_fd), SMB_VFS_OP_SYS_ACL_SET_FD,
-        		SMB_VFS_LAYER_TRANSPARENT },
-
+	
+	{ SMB_VFS_OP(vfs_gpfs_kernel_flock), 
+	  SMB_VFS_OP_KERNEL_FLOCK,
+	  SMB_VFS_LAYER_OPAQUE },
+	
+        { SMB_VFS_OP(vfs_gpfs_setlease), 
+	  SMB_VFS_OP_LINUX_SETLEASE,
+	  SMB_VFS_LAYER_OPAQUE },
+	
+        { SMB_VFS_OP(gpfsacl_fget_nt_acl), 
+	  SMB_VFS_OP_FGET_NT_ACL,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_get_nt_acl), 
+	  SMB_VFS_OP_GET_NT_ACL,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_fset_nt_acl), 
+	  SMB_VFS_OP_FSET_NT_ACL,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_set_nt_acl), 
+	  SMB_VFS_OP_SET_NT_ACL,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_sys_acl_get_file), 
+	  SMB_VFS_OP_SYS_ACL_GET_FILE,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_sys_acl_get_fd), 
+	  SMB_VFS_OP_SYS_ACL_GET_FD,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_sys_acl_set_file), 
+	  SMB_VFS_OP_SYS_ACL_SET_FILE,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(gpfsacl_sys_acl_set_fd), 
+	  SMB_VFS_OP_SYS_ACL_SET_FD,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
         { SMB_VFS_OP(gpfsacl_sys_acl_delete_def_file),
-                SMB_VFS_OP_SYS_ACL_DELETE_DEF_FILE,
-                SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(vfs_gpfs_chmod), SMB_VFS_OP_CHMOD,
-                SMB_VFS_LAYER_TRANSPARENT },
-
-        { SMB_VFS_OP(vfs_gpfs_fchmod), SMB_VFS_OP_FCHMOD,
-                SMB_VFS_LAYER_TRANSPARENT },
+	  SMB_VFS_OP_SYS_ACL_DELETE_DEF_FILE,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(vfs_gpfs_chmod), 
+	  SMB_VFS_OP_CHMOD,
+	  SMB_VFS_LAYER_TRANSPARENT },
+	
+        { SMB_VFS_OP(vfs_gpfs_fchmod), 
+	  SMB_VFS_OP_FCHMOD,
+	  SMB_VFS_LAYER_TRANSPARENT },
 
         { SMB_VFS_OP(NULL), SMB_VFS_OP_NOOP, SMB_VFS_LAYER_NOOP }
 
diff --git a/source3/modules/vfs_gpfs.h b/source3/modules/vfs_gpfs.h
new file mode 100644
index 0000000000..3c499b0850
--- /dev/null
+++ b/source3/modules/vfs_gpfs.h
@@ -0,0 +1,32 @@
+/*
+   Unix SMB/CIFS implementation.
+   Wrap gpfs calls in vfs functions.
+ 
+   Copyright (C) Christian Ambach <cambach1@de.ibm.com> 2006
+   
+   Major code contributions by Chetan Shringarpure <chetan.sh@in.ibm.com>
+                            and Gomati Mohanan <gomati.mohanan@in.ibm.com>
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+  
+
+*/
+
+bool set_gpfs_sharemode(files_struct *fsp, uint32 access_mask,
+			uint32 share_access);
+int set_gpfs_lease(int fd, int leasetype);
+int smbd_gpfs_getacl(char *pathname, int flags, void *acl);
+int smbd_gpfs_putacl(char *pathname, int flags, void *acl);
+void init_gpfs(void);
diff --git a/source3/modules/vfs_prealloc.c b/source3/modules/vfs_prealloc.c
index 2d64bc0184..cb3508dc30 100644
--- a/source3/modules/vfs_prealloc.c
+++ b/source3/modules/vfs_prealloc.c
@@ -47,11 +47,16 @@
 #define lock_type struct flock64
 #endif
 
+#ifdef HAVE_GPFS
+#include "gpfs_gpl.h"
+#endif
+
 #define MODULE "prealloc"
 static int module_debug;
 
 static int preallocate_space(int fd, SMB_OFF_T size)
 {
+#ifndef HAVE_GPFS
 	lock_type fl = {0};
 	int err;
 
@@ -78,6 +83,9 @@ static int preallocate_space(int fd, SMB_OFF_T size)
 	err = -1;
 	errno = ENOSYS;
 #endif
+#else /* GPFS uses completely different interface */
+       err = gpfs_prealloc(fd, (gpfs_off64_t)0, (gpfs_off64_t)size);
+#endif
 
 	if (err) {
 		DEBUG(module_debug,
diff --git a/source3/modules/vfs_shadow_copy2.c b/source3/modules/vfs_shadow_copy2.c
new file mode 100644
index 0000000000..54fc672b9c
--- /dev/null
+++ b/source3/modules/vfs_shadow_copy2.c
@@ -0,0 +1,637 @@
+/* 
+ * implementation of an Shadow Copy module - version 2
+ *
+ * Copyright (C) Andrew Tridgell     2007
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *  
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *  
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include "includes.h"
+
+/*
+
+  This is a 2nd implemetation of a shadow copy module for exposing
+  snapshots to windows clients as shadow copies. This version has the
+  following features:
+
+     1) you don't need to populate your shares with symlinks to the
+     snapshots. This can be very important when you have thousands of
+     shares, or use [homes]
+
+     2) the inode number of the files is altered so it is different
+     from the original. This allows the 'restore' button to work
+     without a sharing violation
+
+  Module options:
+
+      shadow:snapdir = <directory where snapshots are kept>
+
+      This is the directory containing the @GMT-* snapshot directories. If it is an absolute
+      path it is used as-is. If it is a relative path, then it is taken relative to the mount
+      point of the filesystem that the root of this share is on
+
+      shadow:basedir = <base directory that snapshots are from>
+
+      This is an optional parameter that specifies the directory that
+      the snapshots are relative to. It defaults to the filesystem
+      mount point
+
+      shadow:fixinodes = yes/no
+
+      If you enable shadow:fixinodes then this module will modify the
+      apparent inode number of files in the snapshot directories using
+      a hash of the files path. This is needed for snapshot systems
+      where the snapshots have the same device:inode number as the
+      original files (such as happens with GPFS snapshots). If you
+      don't set this option then the 'restore' button in the shadow
+      copy UI will fail with a sharing violation.
+
+  Note that the directory names in the snapshot directory must take the form
+  @GMT-YYYY.MM.DD-HH.MM.SS
+  
+  The following command would generate a correctly formatted directory name:
+     date -u +@GMT-%Y.%m.%d-%H.%M.%S
+  
+ */
+
+static int vfs_shadow_copy2_debug_level = DBGC_VFS;
+
+#undef DBGC_CLASS
+#define DBGC_CLASS vfs_shadow_copy2_debug_level
+
+#define GMT_NAME_LEN 24 /* length of a @GMT- name */
+
+/*
+  make very sure it is one of our special names 
+ */
+static inline bool shadow_copy2_match_name(const char *name)
+{
+	unsigned year, month, day, hr, min, sec;
+	if (name[0] != '@') return False;
+	if (strncmp(name, "@GMT-", 5) != 0) return False;
+	if (sscanf(name, "@GMT-%04u.%02u.%02u-%02u.%02u.%02u", &year, &month,
+		   &day, &hr, &min, &sec) != 6) {
+		return False;
+	}
+	if (name[24] != 0 && name[24] != '/') {
+		return False;
+	}
+	return True;
+}
+
+/*
+  convert a name to the shadow directory
+ */
+
+#define _SHADOW2_NEXT(op, args, rtype, eret, extra) do { \
+	const char *name = fname; \
+	if (shadow_copy2_match_name(fname)) { \
+		char *name2; \
+		rtype ret; \
+		name2 = convert_shadow2_name(handle, fname); \
+		if (name2 == NULL) { \
+			errno = EINVAL; \
+			return eret; \
+		} \
+		name = name2; \
+		ret = SMB_VFS_NEXT_ ## op args; \
+		talloc_free(name2); \
+		if (ret != eret) extra; \
+		return ret; \
+	} else { \
+		return SMB_VFS_NEXT_ ## op args; \
+	} \
+} while (0)
+
+/*
+  convert a name to the shadow directory: NTSTATUS-specific handling
+ */
+
+#define _SHADOW2_NTSTATUS_NEXT(op, args, eret, extra) do { \
+        const char *name = fname; \
+        if (shadow_copy2_match_name(fname)) { \
+                char *name2; \
+                NTSTATUS ret; \
+                name2 = convert_shadow2_name(handle, fname); \
+                if (name2 == NULL) { \
+                        errno = EINVAL; \
+                        return eret; \
+                } \
+                name = name2; \
+                ret = SMB_VFS_NEXT_ ## op args; \
+                talloc_free(name2); \
+                if (!NT_STATUS_EQUAL(ret, eret)) extra; \
+                return ret; \
+        } else { \
+                return SMB_VFS_NEXT_ ## op args; \
+        } \
+} while (0)
+
+#define SHADOW2_NTSTATUS_NEXT(op, args, eret) _SHADOW2_NTSTATUS_NEXT(op, args, eret, )
+
+#define SHADOW2_NEXT(op, args, rtype, eret) _SHADOW2_NEXT(op, args, rtype, eret, )
+
+#define SHADOW2_NEXT2(op, args) do { \
+	if (shadow_copy2_match_name(oldname) || shadow_copy2_match_name(newname)) { \
+		errno = EROFS; \
+		return -1; \
+	} else { \
+		return SMB_VFS_NEXT_ ## op args; \
+	} \
+} while (0)
+
+
+/*
+  find the mount point of a filesystem
+ */
+static char *find_mount_point(TALLOC_CTX *mem_ctx, vfs_handle_struct *handle)
+{
+	char *path = talloc_strdup(mem_ctx, handle->conn->connectpath);
+	dev_t dev;
+	struct stat st;
+	char *p;
+
+	if (stat(path, &st) != 0) {
+		talloc_free(path);
+		return NULL;
+	}
+
+	dev = st.st_dev;
+
+	while ((p = strrchr(path, '/')) && p > path) {
+		*p = 0;
+		if (stat(path, &st) != 0) {
+			talloc_free(path);
+			return NULL;
+		}
+		if (st.st_dev != dev) {
+			*p = '/';
+			break;
+		}
+	}
+
+	return path;	
+}
+
+/*
+  work out the location of the snapshot for this share
+ */
+static const char *shadow_copy2_find_snapdir(TALLOC_CTX *mem_ctx, vfs_handle_struct *handle)
+{
+	const char *snapdir;
+	char *mount_point;
+	const char *ret;
+
+	snapdir = lp_parm_const_string(SNUM(handle->conn), "shadow", "snapdir", NULL);
+	if (snapdir == NULL) {
+		return NULL;
+	}
+	/* if its an absolute path, we're done */
+	if (*snapdir == '/') {
+		return snapdir;
+	}
+
+	/* other its relative to the filesystem mount point */
+	mount_point = find_mount_point(mem_ctx, handle);
+	if (mount_point == NULL) {
+		return NULL;
+	}
+
+	ret = talloc_asprintf(mem_ctx, "%s/%s", mount_point, snapdir);
+	talloc_free(mount_point);
+	return ret;
+}
+
+/*
+  work out the location of the base directory for snapshots of this share
+ */
+static const char *shadow_copy2_find_basedir(TALLOC_CTX *mem_ctx, vfs_handle_struct *handle)
+{
+	const char *basedir = lp_parm_const_string(SNUM(handle->conn), "shadow", "basedir", NULL);
+
+	/* other its the filesystem mount point */
+	if (basedir == NULL) {
+		basedir = find_mount_point(mem_ctx, handle);
+	}
+
+	return basedir;
+}
+
+/*
+  convert a filename from a share relative path, to a path in the
+  snapshot directory
+ */
+static char *convert_shadow2_name(vfs_handle_struct *handle, const char *fname)
+{
+	TALLOC_CTX *tmp_ctx = talloc_new(handle->data);
+	const char *snapdir, *relpath, *baseoffset, *basedir;
+	size_t baselen;
+	char *ret;
+
+	snapdir = shadow_copy2_find_snapdir(tmp_ctx, handle);
+	if (snapdir == NULL) {
+		DEBUG(2,("no snapdir found for share at %s\n", handle->conn->connectpath));
+		talloc_free(tmp_ctx);
+		return NULL;
+	}
+
+	basedir = shadow_copy2_find_basedir(tmp_ctx, handle);
+	if (basedir == NULL) {
+		DEBUG(2,("no basedir found for share at %s\n", handle->conn->connectpath));
+		talloc_free(tmp_ctx);
+		return NULL;
+	}
+
+	relpath = fname + GMT_NAME_LEN;
+	baselen = strlen(basedir);
+	baseoffset = handle->conn->connectpath + baselen;
+
+	/* some sanity checks */
+	if (strncmp(basedir, handle->conn->connectpath, baselen) != 0 ||
+	    (handle->conn->connectpath[baselen] != 0 && handle->conn->connectpath[baselen] != '/')) {
+		DEBUG(0,("convert_shadow2_name: basedir %s is not a parent of %s\n",
+			 basedir, handle->conn->connectpath));
+		talloc_free(tmp_ctx);
+		return NULL;
+	}
+
+	if (*relpath == '/') relpath++;
+	if (*baseoffset == '/') baseoffset++;
+
+	ret = talloc_asprintf(handle->data, "%s/%.*s/%s/%s", 
+			      snapdir, 
+			      GMT_NAME_LEN, fname, 
+			      baseoffset, 
+			      relpath);
+	DEBUG(6,("convert_shadow2_name: '%s' -> '%s'\n", fname, ret));
+	talloc_free(tmp_ctx);
+	return ret;
+}
+
+
+/*
+  simple string hash
+ */
+static uint32 string_hash(const char *s)
+{
+        uint32 n = 0;
+	while (*s) {
+                n = ((n << 5) + n) ^ (uint32)(*s++);
+        }
+        return n;
+}
+
+/*
+  modify a sbuf return to ensure that inodes in the shadow directory
+  are different from those in the main directory
+ */
+static void convert_sbuf(vfs_handle_struct *handle, const char *fname, SMB_STRUCT_STAT *sbuf)
+{
+	if (lp_parm_bool(SNUM(handle->conn), "shadow", "fixinodes", False)) {		
+		/* some snapshot systems, like GPFS, return the name
+		   device:inode for the snapshot files as the current
+		   files. That breaks the 'restore' button in the shadow copy
+		   GUI, as the client gets a sharing violation.
+
+		   This is a crude way of allowing both files to be
+		   open at once. It has a slight chance of inode
+		   number collision, but I can't see a better approach
+		   without significant VFS changes
+		*/
+		uint32_t shash = string_hash(fname) & 0xFF000000;
+		if (shash == 0) {
+			shash = 1;
+		}
+		sbuf->st_ino ^= shash;
+	}
+}
+
+static int shadow_copy2_rename(vfs_handle_struct *handle,
+			const char *oldname, const char *newname)
+{
+	SHADOW2_NEXT2(RENAME, (handle, oldname, newname));
+}
+
+static int shadow_copy2_symlink(vfs_handle_struct *handle,
+				const char *oldname, const char *newname)
+{
+	SHADOW2_NEXT2(SYMLINK, (handle, oldname, newname));
+}
+
+static int shadow_copy2_link(vfs_handle_struct *handle,
+			  const char *oldname, const char *newname)
+{
+	SHADOW2_NEXT2(LINK, (handle, oldname, newname));
+}
+
+static int shadow_copy2_open(vfs_handle_struct *handle,
+			     const char *fname, files_struct *fsp, int flags, mode_t mode)
+{
+	SHADOW2_NEXT(OPEN, (handle, name, fsp, flags, mode), int, -1);
+}
+
+static SMB_STRUCT_DIR *shadow_copy2_opendir(vfs_handle_struct *handle,
+			  const char *fname, const char *mask, uint32 attr)
+{
+        SHADOW2_NEXT(OPENDIR, (handle, name, mask, attr), void*, NULL);
+}
+
+static int shadow_copy2_stat(vfs_handle_struct *handle,
+		      const char *fname, SMB_STRUCT_STAT *sbuf)
+{
+        _SHADOW2_NEXT(STAT, (handle, name, sbuf), int, -1, convert_sbuf(handle, fname, sbuf));
+}
+
+static int shadow_copy2_lstat(vfs_handle_struct *handle,
+		       const char *fname, SMB_STRUCT_STAT *sbuf)
+{
+        _SHADOW2_NEXT(LSTAT, (handle, name, sbuf), int, -1, convert_sbuf(handle, fname, sbuf));
+}
+
+static int shadow_copy2_fstat(vfs_handle_struct *handle, files_struct *fsp, SMB_STRUCT_STAT *sbuf)
+{
+	int ret = SMB_VFS_NEXT_FSTAT(handle, fsp, sbuf);
+	if (ret == 0 && shadow_copy2_match_name(fsp->fsp_name)) {
+		convert_sbuf(handle, fsp->fsp_name, sbuf);
+	}
+	return ret;
+}
+
+static int shadow_copy2_unlink(vfs_handle_struct *handle, const char *fname)
+{
+        SHADOW2_NEXT(UNLINK, (handle, name), int, -1);
+}
+
+static int shadow_copy2_chmod(vfs_handle_struct *handle,
+		       const char *fname, mode_t mode)
+{
+        SHADOW2_NEXT(CHMOD, (handle, name, mode), int, -1);
+}
+
+static int shadow_copy2_chown(vfs_handle_struct *handle,
+		       const char *fname, uid_t uid, gid_t gid)
+{
+        SHADOW2_NEXT(CHOWN, (handle, name, uid, gid), int, -1);
+}
+
+static int shadow_copy2_chdir(vfs_handle_struct *handle,
+		       const char *fname)
+{
+	SHADOW2_NEXT(CHDIR, (handle, name), int, -1);
+}
+
+static int shadow_copy2_ntimes(vfs_handle_struct *handle,
+		       const char *fname, const struct timespec ts[2])
+{
+        SHADOW2_NEXT(NTIMES, (handle, name, ts), int, -1);
+}
+
+static int shadow_copy2_readlink(vfs_handle_struct *handle,
+				 const char *fname, char *buf, size_t bufsiz)
+{
+        SHADOW2_NEXT(READLINK, (handle, name, buf, bufsiz), int, -1);
+}
+
+static int shadow_copy2_mknod(vfs_handle_struct *handle,
+		       const char *fname, mode_t mode, SMB_DEV_T dev)
+{
+        SHADOW2_NEXT(MKNOD, (handle, name, mode, dev), int, -1);
+}
+
+static char *shadow_copy2_realpath(vfs_handle_struct *handle,
+			    const char *fname, char *resolved_path)
+{
+        SHADOW2_NEXT(REALPATH, (handle, name, resolved_path), void*, NULL);
+}
+
+static NTSTATUS shadow_copy2_get_nt_acl(vfs_handle_struct *handle,
+			       const char *fname, uint32 security_info,
+			       struct security_descriptor **ppdesc)
+{
+        SHADOW2_NTSTATUS_NEXT(GET_NT_ACL, (handle, name, security_info, ppdesc), NT_STATUS_ACCESS_DENIED);
+}
+
+static NTSTATUS shadow_copy2_set_nt_acl(vfs_handle_struct *handle, files_struct *fsp,
+			     const char *fname, uint32 security_info_sent,
+			     struct security_descriptor *psd)
+{
+        SHADOW2_NTSTATUS_NEXT(SET_NT_ACL, (handle, fsp, name, security_info_sent, psd), NT_STATUS_ACCESS_DENIED);
+}
+
+static int shadow_copy2_mkdir(vfs_handle_struct *handle,  const char *fname, mode_t mode)
+{
+        SHADOW2_NEXT(MKDIR, (handle, name, mode), int, -1);
+}
+
+static int shadow_copy2_rmdir(vfs_handle_struct *handle,  const char *fname)
+{
+        SHADOW2_NEXT(RMDIR, (handle, name), int, -1);
+}
+
+static int shadow_copy2_chflags(vfs_handle_struct *handle, const char *fname, int flags)
+{
+        SHADOW2_NEXT(CHFLAGS, (handle, name, flags), int, -1);
+}
+
+static ssize_t shadow_copy2_getxattr(vfs_handle_struct *handle,
+				  const char *fname, const char *aname, void *value, size_t size)
+{
+        SHADOW2_NEXT(GETXATTR, (handle, name, aname, value, size), ssize_t, -1);
+}
+
+static ssize_t shadow_copy2_lgetxattr(vfs_handle_struct *handle,
+				      const char *fname, const char *aname, void *value, size_t size)
+{
+        SHADOW2_NEXT(LGETXATTR, (handle, name, aname, value, size), ssize_t, -1);
+}
+
+static ssize_t shadow_copy2_listxattr(struct vfs_handle_struct *handle, const char *fname, 
+				      char *list, size_t size)
+{
+	SHADOW2_NEXT(LISTXATTR, (handle, name, list, size), ssize_t, -1);
+}
+
+static int shadow_copy2_removexattr(struct vfs_handle_struct *handle, const char *fname, 
+				    const char *aname)
+{
+	SHADOW2_NEXT(REMOVEXATTR, (handle, name, aname), int, -1);
+}
+
+static int shadow_copy2_lremovexattr(struct vfs_handle_struct *handle, const char *fname, 
+				     const char *aname)
+{
+	SHADOW2_NEXT(LREMOVEXATTR, (handle, name, aname), int, -1);
+}
+
+static int shadow_copy2_setxattr(struct vfs_handle_struct *handle, const char *fname, 
+				 const char *aname, const void *value, size_t size, int flags)
+{
+	SHADOW2_NEXT(SETXATTR, (handle, name, aname, value, size, flags), int, -1);
+}
+
+static int shadow_copy2_lsetxattr(struct vfs_handle_struct *handle, const char *fname, 
+				  const char *aname, const void *value, size_t size, int flags)
+{
+	SHADOW2_NEXT(LSETXATTR, (handle, name, aname, value, size, flags), int, -1);
+}
+
+static int shadow_copy2_chmod_acl(vfs_handle_struct *handle,
+			   const char *fname, mode_t mode)
+{
+        /* If the underlying VFS doesn't have ACL support... */
+        if (!handle->vfs_next.ops.chmod_acl) {
+                errno = ENOSYS;
+                return -1;
+        }
+        SHADOW2_NEXT(CHMOD_ACL, (handle, name, mode), int, -1);
+}
+
+static int shadow_copy2_get_shadow_copy2_data(vfs_handle_struct *handle, 
+					      files_struct *fsp, 
+					      SHADOW_COPY_DATA *shadow_copy2_data, 
+					      bool labels)
+{
+	SMB_STRUCT_DIR *p;
+	const char *snapdir;
+	SMB_STRUCT_DIRENT *d;
+	TALLOC_CTX *tmp_ctx = talloc_new(handle->data);
+
+	snapdir = shadow_copy2_find_snapdir(tmp_ctx, handle);
+	if (snapdir == NULL) {
+		DEBUG(0,("shadow:snapdir not found for %s in get_shadow_copy_data\n",
+			 handle->conn->connectpath));
+		errno = EINVAL;
+		talloc_free(tmp_ctx);
+		return -1;
+	}
+
+	p = SMB_VFS_NEXT_OPENDIR(handle, snapdir, NULL, 0);
+
+	if (!p) {
+		DEBUG(0,("shadow_copy2: SMB_VFS_NEXT_OPENDIR() failed for '%s' - %s\n", 
+			 snapdir, strerror(errno)));
+		talloc_free(tmp_ctx);
+		return -1;
+	}
+
+	talloc_free(tmp_ctx);
+
+	shadow_copy2_data->num_volumes = 0;
+	shadow_copy2_data->labels      = NULL;
+
+	while ((d = SMB_VFS_NEXT_READDIR(handle, p))) {
+		SHADOW_COPY_LABEL *tlabels;
+
+		/* ignore names not of the right form in the snapshot directory */
+		if (!shadow_copy2_match_name(d->d_name)) {
+			continue;
+		}
+
+		if (!labels) {
+			/* the caller doesn't want the labels */
+			shadow_copy2_data->num_volumes++;
+			continue;
+		}
+
+		tlabels = talloc_realloc(shadow_copy2_data->mem_ctx,
+					 shadow_copy2_data->labels,
+					 SHADOW_COPY_LABEL, shadow_copy2_data->num_volumes+1);
+		if (tlabels == NULL) {
+			DEBUG(0,("shadow_copy2: out of memory\n"));
+			SMB_VFS_NEXT_CLOSEDIR(handle, p);
+			return -1;
+		}
+
+		strlcpy(tlabels[shadow_copy2_data->num_volumes], d->d_name, sizeof(*tlabels));
+		shadow_copy2_data->num_volumes++;
+		shadow_copy2_data->labels = tlabels;
+	}
+
+	SMB_VFS_NEXT_CLOSEDIR(handle,p);
+	return 0;
+}
+
+/* VFS operations structure */
+
+static vfs_op_tuple shadow_copy2_ops[] = {
+        {SMB_VFS_OP(shadow_copy2_opendir),  SMB_VFS_OP_OPENDIR,  SMB_VFS_LAYER_TRANSPARENT},
+
+	/* directory operations */
+        {SMB_VFS_OP(shadow_copy2_mkdir),       SMB_VFS_OP_MKDIR,       SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_rmdir),       SMB_VFS_OP_RMDIR,       SMB_VFS_LAYER_TRANSPARENT},
+
+	/* xattr and flags operations */
+        {SMB_VFS_OP(shadow_copy2_chflags),     SMB_VFS_OP_CHFLAGS,     SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_getxattr),    SMB_VFS_OP_GETXATTR,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_lgetxattr),   SMB_VFS_OP_LGETXATTR,   SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_listxattr),   SMB_VFS_OP_LISTXATTR,   SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_removexattr), SMB_VFS_OP_REMOVEXATTR, SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_lremovexattr),SMB_VFS_OP_LREMOVEXATTR,SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_setxattr),    SMB_VFS_OP_SETXATTR,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_lsetxattr),   SMB_VFS_OP_LSETXATTR,   SMB_VFS_LAYER_TRANSPARENT},
+
+        /* File operations */
+        {SMB_VFS_OP(shadow_copy2_open),       SMB_VFS_OP_OPEN,     SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_rename),     SMB_VFS_OP_RENAME,   SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_stat),       SMB_VFS_OP_STAT,     SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_lstat),      SMB_VFS_OP_LSTAT,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_fstat),      SMB_VFS_OP_FSTAT,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_unlink),     SMB_VFS_OP_UNLINK,   SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_chmod),      SMB_VFS_OP_CHMOD,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_chown),      SMB_VFS_OP_CHOWN,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_chdir),      SMB_VFS_OP_CHDIR,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_ntimes),     SMB_VFS_OP_NTIMES,   SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_symlink),    SMB_VFS_OP_SYMLINK,  SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_readlink),   SMB_VFS_OP_READLINK, SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_link),       SMB_VFS_OP_LINK,     SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_mknod),      SMB_VFS_OP_MKNOD,    SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_realpath),   SMB_VFS_OP_REALPATH, SMB_VFS_LAYER_TRANSPARENT},
+
+        /* NT File ACL operations */
+        {SMB_VFS_OP(shadow_copy2_get_nt_acl), SMB_VFS_OP_GET_NT_ACL, SMB_VFS_LAYER_TRANSPARENT},
+        {SMB_VFS_OP(shadow_copy2_set_nt_acl), SMB_VFS_OP_SET_NT_ACL, SMB_VFS_LAYER_TRANSPARENT},
+
+        /* POSIX ACL operations */
+        {SMB_VFS_OP(shadow_copy2_chmod_acl), SMB_VFS_OP_CHMOD_ACL, SMB_VFS_LAYER_TRANSPARENT},
+
+	/* special shadown copy op */
+	{SMB_VFS_OP(shadow_copy2_get_shadow_copy2_data), 
+	 SMB_VFS_OP_GET_SHADOW_COPY_DATA,SMB_VFS_LAYER_OPAQUE},
+
+	{SMB_VFS_OP(NULL), SMB_VFS_OP_NOOP, SMB_VFS_LAYER_NOOP}
+};
+
+NTSTATUS vfs_shadow_copy2_init(void);
+NTSTATUS vfs_shadow_copy2_init(void)
+{
+	NTSTATUS ret;
+
+	ret = smb_register_vfs(SMB_VFS_INTERFACE_VERSION, "shadow_copy2", shadow_copy2_ops);
+
+	if (!NT_STATUS_IS_OK(ret))
+		return ret;
+
+	vfs_shadow_copy2_debug_level = debug_add_class("shadow_copy2");
+	if (vfs_shadow_copy2_debug_level == -1) {
+		vfs_shadow_copy2_debug_level = DBGC_VFS;
+		DEBUG(0, ("%s: Couldn't register custom debugging class!\n",
+			"vfs_shadow_copy2_init"));
+	} else {
+		DEBUG(10, ("%s: Debug class number of '%s': %d\n", 
+			"vfs_shadow_copy2_init","shadow_copy2",vfs_shadow_copy2_debug_level));
+	}
+
+	return ret;
+}
diff --git a/source3/modules/vfs_tsmsm.c b/source3/modules/vfs_tsmsm.c
new file mode 100644
index 0000000000..fe791b24bf
--- /dev/null
+++ b/source3/modules/vfs_tsmsm.c
@@ -0,0 +1,338 @@
+/*
+  Unix SMB/CIFS implementation.
+  Samba VFS module for handling offline files
+  with Tivoli Storage Manager Space Management
+
+  (c) Alexander Bokovoy, 2007
+  (c) Andrew Tridgell, 2007
+  
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+/*
+  This VFS module accepts following options:
+  tsmsm: hsm script = <path to hsm script> (/bin/true by default, i.e. does nothing)
+         hsm script should point to a shell script which accepts two arguments:
+	 <operation> <filepath>
+	 where <operation> is currently 'offline' to set offline status of the <filepath>
+
+  tsmsm: online ratio = ratio to check reported size against actual file size (0.5 by default)
+
+  The TSMSM VFS module tries to avoid calling expensive DMAPI calls with some heuristics
+  based on the fact that number of blocks reported of a file multiplied by 512 will be
+  bigger than 'online ratio' of actual size for online (non-migrated) files.
+
+  If checks fail, we call DMAPI and ask for specific IBM attribute which present for
+  offline (migrated) files. If this attribute presents, we consider file offline.
+ */
+
+#include "includes.h"
+
+#ifndef USE_DMAPI
+#error "This module requires DMAPI support!"
+#endif
+
+#ifdef HAVE_XFS_DMAPI_H
+#include <xfs/dmapi.h>
+#elif defined(HAVE_SYS_DMI_H)
+#include <sys/dmi.h>
+#elif defined(HAVE_SYS_JFSDMAPI_H)
+#include <sys/jfsdmapi.h>
+#elif defined(HAVE_SYS_DMAPI_H)
+#include <sys/dmapi.h>
+#elif defined(HAVE_DMAPI_H)
+#include <dmapi.h>
+#endif
+
+#ifndef _ISOC99_SOURCE
+#define _ISOC99_SOURCE 
+#endif
+
+#include <math.h> 
+
+/* optimisation tunables - used to avoid the DMAPI slow path */
+#define FILE_IS_ONLINE_RATIO      0.5
+#define DM_ATTRIB_OBJECT "IBMObj"
+#define DM_ATTRIB_MIGRATED "IBMMig"
+
+struct tsmsm_struct {
+	dm_sessid_t sid;
+	float online_ratio;
+	char *hsmscript;
+};
+
+#define TSM_STRINGIFY(a) #a
+#define TSM_TOSTRING(a) TSM_STRINGIFY(a)
+
+static void tsmsm_free_data(void **pptr) {
+	struct tsmsm_struct **tsmd = (struct tsmsm_struct **)pptr;
+	if(!tsmd) return;
+	TALLOC_FREE(*tsmd);
+}
+
+static int tsmsm_connect(struct vfs_handle_struct *handle,
+			 const char *service,
+			 const char *user) {
+	struct tsmsm_struct *tsmd = TALLOC_ZERO_P(handle, struct tsmsm_struct);
+	const char *hsmscript, *tsmname;
+	const char *fres;
+	
+	if (!tsmd) {
+		DEBUG(0,("tsmsm_connect: out of memory!\n"));
+		return -1;
+	}
+
+	tsmd->sid = *(dm_sessid_t*) dmapi_get_current_session();
+
+	if (tsmd->sid == DM_NO_SESSION) {
+		DEBUG(0,("tsmsm_connect: no DMAPI session for Samba is available!\n"));
+		TALLOC_FREE(tsmd);
+		return -1;
+	}
+
+	tsmname = (handle->param ? handle->param : "tsmsm");
+	hsmscript = lp_parm_const_string(SNUM(handle->conn), tsmname,
+					 "hsm script", NULL);
+	if (hsmscript) {
+		tsmd->hsmscript = talloc_strdup(tsmd, hsmscript);
+		if(!tsmd->hsmscript) {
+			DEBUG(1, ("tsmsm_connect: can't allocate memory for hsm script path"));
+			TALLOC_FREE(tsmd);
+			return -1;
+		}
+	} else {
+		DEBUG(1, ("tsmsm_connect: can't call hsm script because it "
+			  "is not set to anything in the smb.conf\n"
+			  "Use %s: 'hsm script = path' to set it\n",
+			  tsmname));
+		TALLOC_FREE(tsmd);
+		return -1;
+	}
+
+	fres = lp_parm_const_string(SNUM(handle->conn), tsmname, 
+				    "online ratio", TSM_TOSTRING(FILE_IS_ONLINE_RATIO));
+	tsmd->online_ratio = strtof(fres, NULL);
+	if((tsmd->online_ratio == (float)0) || ((errno == ERANGE) &&
+						((tsmd->online_ratio == HUGE_VALF) ||
+						 (tsmd->online_ratio == HUGE_VALL)))) {
+		DEBUG(1, ("tsmsm_connect: error while getting online ratio from smb.conf."
+			  "Default to %s.\n", TSM_TOSTRING(FILE_IS_ONLINE_RATIO)));
+		tsmd->online_ratio = FILE_IS_ONLINE_RATIO;
+	}
+
+        /* Store the private data. */
+        SMB_VFS_HANDLE_SET_DATA(handle, tsmd, tsmsm_free_data,
+                                struct tsmsm_struct, return -1);
+        return SMB_VFS_NEXT_CONNECT(handle, service, user); 
+}
+
+static int tsmsm_is_offline(struct vfs_handle_struct *handle, 
+			    struct connection_struct *conn,
+			    const char *path,
+			    SMB_STRUCT_STAT *stbuf,
+			    bool *offline) {
+	struct tsmsm_struct *tsmd = (struct tsmsm_struct *) handle->data;
+	void *dmhandle = NULL;
+	size_t dmhandle_len = 0;
+	size_t rlen;
+	dm_attrname_t dmname;
+	int ret;
+
+        /* if the file has more than FILE_IS_ONLINE_RATIO of blocks available,
+	   then assume it is not offline (it may not be 100%, as it could be sparse) */
+	if (512 * (off_t)stbuf->st_blocks >= stbuf->st_size * tsmd->online_ratio) {
+		*offline = false;
+		DEBUG(10,("%s not offline: st_blocks=%ld st_size=%ld online_ratio=%.2f\n", 
+			  path, stbuf->st_blocks, stbuf->st_size, tsmd->online_ratio));
+		return 0;
+	}
+	
+        /* using POSIX capabilities does not work here. It's a slow path, so 
+	 * become_root() is just as good anyway (tridge) 
+	 */
+
+	/* Also, AIX has DMAPI but no POSIX capablities support. In this case,
+	 * we need to be root to do DMAPI manipulations.
+	 */
+	become_root();
+
+	/* go the slow DMAPI route */
+	if (dm_path_to_handle((char*)path, &dmhandle, &dmhandle_len) != 0) {
+		ret = -1;
+		DEBUG(2,("dm_path_to_handle failed - assuming offline (%s) - %s\n", 
+			 path, strerror(errno)));
+		*offline = True;
+		goto done;
+	}
+
+	memset(&dmname, 0, sizeof(dmname));
+	strlcpy((char *)&dmname.an_chars[0], DM_ATTRIB_OBJECT, sizeof(dmname.an_chars));
+
+	ret = dm_get_dmattr(tsmd->sid, dmhandle, dmhandle_len, 
+			    DM_NO_TOKEN, &dmname, 0, NULL, &rlen);
+
+	/* its offline if the IBMObj attribute exists */
+	*offline = (ret == 0 || (ret == -1 && errno == E2BIG));
+
+	DEBUG(10,("dm_get_dmattr %s ret=%d (%s)\n", path, ret, strerror(errno)));
+
+	ret = 0;
+
+	dm_handle_free(dmhandle, dmhandle_len);	
+
+done:
+	unbecome_root();
+	return ret;
+}
+
+
+static bool tsmsm_aio_force(struct vfs_handle_struct *handle, struct files_struct *fsp)
+{
+	SMB_STRUCT_STAT sbuf;
+	struct tsmsm_struct *tsmd = (struct tsmsm_struct *) handle->data;
+	/* see if the file might be offline. This is called before each IO
+	   to ensure we use AIO if the file is offline. We don't do the full dmapi
+	   call as that would be too slow, instead we err on the side of using AIO
+	   if the file might be offline
+	*/
+	if(SMB_VFS_FSTAT(fsp, &sbuf) == 0) {
+		DEBUG(10,("tsmsm_aio_force st_blocks=%ld st_size=%ld online_ratio=%.2f\n", 
+			  sbuf.st_blocks, sbuf.st_size, tsmd->online_ratio));
+		return !(512 * (off_t)sbuf.st_blocks >= sbuf.st_size * tsmd->online_ratio);
+	}
+	return False;
+}
+
+static ssize_t tsmsm_aio_return(struct vfs_handle_struct *handle, struct files_struct *fsp, 
+				SMB_STRUCT_AIOCB *aiocb)
+{
+	ssize_t result;
+
+	result = SMB_VFS_NEXT_AIO_RETURN(handle, fsp, aiocb);
+	if(result >= 0) {
+		notify_fname(handle->conn, NOTIFY_ACTION_MODIFIED,
+			     FILE_NOTIFY_CHANGE_ATTRIBUTES,
+			     fsp->fsp_name);
+	}
+
+	return result;
+}
+
+static ssize_t tsmsm_sendfile(vfs_handle_struct *handle, int tofd, files_struct *fsp, const DATA_BLOB *hdr,
+			      SMB_OFF_T offset, size_t n)
+{
+	bool file_online = tsmsm_aio_force(handle, fsp);
+
+	if(!file_online) 
+	    return ENOSYS;
+	    
+	return SMB_VFS_NEXT_SENDFILE(handle, tofd, fsp, hdr, offset, n);
+}
+
+/* We do overload pread to allow notification when file becomes online after offline status */
+/* We don't intercept SMB_VFS_READ here because all file I/O now goes through SMB_VFS_PREAD instead */
+static ssize_t tsmsm_pread(struct vfs_handle_struct *handle, struct files_struct *fsp, 
+			   void *data, size_t n, SMB_OFF_T offset) {
+	ssize_t result;
+	bool notify_online = tsmsm_aio_force(handle, fsp);
+
+	result = SMB_VFS_NEXT_PREAD(handle, fsp, data, n, offset);
+	if((result != -1) && notify_online) {
+	    /* We can't actually force AIO at this point (came here not from reply_read_and_X) 
+	       what we can do is to send notification that file became online
+	    */
+		notify_fname(handle->conn, NOTIFY_ACTION_MODIFIED,
+			     FILE_NOTIFY_CHANGE_ATTRIBUTES,
+			     fsp->fsp_name);
+	}
+
+	return result;
+}
+
+static ssize_t tsmsm_pwrite(struct vfs_handle_struct *handle, struct files_struct *fsp, 
+			   void *data, size_t n, SMB_OFF_T offset) {
+	ssize_t result;
+	bool notify_online = tsmsm_aio_force(handle, fsp);
+
+	result = SMB_VFS_NEXT_PWRITE(handle, fsp, data, n, offset);
+	if((result != -1) && notify_online) {
+	    /* We can't actually force AIO at this point (came here not from reply_read_and_X) 
+	       what we can do is to send notification that file became online
+	    */
+		notify_fname(handle->conn, NOTIFY_ACTION_MODIFIED,
+			     FILE_NOTIFY_CHANGE_ATTRIBUTES,
+			     fsp->fsp_name);
+	}
+
+	return result;
+}
+
+static int tsmsm_set_offline(struct vfs_handle_struct *handle, struct connection_struct *conn, 
+			     const char *path) {
+	struct tsmsm_struct *tsmd = (struct tsmsm_struct *) handle->data;
+	int result = 0;
+	char *command;
+
+	/* Now, call the script */
+	command = talloc_asprintf(tsmd, "%s offline \"%s\"", tsmd->hsmscript, path);
+	if(!command) {
+		DEBUG(1, ("tsmsm_set_offline: can't allocate memory to run hsm script"));
+		return -1;
+	}
+	DEBUG(10, ("tsmsm_set_offline: Running [%s]\n", command));
+	if((result = smbrun(command, NULL)) != 0) {
+		DEBUG(1,("tsmsm_set_offline: Running [%s] returned %d\n", command, result));
+	}
+	TALLOC_FREE(command);
+	return result;
+}
+
+static bool tsmsm_is_remotestorage(struct vfs_handle_struct *handle, struct connection_struct *conn, const char *path) {
+	return True;
+}
+
+static vfs_op_tuple vfs_tsmsm_ops[] = {
+
+	/* Disk operations */
+
+	{SMB_VFS_OP(tsmsm_connect),	SMB_VFS_OP_CONNECT,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_aio_force),	SMB_VFS_OP_AIO_FORCE,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_aio_return),	SMB_VFS_OP_AIO_RETURN,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_pread),	SMB_VFS_OP_PREAD,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_pwrite),	SMB_VFS_OP_PWRITE,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_sendfile),	SMB_VFS_OP_SENDFILE,
+	 SMB_VFS_LAYER_TRANSPARENT},
+	{SMB_VFS_OP(tsmsm_is_offline),SMB_VFS_OP_IS_OFFLINE,
+	 SMB_VFS_LAYER_OPAQUE},
+	{SMB_VFS_OP(tsmsm_set_offline),SMB_VFS_OP_SET_OFFLINE,
+	 SMB_VFS_LAYER_OPAQUE},
+	{SMB_VFS_OP(tsmsm_is_remotestorage),SMB_VFS_OP_IS_REMOTESTORAGE,
+	 SMB_VFS_LAYER_OPAQUE},
+
+	/* Finish VFS operations definition */
+
+	{SMB_VFS_OP(NULL),		SMB_VFS_OP_NOOP,
+	 SMB_VFS_LAYER_NOOP}
+};
+
+NTSTATUS vfs_tsmsm_init(void);
+NTSTATUS vfs_tsmsm_init(void)
+{
+	return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
+				"tsmsm", vfs_tsmsm_ops);
+}
diff --git a/source3/smbd/connection.c b/source3/smbd/connection.c
index 016c8adb1b..d7063c9989 100644
--- a/source3/smbd/connection.c
+++ b/source3/smbd/connection.c
@@ -118,6 +118,15 @@ int count_current_connections( const char *sharename, bool clear  )
 }
 
 /****************************************************************************
+ Count the number of connections open across all shares.
+****************************************************************************/
+
+int count_all_current_connections(void)
+{
+        return count_current_connections(NULL, True /* clear stale entries */);
+}
+
+/****************************************************************************
  Claim an entry in the connections database.
 ****************************************************************************/
 
diff --git a/source3/smbd/dmapi.c b/source3/smbd/dmapi.c
index 05e9138ea9..620baf199e 100644
--- a/source3/smbd/dmapi.c
+++ b/source3/smbd/dmapi.c
@@ -46,7 +46,7 @@ bool dmapi_have_session(void) { return False; }
 #define DMAPI_SESSION_NAME "samba"
 #define DMAPI_TRACE 10
 
-static dm_sessid_t dmapi_session = DM_NO_SESSION;
+static dm_sessid_t samba_dmapi_session = DM_NO_SESSION;
 
 /* Initialise the DMAPI interface. Make sure that we only end up initialising
  * once per process to avoid resource leaks across different DMAPI
@@ -75,7 +75,7 @@ static int init_dmapi_service(void)
 
 bool dmapi_have_session(void)
 {
-	return dmapi_session != DM_NO_SESSION;
+	return samba_dmapi_session != DM_NO_SESSION;
 }
 
 static dm_sessid_t *realloc_session_list(dm_sessid_t * sessions, int count)
@@ -110,7 +110,7 @@ int dmapi_init_session(void)
 	 */
 	SMB_WARN(getuid() == 0, "dmapi_init_session must be called as root");
 
-	dmapi_session = DM_NO_SESSION;
+	samba_dmapi_session = DM_NO_SESSION;
 	if (init_dmapi_service() < 0) {
 		return -1;
 	}
@@ -139,7 +139,7 @@ retry:
 		err = dm_query_session(sessions[i], sizeof(buf), buf, &buflen);
 		buf[sizeof(buf) - 1] = '\0';
 		if (err == 0 && strcmp(DMAPI_SESSION_NAME, buf) == 0) {
-			dmapi_session = sessions[i];
+			samba_dmapi_session = sessions[i];
 			DEBUGADD(DMAPI_TRACE,
 				("attached to existing DMAPI session "
 				 "named '%s'\n", buf));
@@ -150,16 +150,15 @@ retry:
 	TALLOC_FREE(sessions);
 
 	/* No session already defined. */
-	if (dmapi_session == DM_NO_SESSION) {
-		err = dm_create_session(DM_NO_SESSION,
-					CONST_DISCARD(char *,
-						      DMAPI_SESSION_NAME),
-					&dmapi_session);
+	if (samba_dmapi_session == DM_NO_SESSION) {
+		err = dm_create_session(DM_NO_SESSION, 
+					CONST_DISCARD(char *, DMAPI_SESSION_NAME),
+					&samba_dmapi_session);
 		if (err < 0) {
 			DEBUGADD(DMAPI_TRACE,
 				("failed to create new DMAPI session: %s\n",
 				strerror(errno)));
-			dmapi_session = DM_NO_SESSION;
+			samba_dmapi_session = DM_NO_SESSION;
 			return -1;
 		}
 
@@ -185,22 +184,22 @@ static int reattach_dmapi_session(void)
 	char	buf[DM_SESSION_INFO_LEN];
 	size_t	buflen;
 
-	if (dmapi_session != DM_NO_SESSION ) {
+	if (samba_dmapi_session != DM_NO_SESSION ) {
 		become_root();
 
 		/* NOTE: On Linux, this call opens /dev/dmapi, costing us a
 		 * file descriptor. Ideally, we would close this when we fork.
 		 */
 		if (init_dmapi_service() < 0) {
-			dmapi_session = DM_NO_SESSION;
+			samba_dmapi_session = DM_NO_SESSION;
 			unbecome_root();
 			return -1;
 		}
 
-		if (dm_query_session(dmapi_session, sizeof(buf),
+		if (dm_query_session(samba_dmapi_session, sizeof(buf),
 			    buf, &buflen) < 0) {
 			/* Session is stale. Disable DMAPI. */
-			dmapi_session = DM_NO_SESSION;
+			samba_dmapi_session = DM_NO_SESSION;
 			unbecome_root();
 			return -1;
 		}
@@ -214,33 +213,42 @@ static int reattach_dmapi_session(void)
 	return 0;
 }
 
-uint32 dmapi_file_flags(const char * const path)
+/* If a DMAPI session has been initialised, then we need to make sure
+ * we are attached to it and have the correct privileges. This is
+ * necessary to be able to do DMAPI operations across a fork(2). If
+ * it fails, there is no likelihood of that failure being transient.
+ *
+ * Note that this use of the static attached flag relies on the fact
+ * that dmapi_file_flags() is never called prior to forking the
+ * per-client server process.
+ */
+const void * dmapi_get_current_session(void) 
 {
 	static int attached = 0;
+	if (dmapi_have_session() && !attached) {
+		attached++;
+		if (reattach_dmapi_session() < 0) {
+			return DM_NO_SESSION;
+		}
+	}
+	return &samba_dmapi_session;
+}
 
+uint32 dmapi_file_flags(const char * const path)
+{
+	dm_sessid_t dmapi_session;
 	int		err;
 	dm_eventset_t   events = {0};
 	uint		nevents;
 
-	void	*dm_handle;
-	size_t	dm_handle_len;
+	void	*dm_handle = NULL;
+	size_t	dm_handle_len = 0;
 
 	uint32	flags = 0;
 
-	/* If a DMAPI session has been initialised, then we need to make sure
-	 * we are attached to it and have the correct privileges. This is
-	 * necessary to be able to do DMAPI operations across a fork(2). If
-	 * it fails, there is no liklihood of that failure being transient.
-	 *
-	 * Note that this use of the static attached flag relies on the fact
-	 * that dmapi_file_flags() is never called prior to forking the
-	 * per-client server process.
-	 */
-	if (dmapi_have_session() && !attached) {
-		attached++;
-		if (reattach_dmapi_session() < 0) {
-			return 0;
-		}
+	dmapi_session = *(dm_sessid_t*) dmapi_get_current_session();
+	if (dmapi_session == DM_NO_SESSION) {
+		return 0;
 	}
 
 	/* AIX has DMAPI but no POSIX capablities support. In this case,
@@ -313,4 +321,5 @@ done:
 	return flags;
 }
 
+
 #endif /* USE_DMAPI */
diff --git a/source3/smbd/dosmode.c b/source3/smbd/dosmode.c
index d3813f9b41..2021621dfa 100644
--- a/source3/smbd/dosmode.c
+++ b/source3/smbd/dosmode.c
@@ -31,23 +31,6 @@ static int set_sparse_flag(const SMB_STRUCT_STAT * const sbuf)
 }
 
 /****************************************************************************
- Work out whether this file is offline
-****************************************************************************/
-
-static uint32 set_offline_flag(connection_struct *conn, const char *const path)
-{
-	if (ISDOT(path) || ISDOTDOT(path)) {
-		return 0;
-	}
-
-	if (!lp_dmapi_support(SNUM(conn)) || !dmapi_have_session()) {
-		return 0;
-	}
-
-	return dmapi_file_flags(path);
-}
-
-/****************************************************************************
  Change a dos mode to a unix mode.
     Base permission for files:
          if creating file and inheriting (i.e. parent_dir != NULL)
@@ -366,6 +349,8 @@ uint32 dos_mode_msdfs(connection_struct *conn, const char *path,SMB_STRUCT_STAT
 uint32 dos_mode(connection_struct *conn, const char *path,SMB_STRUCT_STAT *sbuf)
 {
 	uint32 result = 0;
+	bool offline;
+	int ret;
 
 	DEBUG(8,("dos_mode: %s\n", path));
 
@@ -395,8 +380,10 @@ uint32 dos_mode(connection_struct *conn, const char *path,SMB_STRUCT_STAT *sbuf)
 		result |= dos_mode_from_sbuf(conn, path, sbuf);
 	}
 
-	if (S_ISREG(sbuf->st_mode)) {
-		result |= set_offline_flag(conn, path);
+	
+	ret = SMB_VFS_IS_OFFLINE(conn, path, sbuf, &offline);
+	if (S_ISREG(sbuf->st_mode) && (ret == 0) && offline) {
+		result |= FILE_ATTRIBUTE_OFFLINE;
 	}
 
 	/* Optimization : Only call is_hidden_path if it's not already
@@ -432,7 +419,7 @@ int file_set_dosmode(connection_struct *conn, const char *fname,
 	int mask=0;
 	mode_t tmp;
 	mode_t unixmode;
-	int ret = -1;
+	int ret = -1, lret = -1;
 
 	/* We only allow READONLY|HIDDEN|SYSTEM|DIRECTORY|ARCHIVE here. */
 	dosmode &= SAMBA_ATTRIBUTES_MASK;
@@ -505,10 +492,21 @@ int file_set_dosmode(connection_struct *conn, const char *fname,
 		unixmode |= (st->st_mode & (S_IWUSR|S_IWGRP|S_IWOTH));
 	}
 
-	if ((ret = SMB_VFS_CHMOD(conn,fname,unixmode)) == 0) {
-		if (!newfile) {
+	if (dosmode & FILE_ATTRIBUTE_OFFLINE) {
+		lret = SMB_VFS_SET_OFFLINE(conn, fname);
+		if (lret == -1) {
+			DEBUG(0, ("set_dos_mode: client has asked to set "
+				  "FILE_ATTRIBUTE_OFFLINE to %s/%s but there was "
+				  "an error while setting it or it is not supported.\n", 
+				  parent_dir, fname));
+		}
+	}
+
+	ret = SMB_VFS_CHMOD(conn, fname, unixmode);
+	if (ret == 0) {
+		if(!newfile || (lret != -1)) {
 			notify_fname(conn, NOTIFY_ACTION_MODIFIED,
-				FILE_NOTIFY_CHANGE_ATTRIBUTES, fname);
+				     FILE_NOTIFY_CHANGE_ATTRIBUTES, fname);
 		}
 		st->st_mode = unixmode;
 		return 0;
diff --git a/source3/smbd/oplock_linux.c b/source3/smbd/oplock_linux.c
index 05021b6c74..fa7cb42bc6 100644
--- a/source3/smbd/oplock_linux.c
+++ b/source3/smbd/oplock_linux.c
@@ -93,17 +93,27 @@ static void set_capability(unsigned capability)
 		return;
 	}
 
-	data.effective |= (1<<capability);
+	if (0 == (data.effective & (1<<capability))) {
+		data.effective |= (1<<capability);
 
-	if (capset(&header, &data) == -1) {
-		DEBUG(3,("Unable to set %d capability (%s)\n", 
-			 capability, strerror(errno)));
+		if (capset(&header, &data) == -1) {
+			DEBUG(3,("Unable to set %d capability (%s)\n", 
+				 capability, strerror(errno)));
+		}
 	}
 }
 
 /*
- Call to set the kernel lease signal handler
-*/
+ * public function to get linux lease capability. Needed by some VFS modules (eg. gpfs.c)
+ */
+void linux_set_lease_capability(void)
+{
+	set_capability(CAP_LEASE);
+}
+
+/* 
+ * Call to set the kernel lease signal handler
+ */
 int linux_set_lease_sighandler(int fd)
 {
         if (fcntl(fd, F_SETSIG, RT_SIGNAL_LEASE) == -1) {
diff --git a/source3/smbd/posix_acls.c b/source3/smbd/posix_acls.c
index 5f18615f66..6cec39f9c0 100644
--- a/source3/smbd/posix_acls.c
+++ b/source3/smbd/posix_acls.c
@@ -3413,6 +3413,9 @@ NTSTATUS set_nt_acl(files_struct *fsp, uint32 security_info_sent, SEC_DESC *psd)
 	bool acl_perms = False;
 	mode_t orig_mode = (mode_t)0;
 	NTSTATUS status;
+	uid_t orig_uid;
+	gid_t orig_gid;
+	bool need_chown = False;
 
 	DEBUG(10,("set_nt_acl: called for file %s\n", fsp->fsp_name ));
 
@@ -3435,6 +3438,8 @@ NTSTATUS set_nt_acl(files_struct *fsp, uint32 security_info_sent, SEC_DESC *psd)
 
 	/* Save the original elements we check against. */
 	orig_mode = sbuf.st_mode;
+	orig_uid = sbuf.st_uid;
+	orig_gid = sbuf.st_gid;
 
 	/*
 	 * Unpack the user/group/world id's.
@@ -3449,7 +3454,11 @@ NTSTATUS set_nt_acl(files_struct *fsp, uint32 security_info_sent, SEC_DESC *psd)
 	 * Do we need to chown ?
 	 */
 
-	if (((user != (uid_t)-1) && (sbuf.st_uid != user)) || (( grp != (gid_t)-1) && (sbuf.st_gid != grp))) {
+	if (((user != (uid_t)-1) && (orig_uid != user)) || (( grp != (gid_t)-1) && (orig_gid != grp))) {
+		need_chown = True;
+	}
+
+	if (need_chown && (user == (uid_t)-1 || user == current_user.ut.uid)) {
 
 		DEBUG(3,("set_nt_acl: chown %s. uid = %u, gid = %u.\n",
 				fsp->fsp_name, (unsigned int)user, (unsigned int)grp ));
@@ -3487,6 +3496,11 @@ NTSTATUS set_nt_acl(files_struct *fsp, uint32 security_info_sent, SEC_DESC *psd)
 
 		/* Save the original elements we check against. */
 		orig_mode = sbuf.st_mode;
+		orig_uid = sbuf.st_uid;
+		orig_gid = sbuf.st_gid;
+
+		/* We did chown already, drop the flag */
+		need_chown = False;
 	}
 
 	create_file_sids(&sbuf, &file_owner_sid, &file_grp_sid);
@@ -3630,6 +3644,21 @@ NTSTATUS set_nt_acl(files_struct *fsp, uint32 security_info_sent, SEC_DESC *psd)
 		free_canon_ace_list(dir_ace_list); 
 	}
 
+	/* Any chown pending? */
+	if (need_chown) {
+		DEBUG(3,("set_nt_acl: chown %s. uid = %u, gid = %u.\n",
+			 fsp->fsp_name, (unsigned int)user, (unsigned int)grp ));
+		
+		if(try_chown( fsp->conn, fsp->fsp_name, user, grp) == -1) {
+			DEBUG(3,("set_nt_acl: chown %s, %u, %u failed. Error = %s.\n",
+				 fsp->fsp_name, (unsigned int)user, (unsigned int)grp, strerror(errno) ));
+			if (errno == EPERM) {
+				return NT_STATUS_INVALID_OWNER;
+			}
+			return map_nt_error_from_unix(errno);
+		}
+	}
+	
 	return NT_STATUS_OK;
 }
 
diff --git a/source3/smbd/reply.c b/source3/smbd/reply.c
index e2316ef120..381ddfe151 100644
--- a/source3/smbd/reply.c
+++ b/source3/smbd/reply.c
@@ -3329,8 +3329,12 @@ void reply_read_and_X(struct smb_request *req)
 		return;
 	}
 
-	if (!big_readX
-	    && schedule_aio_read_and_X(conn, req, fsp, startpos, smb_maxcnt)) {
+	/* It is possible for VFS modules to selectively decide whether Async I/O should be used
+	   for the file or not.
+        */
+	if ((SMB_VFS_AIO_FORCE(fsp)) &&
+	    !big_readX &&
+	    schedule_aio_read_and_X(conn, req, fsp, startpos, smb_maxcnt)) {
 		END_PROFILE(SMBreadX);
 		return;
 	}
@@ -4001,13 +4005,16 @@ void reply_write_and_X(struct smb_request *req)
 		nwritten = 0;
 	} else {
 
-		if (req->unread_bytes == 0 &&
-				schedule_aio_write_and_X(conn, req, fsp, data,
-							startpos, numtowrite)) {
+		/* It is possible for VFS modules to selectively decide whether Async I/O
+		   should be used for the file or not.
+		*/
+		if ((SMB_VFS_AIO_FORCE(fsp)) && (req->unread_bytes == 0) &&
+		    schedule_aio_write_and_X(conn, req, fsp, data, startpos,
+					     numtowrite)) {
 			END_PROFILE(SMBwriteX);
 			return;
 		}
-
+		
 		nwritten = write_file(req,fsp,data,startpos,numtowrite);
 	}
 
diff --git a/source3/smbd/server.c b/source3/smbd/server.c
index 8371d17f10..db241103ed 100644
--- a/source3/smbd/server.c
+++ b/source3/smbd/server.c
@@ -268,10 +268,20 @@ static void add_child_pid(pid_t pid)
 	num_children += 1;
 }
 
-static void remove_child_pid(pid_t pid)
+static void remove_child_pid(pid_t pid, bool unclean_shutdown)
 {
 	struct child_pid *child;
 
+	if (unclean_shutdown) {
+		/* a child terminated uncleanly so tickle all processes to see 
+		   if they can grab any of the pending locks
+		*/
+		messaging_send_buf(smbd_messaging_context(), procid_self(), 
+				   MSG_SMB_BRL_VALIDATE, NULL, 0);
+		message_send_all(smbd_messaging_context(), 
+				 MSG_SMB_UNLOCK, NULL, 0, NULL);
+	}
+
 	if (lp_max_smbd_processes() == 0) {
 		/* Don't bother with the child list if we don't care anyway */
 		return;
@@ -560,10 +570,27 @@ static bool open_sockets_smbd(bool is_daemon, bool interactive, const char *smb_
 
 		if (got_sig_cld) {
 			pid_t pid;
+			int status;
+
 			got_sig_cld = False;
 
-			while ((pid = sys_waitpid(-1, NULL, WNOHANG)) > 0) {
-				remove_child_pid(pid);
+			while ((pid = sys_waitpid(-1, &status, WNOHANG)) > 0) {
+				bool unclean_shutdown = False;
+				
+				/* If the child terminated normally, assume
+				   it was an unclean shutdown unless the
+				   status is 0 
+				*/
+				if (WIFEXITED(status)) {
+					unclean_shutdown = WEXITSTATUS(status);
+				}
+				/* If the child terminated due to a signal
+				   we always assume it was unclean.
+				*/
+				if (WIFSIGNALED(status)) {
+					unclean_shutdown = True;
+				}
+				remove_child_pid(pid, unclean_shutdown);
 			}
 		}
 
@@ -603,6 +630,15 @@ static bool open_sockets_smbd(bool is_daemon, bool interactive, const char *smb_
 
 			continue;
 		}
+		
+
+		/* If the idle timeout fired and we don't have any connected
+		 * users, exit gracefully. We should be running under a process
+		 * controller that will restart us if necessry.
+		 */
+		if (num == 0 && count_all_current_connections() == 0) {
+			exit_server_cleanly("idle timeout");
+		}
 
 		/* process pending nDNS responses */
 		if (dns_register_smbd_reply(dns_reg, &r_fds, &idle_timeout)) {
@@ -906,6 +942,29 @@ void exit_server_fault(void)
 	exit_server("critical server fault");
 }
 
+
+/****************************************************************************
+received when we should release a specific IP
+****************************************************************************/
+static void msg_release_ip(struct messaging_context *msg_ctx, void *private_data, 
+                    	   uint32_t msg_type, struct server_id server_id, DATA_BLOB *data)
+{
+	const char *ip = (const char *)data->data;
+	char addr[INET6_ADDRSTRLEN];
+
+	if (strcmp(client_socket_addr(get_client_fd(),addr,sizeof(addr)), ip) == 0) {
+		/* we can't afford to do a clean exit - that involves
+		   database writes, which would potentially mean we
+		   are still running after the failover has finished -
+		   we have to get rid of this process ID straight
+		   away */
+		DEBUG(0,("Got release IP message for our IP %s - exiting immediately\n",
+			ip));
+		_exit(0);
+	}
+}
+
+
 /****************************************************************************
  Initialise connect, service and file structs.
 ****************************************************************************/
@@ -1305,6 +1364,8 @@ extern void build_options(bool screen);
 	/* register our message handlers */
 	messaging_register(smbd_messaging_context(), NULL,
 			   MSG_SMB_FORCE_TDIS, msg_force_tdis);
+	messaging_register(smbd_messaging_context(), NULL,
+			   MSG_SMB_RELEASE_IP, msg_release_ip);
 
 	if ((lp_keepalive() != 0)
 	    && !(event_add_idle(smbd_event_context(), NULL,
diff --git a/source3/smbd/trans2.c b/source3/smbd/trans2.c
index bf6802f2a6..5729ab5349 100644
--- a/source3/smbd/trans2.c
+++ b/source3/smbd/trans2.c
@@ -2373,8 +2373,8 @@ static void call_trans2qfsinfo(connection_struct *conn,
 	const char *vname = volume_label(SNUM(conn));
 	int snum = SNUM(conn);
 	char *fstype = lp_fstype(SNUM(conn));
-	int quota_flag = 0;
-
+	uint32 additional_flags = 0;
+	
 	if (total_params < 2) {
 		reply_nterror(req, NT_STATUS_INVALID_PARAMETER);
 		return;
@@ -2487,16 +2487,23 @@ cBytesSector=%u, cUnitTotal=%u, cUnitAvail=%d\n", (unsigned int)st.st_dev, (unsi
 		case SMB_QUERY_FS_ATTRIBUTE_INFO:
 		case SMB_FS_ATTRIBUTE_INFORMATION:
 
-
+			additional_flags = 0;
 #if defined(HAVE_SYS_QUOTAS)
-			quota_flag = FILE_VOLUME_QUOTAS;
+			additional_flags |= FILE_VOLUME_QUOTAS;
 #endif
 
+			if(lp_nt_acl_support(SNUM(conn))) {
+				additional_flags |= FILE_PERSISTENT_ACLS;
+			}
+			
+			if(SMB_VFS_IS_REMOTESTORAGE(conn, lp_pathname(SNUM(conn)))) {
+				additional_flags |= FILE_SUPPORTS_REMOTE_STORAGE;
+				additional_flags |= FILE_SUPPORTS_REPARSE_POINTS;
+			}
+			
 			SIVAL(pdata,0,FILE_CASE_PRESERVED_NAMES|FILE_CASE_SENSITIVE_SEARCH|
-				(lp_nt_acl_support(SNUM(conn)) ? FILE_PERSISTENT_ACLS : 0)|
-				FILE_SUPPORTS_OBJECT_IDS|
-				FILE_UNICODE_ON_DISK|
-				quota_flag); /* FS ATTRIBUTES */
+				FILE_SUPPORTS_OBJECT_IDS|FILE_UNICODE_ON_DISK|
+				additional_flags); /* FS ATTRIBUTES */
 
 			SIVAL(pdata,4,255); /* Max filename component length */
 			/* NOTE! the fstype must *not* be null terminated or win98 won't recognise it
diff --git a/source3/winbindd/idmap_tdb2.c b/source3/winbindd/idmap_tdb2.c
new file mode 100644
index 0000000000..ab89e615f7
--- /dev/null
+++ b/source3/winbindd/idmap_tdb2.c
@@ -0,0 +1,1017 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   idmap TDB2 backend, used for clustered Samba setups.
+
+   This uses 2 tdb files. One is permanent, and is in shared storage
+   on the cluster (using "tdb:idmap2.tdb =" in smb.conf). The other is a
+   temporary cache tdb on local storage.
+
+   Copyright (C) Andrew Tridgell 2007
+
+   This is heavily based upon idmap_tdb.c, which is:
+
+   Copyright (C) Tim Potter 2000
+   Copyright (C) Jim McDonough <jmcd@us.ibm.com> 2003
+   Copyright (C) Jeremy Allison 2006
+   Copyright (C) Simo Sorce 2003-2006
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include "includes.h"
+#include "winbindd.h"
+
+#undef DBGC_CLASS
+#define DBGC_CLASS DBGC_IDMAP
+
+/* High water mark keys */
+#define HWM_GROUP  "GROUP HWM"
+#define HWM_USER   "USER HWM"
+
+static struct idmap_tdb2_state {
+	/* User and group id pool */
+	uid_t low_uid, high_uid;               /* Range of uids to allocate */
+	gid_t low_gid, high_gid;               /* Range of gids to allocate */
+	const char *idmap_script;
+} idmap_tdb2_state;
+
+
+
+/* tdb context for the local cache tdb */
+static TDB_CONTEXT *idmap_tdb2_tmp;
+
+/* handle to the permanent tdb */
+static struct db_context *idmap_tdb2_perm;
+
+/*
+  open the cache tdb
+ */
+static NTSTATUS idmap_tdb2_open_cache_db(void)
+{
+	const char *db_path;
+
+	if (idmap_tdb2_tmp) {
+		/* its already open */
+		return NT_STATUS_OK;
+	}
+
+	db_path = lock_path("idmap2_cache.tdb");
+
+	/* Open idmap repository */
+	if (!(idmap_tdb2_tmp = tdb_open_log(db_path, 0, TDB_CLEAR_IF_FIRST, O_RDWR|O_CREAT, 0644))) {
+		DEBUG(0, ("Unable to open cache idmap database '%s'\n", db_path));
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+
+	return NT_STATUS_OK;
+}
+
+
+static NTSTATUS idmap_tdb2_alloc_load(void);
+
+/*
+  open the permanent tdb
+ */
+static NTSTATUS idmap_tdb2_open_perm_db(void)
+{
+	char *db_path;
+	
+	if (idmap_tdb2_perm) {
+		/* its already open */
+		return NT_STATUS_OK;
+	}
+
+	db_path = lp_parm_talloc_string(-1, "tdb", "idmap2.tdb", NULL);
+	if (db_path == NULL) {
+		/* fall back to the private directory, which, despite
+		   its name, is usually on shared storage */
+		db_path = talloc_asprintf(NULL, "%s/idmap2.tdb", lp_private_dir());
+	}
+	NT_STATUS_HAVE_NO_MEMORY(db_path);
+
+	/* Open idmap repository */
+	idmap_tdb2_perm = db_open(NULL, db_path, 0, TDB_DEFAULT,
+				  O_RDWR|O_CREAT, 0644);
+	TALLOC_FREE(db_path);
+
+	if (idmap_tdb2_perm == NULL) {
+		DEBUG(0, ("Unable to open permanent idmap database '%s'\n",
+			  db_path));
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+
+	/* load the ranges and high/low water marks */
+	return idmap_tdb2_alloc_load();
+}
+
+
+/*
+  load the idmap allocation ranges and high/low water marks
+*/
+static NTSTATUS idmap_tdb2_alloc_load(void)
+{
+	const char *range;
+	uid_t low_uid = 0;
+	uid_t high_uid = 0;
+	gid_t low_gid = 0;
+	gid_t high_gid = 0;
+
+	/* load ranges */
+	idmap_tdb2_state.low_uid = 0;
+	idmap_tdb2_state.high_uid = 0;
+	idmap_tdb2_state.low_gid = 0;
+	idmap_tdb2_state.high_gid = 0;
+
+	/* see if a idmap script is configured */
+	idmap_tdb2_state.idmap_script = lp_parm_const_string(-1, "idmap", "script", NULL);
+
+	if (idmap_tdb2_state.idmap_script) {
+		DEBUG(1, ("using idmap script '%s'\n", idmap_tdb2_state.idmap_script));
+	}
+
+	range = lp_parm_const_string(-1, "idmap alloc config", "range", NULL);
+	if (range && range[0]) {
+		unsigned low_id, high_id;
+		if (sscanf(range, "%u - %u", &low_id, &high_id) == 2) {
+			if (low_id < high_id) {
+				idmap_tdb2_state.low_gid = idmap_tdb2_state.low_uid = low_id;
+				idmap_tdb2_state.high_gid = idmap_tdb2_state.high_uid = high_id;
+			} else {
+				DEBUG(1, ("ERROR: invalid idmap alloc range [%s]", range));
+			}
+		} else {
+			DEBUG(1, ("ERROR: invalid syntax for idmap alloc config:range [%s]", range));
+		}
+	}
+
+	/* Create high water marks for group and user id */
+	if (lp_idmap_uid(&low_uid, &high_uid)) {
+		idmap_tdb2_state.low_uid = low_uid;
+		idmap_tdb2_state.high_uid = high_uid;
+	}
+
+	if (lp_idmap_gid(&low_gid, &high_gid)) {
+		idmap_tdb2_state.low_gid = low_gid;
+		idmap_tdb2_state.high_gid = high_gid;
+	}
+
+	if (idmap_tdb2_state.high_uid <= idmap_tdb2_state.low_uid) {
+		DEBUG(1, ("idmap uid range missing or invalid\n"));
+		DEBUGADD(1, ("idmap will be unable to map foreign SIDs\n"));
+		return NT_STATUS_UNSUCCESSFUL;
+	} else {
+		uint32 low_id;
+
+		if (((low_id = dbwrap_fetch_int32(idmap_tdb2_perm,
+						  HWM_USER)) == -1) ||
+		    (low_id < idmap_tdb2_state.low_uid)) {
+			if (dbwrap_store_int32(
+				    idmap_tdb2_perm, HWM_USER,
+				    idmap_tdb2_state.low_uid) == -1) {
+				DEBUG(0, ("Unable to initialise user hwm in idmap database\n"));
+				return NT_STATUS_INTERNAL_DB_ERROR;
+			}
+		}
+	}
+
+	if (idmap_tdb2_state.high_gid <= idmap_tdb2_state.low_gid) {
+		DEBUG(1, ("idmap gid range missing or invalid\n"));
+		DEBUGADD(1, ("idmap will be unable to map foreign SIDs\n"));
+		return NT_STATUS_UNSUCCESSFUL;
+	} else {
+		uint32 low_id;
+
+		if (((low_id = dbwrap_fetch_int32(idmap_tdb2_perm,
+						  HWM_GROUP)) == -1) ||
+		    (low_id < idmap_tdb2_state.low_gid)) {
+			if (dbwrap_store_int32(
+				    idmap_tdb2_perm, HWM_GROUP,
+				    idmap_tdb2_state.low_gid) == -1) {
+				DEBUG(0, ("Unable to initialise group hwm in idmap database\n"));
+				return NT_STATUS_INTERNAL_DB_ERROR;
+			}
+		}
+	}
+
+	return NT_STATUS_OK;
+}
+
+
+/*
+  Initialise idmap alloc database. 
+*/
+static NTSTATUS idmap_tdb2_alloc_init(const char *params)
+{
+	/* nothing to do - we want to avoid opening the permanent
+	   database if possible. Instead we load the params when we
+	   first need it. */
+	return NT_STATUS_OK;
+}
+
+
+/*
+  Allocate a new id. 
+*/
+static NTSTATUS idmap_tdb2_allocate_id(struct unixid *xid)
+{
+	bool ret;
+	const char *hwmkey;
+	const char *hwmtype;
+	uint32_t high_hwm;
+	uint32_t hwm;
+	NTSTATUS status;
+
+	status = idmap_tdb2_open_perm_db();
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	/* Get current high water mark */
+	switch (xid->type) {
+
+	case ID_TYPE_UID:
+		hwmkey = HWM_USER;
+		hwmtype = "UID";
+		high_hwm = idmap_tdb2_state.high_uid;
+		break;
+
+	case ID_TYPE_GID:
+		hwmkey = HWM_GROUP;
+		hwmtype = "GID";
+		high_hwm = idmap_tdb2_state.high_gid;
+		break;
+
+	default:
+		DEBUG(2, ("Invalid ID type (0x%x)\n", xid->type));
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	if ((hwm = dbwrap_fetch_int32(idmap_tdb2_perm, hwmkey)) == -1) {
+		return NT_STATUS_INTERNAL_DB_ERROR;
+	}
+
+	/* check it is in the range */
+	if (hwm > high_hwm) {
+		DEBUG(1, ("Fatal Error: %s range full!! (max: %lu)\n", 
+			  hwmtype, (unsigned long)high_hwm));
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+
+	/* fetch a new id and increment it */
+	ret = dbwrap_change_uint32_atomic(idmap_tdb2_perm, hwmkey, &hwm, 1);
+	if (ret == -1) {
+		DEBUG(1, ("Fatal error while fetching a new %s value\n!", hwmtype));
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+
+	/* recheck it is in the range */
+	if (hwm > high_hwm) {
+		DEBUG(1, ("Fatal Error: %s range full!! (max: %lu)\n", 
+			  hwmtype, (unsigned long)high_hwm));
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+	
+	xid->id = hwm;
+	DEBUG(10,("New %s = %d\n", hwmtype, hwm));
+
+	return NT_STATUS_OK;
+}
+
+/*
+  Get current highest id. 
+*/
+static NTSTATUS idmap_tdb2_get_hwm(struct unixid *xid)
+{
+	const char *hwmkey;
+	const char *hwmtype;
+	uint32_t hwm;
+	uint32_t high_hwm;
+
+	/* Get current high water mark */
+	switch (xid->type) {
+
+	case ID_TYPE_UID:
+		hwmkey = HWM_USER;
+		hwmtype = "UID";
+		high_hwm = idmap_tdb2_state.high_uid;
+		break;
+
+	case ID_TYPE_GID:
+		hwmkey = HWM_GROUP;
+		hwmtype = "GID";
+		high_hwm = idmap_tdb2_state.high_gid;
+		break;
+
+	default:
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	if ((hwm = dbwrap_fetch_int32(idmap_tdb2_perm, hwmkey)) == -1) {
+		return NT_STATUS_INTERNAL_DB_ERROR;
+	}
+
+	xid->id = hwm;
+
+	/* Warn if it is out of range */
+	if (hwm >= high_hwm) {
+		DEBUG(0, ("Warning: %s range full!! (max: %lu)\n", 
+			  hwmtype, (unsigned long)high_hwm));
+	}
+
+	return NT_STATUS_OK;
+}
+
+/*
+  Set high id. 
+*/
+static NTSTATUS idmap_tdb2_set_hwm(struct unixid *xid)
+{
+	/* not supported, or we would invalidate the cache tdb on
+	   other nodes */
+	DEBUG(0,("idmap_tdb2_set_hwm not supported\n"));
+	return NT_STATUS_NOT_SUPPORTED;
+}
+
+/*
+  Close the alloc tdb 
+*/
+static NTSTATUS idmap_tdb2_alloc_close(void)
+{
+	/* don't actually close it */
+	return NT_STATUS_OK;
+}
+
+/*
+  IDMAP MAPPING TDB BACKEND
+*/
+struct idmap_tdb2_context {
+	uint32_t filter_low_id;
+	uint32_t filter_high_id;
+};
+
+/*
+  try fetching from the cache tdb, and if that fails then
+  fetch from the permanent tdb
+ */
+static TDB_DATA tdb2_fetch_bystring(TALLOC_CTX *mem_ctx, const char *keystr)
+{
+	TDB_DATA ret;
+	NTSTATUS status;
+
+	ret = tdb_fetch_bystring(idmap_tdb2_tmp, keystr);
+	if (ret.dptr != NULL) {
+		/* got it from cache */
+		unsigned char *tmp;
+
+		tmp = (unsigned char *)talloc_memdup(mem_ctx, ret.dptr,
+						     ret.dsize);
+		SAFE_FREE(ret.dptr);
+		ret.dptr = tmp;
+
+		if (ret.dptr == NULL) {
+			return make_tdb_data(NULL, 0);
+		}
+		return ret;
+	}
+	
+	status = idmap_tdb2_open_perm_db();
+	if (!NT_STATUS_IS_OK(status)) {
+		return ret;
+	}
+
+	/* fetch from the permanent tdb */
+	return dbwrap_fetch_bystring(idmap_tdb2_perm, mem_ctx, keystr);
+}
+
+/*
+  store into both databases
+ */
+static NTSTATUS tdb2_store_bystring(const char *keystr, TDB_DATA data, int flags)
+{
+	NTSTATUS ret;
+	NTSTATUS status = idmap_tdb2_open_perm_db();
+	if (!NT_STATUS_IS_OK(status)) {
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+	ret = dbwrap_store_bystring(idmap_tdb2_perm, keystr, data, flags);
+	if (!NT_STATUS_IS_OK(ret)) {
+		ret = tdb_store_bystring(idmap_tdb2_tmp, keystr, data, flags) ? NT_STATUS_OK : NT_STATUS_UNSUCCESSFUL;
+	}
+	return ret;
+}
+
+/*
+  delete from both databases
+ */
+static NTSTATUS tdb2_delete_bystring(const char *keystr)
+{
+	NTSTATUS ret;
+	NTSTATUS status = idmap_tdb2_open_perm_db();
+	if (!NT_STATUS_IS_OK(status)) {
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+	ret = dbwrap_delete_bystring(idmap_tdb2_perm, keystr);
+	if (!NT_STATUS_IS_OK(ret)) {
+		ret = tdb_delete_bystring(idmap_tdb2_tmp, keystr)  ? NT_STATUS_OK : NT_STATUS_UNSUCCESSFUL;
+	}
+	return ret;
+}
+
+/*
+  Initialise idmap database. 
+*/
+static NTSTATUS idmap_tdb2_db_init(struct idmap_domain *dom)
+{
+	NTSTATUS ret;
+	struct idmap_tdb2_context *ctx;
+	char *config_option = NULL;
+	const char *range;
+	NTSTATUS status;
+
+	status = idmap_tdb2_open_cache_db();
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	ctx = talloc(dom, struct idmap_tdb2_context);
+	if ( ! ctx) {
+		DEBUG(0, ("Out of memory!\n"));
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	config_option = talloc_asprintf(ctx, "idmap config %s", dom->name);
+	if ( ! config_option) {
+		DEBUG(0, ("Out of memory!\n"));
+		ret = NT_STATUS_NO_MEMORY;
+		goto failed;
+	}
+
+	range = lp_parm_const_string(-1, config_option, "range", NULL);
+	if (( ! range) ||
+	    (sscanf(range, "%u - %u", &ctx->filter_low_id, &ctx->filter_high_id) != 2) ||
+	    (ctx->filter_low_id > ctx->filter_high_id)) {
+		ctx->filter_low_id = 0;
+		ctx->filter_high_id = 0;
+	}
+
+	dom->private_data = ctx;
+	dom->initialized = True;
+
+	talloc_free(config_option);
+	return NT_STATUS_OK;
+
+failed:
+	talloc_free(ctx);
+	return ret;
+}
+
+
+/*
+  run a script to perform a mapping
+
+  The script should the following command lines:
+
+      SIDTOID S-1-xxxx
+      IDTOSID UID xxxx
+      IDTOSID GID xxxx
+
+  and should return one of the following as a single line of text
+     UID:xxxx
+     GID:xxxx
+     SID:xxxx
+     ERR:xxxx
+ */
+static NTSTATUS idmap_tdb2_script(struct idmap_tdb2_context *ctx, struct id_map *map,
+				  const char *fmt, ...)
+{
+	va_list ap;
+	char *cmd;
+	FILE *p;
+	char line[64];
+	unsigned long v;
+
+	cmd = talloc_asprintf(ctx, "%s ", idmap_tdb2_state.idmap_script);
+	NT_STATUS_HAVE_NO_MEMORY(cmd);	
+
+	va_start(ap, fmt);
+	cmd = talloc_vasprintf_append(cmd, fmt, ap);
+	va_end(ap);
+	NT_STATUS_HAVE_NO_MEMORY(cmd);
+
+	p = popen(cmd, "r");
+	talloc_free(cmd);
+	if (p == NULL) {
+		return NT_STATUS_NONE_MAPPED;
+	}
+
+	if (fgets(line, sizeof(line)-1, p) == NULL) {
+		pclose(p);
+		return NT_STATUS_NONE_MAPPED;
+	}
+	pclose(p);
+
+	DEBUG(10,("idmap script gave: %s\n", line));
+
+	if (sscanf(line, "UID:%lu", &v) == 1) {
+		map->xid.id   = v;
+		map->xid.type = ID_TYPE_UID;
+	} else if (sscanf(line, "GID:%lu", &v) == 1) {
+		map->xid.id   = v;
+		map->xid.type = ID_TYPE_GID;		
+	} else if (strncmp(line, "SID:S-", 6) == 0) {
+		if (!string_to_sid(map->sid, &line[4])) {
+			DEBUG(0,("Bad SID in '%s' from idmap script %s\n",
+				 line, idmap_tdb2_state.idmap_script));
+			return NT_STATUS_NONE_MAPPED;			
+		}
+	} else {
+		DEBUG(0,("Bad reply '%s' from idmap script %s\n",
+			 line, idmap_tdb2_state.idmap_script));
+		return NT_STATUS_NONE_MAPPED;
+	}
+
+	return NT_STATUS_OK;
+}
+
+
+
+/*
+  Single id to sid lookup function. 
+*/
+static NTSTATUS idmap_tdb2_id_to_sid(struct idmap_tdb2_context *ctx, struct id_map *map)
+{
+	NTSTATUS ret;
+	TDB_DATA data;
+	char *keystr;
+
+	if (!ctx || !map) {
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	/* apply filters before checking */
+	if ((ctx->filter_low_id && (map->xid.id < ctx->filter_low_id)) ||
+	    (ctx->filter_high_id && (map->xid.id > ctx->filter_high_id))) {
+		DEBUG(5, ("Requested id (%u) out of range (%u - %u). Filtered!\n",
+				map->xid.id, ctx->filter_low_id, ctx->filter_high_id));
+		return NT_STATUS_NONE_MAPPED;
+	}
+
+	switch (map->xid.type) {
+
+	case ID_TYPE_UID:
+		keystr = talloc_asprintf(ctx, "UID %lu", (unsigned long)map->xid.id);
+		break;
+		
+	case ID_TYPE_GID:
+		keystr = talloc_asprintf(ctx, "GID %lu", (unsigned long)map->xid.id);
+		break;
+
+	default:
+		DEBUG(2, ("INVALID unix ID type: 0x02%x\n", map->xid.type));
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	/* final SAFE_FREE safe */
+	data.dptr = NULL;
+
+	if (keystr == NULL) {
+		DEBUG(0, ("Out of memory!\n"));
+		ret = NT_STATUS_NO_MEMORY;
+		goto done;
+	}
+
+	DEBUG(10,("Fetching record %s\n", keystr));
+
+	/* Check if the mapping exists */
+	data = tdb2_fetch_bystring(keystr, keystr);
+
+	if (!data.dptr) {
+		fstring sidstr;
+
+		DEBUG(10,("Record %s not found\n", keystr));
+		if (idmap_tdb2_state.idmap_script == NULL) {
+			ret = NT_STATUS_NONE_MAPPED;
+			goto done;
+		}
+
+		ret = idmap_tdb2_script(ctx, map, "IDTOSID %s", keystr);
+
+		/* store it on shared storage */
+		if (!NT_STATUS_IS_OK(ret)) {
+			goto done;
+		}
+
+		if (sid_to_fstring(sidstr, map->sid)) {
+			/* both forward and reverse mappings */
+			tdb2_store_bystring(keystr,
+					    string_term_tdb_data(sidstr), 
+					    TDB_REPLACE);
+			tdb2_store_bystring(sidstr,
+					    string_term_tdb_data(keystr), 
+					    TDB_REPLACE);
+		}
+		goto done;
+	}
+		
+	if (!string_to_sid(map->sid, (const char *)data.dptr)) {
+		DEBUG(10,("INVALID SID (%s) in record %s\n",
+			(const char *)data.dptr, keystr));
+		ret = NT_STATUS_INTERNAL_DB_ERROR;
+		goto done;
+	}
+
+	DEBUG(10,("Found record %s -> %s\n", keystr, (const char *)data.dptr));
+	ret = NT_STATUS_OK;
+
+done:
+	talloc_free(keystr);
+	return ret;
+}
+
+
+/*
+ Single sid to id lookup function. 
+*/
+static NTSTATUS idmap_tdb2_sid_to_id(struct idmap_tdb2_context *ctx, struct id_map *map)
+{
+	NTSTATUS ret;
+	TDB_DATA data;
+	char *keystr;
+	unsigned long rec_id = 0;
+
+	if ((keystr = sid_string_talloc(ctx, map->sid)) == NULL) {
+		DEBUG(0, ("Out of memory!\n"));
+		ret = NT_STATUS_NO_MEMORY;
+		goto done;
+	}
+
+	DEBUG(10,("Fetching record %s\n", keystr));
+
+	/* Check if sid is present in database */
+	data = tdb2_fetch_bystring(keystr, keystr);
+	if (!data.dptr) {
+		fstring idstr;
+
+		DEBUG(10,(__location__ " Record %s not found\n", keystr));
+
+		if (idmap_tdb2_state.idmap_script == NULL) {
+			ret = NT_STATUS_NONE_MAPPED;
+			goto done;
+		}
+			
+		ret = idmap_tdb2_script(ctx, map, "SIDTOID %s", keystr);
+		/* store it on shared storage */
+		if (!NT_STATUS_IS_OK(ret)) {
+			goto done;
+		}
+
+		snprintf(idstr, sizeof(idstr), "%cID %lu", 
+			 map->xid.type == ID_TYPE_UID?'U':'G',
+			 (unsigned long)map->xid.id);
+		/* store both forward and reverse mappings */
+		tdb2_store_bystring(keystr, string_term_tdb_data(idstr),
+				    TDB_REPLACE);
+		tdb2_store_bystring(idstr, string_term_tdb_data(keystr),
+				    TDB_REPLACE);
+		goto done;
+	}
+
+	/* What type of record is this ? */
+	if (sscanf((const char *)data.dptr, "UID %lu", &rec_id) == 1) { /* Try a UID record. */
+		map->xid.id = rec_id;
+		map->xid.type = ID_TYPE_UID;
+		DEBUG(10,("Found uid record %s -> %s \n", keystr, (const char *)data.dptr ));
+		ret = NT_STATUS_OK;
+
+	} else if (sscanf((const char *)data.dptr, "GID %lu", &rec_id) == 1) { /* Try a GID record. */
+		map->xid.id = rec_id;
+		map->xid.type = ID_TYPE_GID;
+		DEBUG(10,("Found gid record %s -> %s \n", keystr, (const char *)data.dptr ));
+		ret = NT_STATUS_OK;
+
+	} else { /* Unknown record type ! */
+		DEBUG(2, ("Found INVALID record %s -> %s\n", keystr, (const char *)data.dptr));
+		ret = NT_STATUS_INTERNAL_DB_ERROR;
+	}
+	
+	/* apply filters before returning result */
+	if ((ctx->filter_low_id && (map->xid.id < ctx->filter_low_id)) ||
+	    (ctx->filter_high_id && (map->xid.id > ctx->filter_high_id))) {
+		DEBUG(5, ("Requested id (%u) out of range (%u - %u). Filtered!\n",
+				map->xid.id, ctx->filter_low_id, ctx->filter_high_id));
+		ret = NT_STATUS_NONE_MAPPED;
+	}
+
+done:
+	talloc_free(keystr);
+	return ret;
+}
+
+/*
+  lookup a set of unix ids. 
+*/
+static NTSTATUS idmap_tdb2_unixids_to_sids(struct idmap_domain *dom, struct id_map **ids)
+{
+	struct idmap_tdb2_context *ctx;
+	NTSTATUS ret;
+	int i;
+
+	/* make sure we initialized */
+	if ( ! dom->initialized) {
+		ret = idmap_tdb2_db_init(dom);
+		if ( ! NT_STATUS_IS_OK(ret)) {
+			return ret;
+		}
+	}
+
+	ctx = talloc_get_type(dom->private_data, struct idmap_tdb2_context);
+
+	for (i = 0; ids[i]; i++) {
+		ret = idmap_tdb2_id_to_sid(ctx, ids[i]);
+		if ( ! NT_STATUS_IS_OK(ret)) {
+
+			/* if it is just a failed mapping continue */
+			if (NT_STATUS_EQUAL(ret, NT_STATUS_NONE_MAPPED)) {
+
+				/* make sure it is marked as unmapped */
+				ids[i]->status = ID_UNMAPPED;
+				continue;
+			}
+			
+			/* some fatal error occurred, return immediately */
+			goto done;
+		}
+
+		/* all ok, id is mapped */
+		ids[i]->status = ID_MAPPED;
+	}
+
+	ret = NT_STATUS_OK;
+
+done:
+	return ret;
+}
+
+/*
+  lookup a set of sids. 
+*/
+static NTSTATUS idmap_tdb2_sids_to_unixids(struct idmap_domain *dom, struct id_map **ids)
+{
+	struct idmap_tdb2_context *ctx;
+	NTSTATUS ret;
+	int i;
+
+	/* make sure we initialized */
+	if ( ! dom->initialized) {
+		ret = idmap_tdb2_db_init(dom);
+		if ( ! NT_STATUS_IS_OK(ret)) {
+			return ret;
+		}
+	}
+
+	ctx = talloc_get_type(dom->private_data, struct idmap_tdb2_context);
+
+	for (i = 0; ids[i]; i++) {
+		ret = idmap_tdb2_sid_to_id(ctx, ids[i]);
+		if ( ! NT_STATUS_IS_OK(ret)) {
+
+			/* if it is just a failed mapping continue */
+			if (NT_STATUS_EQUAL(ret, NT_STATUS_NONE_MAPPED)) {
+
+				/* make sure it is marked as unmapped */
+				ids[i]->status = ID_UNMAPPED;
+				continue;
+			}
+			
+			/* some fatal error occurred, return immediately */
+			goto done;
+		}
+
+		/* all ok, id is mapped */
+		ids[i]->status = ID_MAPPED;
+	}
+
+	ret = NT_STATUS_OK;
+
+done:
+	return ret;
+}
+
+
+/*
+  set a mapping. 
+*/
+static NTSTATUS idmap_tdb2_set_mapping(struct idmap_domain *dom, const struct id_map *map)
+{
+	struct idmap_tdb2_context *ctx;
+	NTSTATUS ret;
+	TDB_DATA data;
+	char *ksidstr, *kidstr;
+	struct db_record *update_lock = NULL;
+	struct db_record *rec = NULL;
+
+	/* make sure we initialized */
+	if ( ! dom->initialized) {
+		ret = idmap_tdb2_db_init(dom);
+		if ( ! NT_STATUS_IS_OK(ret)) {
+			return ret;
+		}
+	}
+
+	if (!map || !map->sid) {
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	ksidstr = kidstr = NULL;
+	data.dptr = NULL;
+
+	/* TODO: should we filter a set_mapping using low/high filters ? */
+	
+	ctx = talloc_get_type(dom->private_data, struct idmap_tdb2_context);
+
+	switch (map->xid.type) {
+
+	case ID_TYPE_UID:
+		kidstr = talloc_asprintf(ctx, "UID %lu", (unsigned long)map->xid.id);
+		break;
+		
+	case ID_TYPE_GID:
+		kidstr = talloc_asprintf(ctx, "GID %lu", (unsigned long)map->xid.id);
+		break;
+
+	default:
+		DEBUG(2, ("INVALID unix ID type: 0x02%x\n", map->xid.type));
+		return NT_STATUS_INVALID_PARAMETER;
+	}
+
+	if (kidstr == NULL) {
+		DEBUG(0, ("ERROR: Out of memory!\n"));
+		ret = NT_STATUS_NO_MEMORY;
+		goto done;
+	}
+
+	if (!(ksidstr = sid_string_talloc(ctx, map->sid))) {
+		DEBUG(0, ("Out of memory!\n"));
+		ret = NT_STATUS_NO_MEMORY;
+		goto done;
+	}
+
+	DEBUG(10, ("Storing %s <-> %s map\n", ksidstr, kidstr));
+
+	/*
+	 * Get us the update lock. This is necessary to get the lock orders
+	 * right, we need to deal with two records under a lock.
+	 */
+
+	if (!(update_lock = idmap_tdb2_perm->fetch_locked(
+		      idmap_tdb2_perm, ctx,
+		      string_term_tdb_data("UPDATELOCK")))) {
+		DEBUG(10,("Failed to lock record %s\n", ksidstr));
+		ret = NT_STATUS_UNSUCCESSFUL;
+		goto done;
+	}
+
+	/*
+	 * *DELETE* previous mappings if any. *
+	 */
+
+	/* First delete indexed on SID */
+
+	if (((rec = idmap_tdb2_perm->fetch_locked(
+		     idmap_tdb2_perm, update_lock,
+		     string_term_tdb_data(ksidstr))) != NULL)
+	    && (rec->value.dsize != 0)) {
+		struct db_record *rec2;
+
+		if ((rec2 = idmap_tdb2_perm->fetch_locked(
+			     idmap_tdb2_perm, update_lock, rec->value))
+		    != NULL) {
+			rec2->delete_rec(rec2);
+			TALLOC_FREE(rec2);
+		}
+
+		rec->delete_rec(rec);
+
+		tdb_delete(idmap_tdb2_tmp, rec->key);
+		tdb_delete(idmap_tdb2_tmp, rec->value);
+	}
+	TALLOC_FREE(rec);
+
+	/* Now delete indexed on unix ID */
+
+	if (((rec = idmap_tdb2_perm->fetch_locked(
+		     idmap_tdb2_perm, update_lock,
+		     string_term_tdb_data(kidstr))) != NULL)
+	    && (rec->value.dsize != 0)) {
+		struct db_record *rec2;
+
+		if ((rec2 = idmap_tdb2_perm->fetch_locked(
+			     idmap_tdb2_perm, update_lock, rec->value))
+		    != NULL) {
+			rec2->delete_rec(rec2);
+			TALLOC_FREE(rec2);
+		}
+
+		rec->delete_rec(rec);
+
+		tdb_delete(idmap_tdb2_tmp, rec->key);
+		tdb_delete(idmap_tdb2_tmp, rec->value);
+	}
+	TALLOC_FREE(rec);
+
+	if (!NT_STATUS_IS_OK(tdb2_store_bystring(ksidstr, string_term_tdb_data(kidstr),
+				TDB_INSERT))) {
+		DEBUG(0, ("Error storing SID -> ID\n"));
+		ret = NT_STATUS_UNSUCCESSFUL;
+		goto done;
+	}
+	if (!NT_STATUS_IS_OK(tdb2_store_bystring(kidstr, string_term_tdb_data(ksidstr),
+				TDB_INSERT))) {
+		DEBUG(0, ("Error storing ID -> SID\n"));
+		/* try to remove the previous stored SID -> ID map */
+		tdb2_delete_bystring(ksidstr);
+		ret = NT_STATUS_UNSUCCESSFUL;
+		goto done;
+	}
+
+	DEBUG(10,("Stored %s <-> %s\n", ksidstr, kidstr));
+	ret = NT_STATUS_OK;
+
+done:
+	talloc_free(ksidstr);
+	talloc_free(kidstr);
+	SAFE_FREE(data.dptr);
+	TALLOC_FREE(update_lock);
+	return ret;
+}
+
+/*
+  remove a mapping. 
+*/
+static NTSTATUS idmap_tdb2_remove_mapping(struct idmap_domain *dom, const struct id_map *map)
+{
+	/* not supported as it would invalidate the cache tdb on other
+	   nodes */
+	DEBUG(0,("idmap_tdb2_remove_mapping not supported\n"));
+	return NT_STATUS_NOT_SUPPORTED;
+}
+
+/*
+  Close the idmap tdb instance
+*/
+static NTSTATUS idmap_tdb2_close(struct idmap_domain *dom)
+{
+	/* don't do anything */
+	return NT_STATUS_OK;
+}
+
+
+/*
+  Dump all mappings out
+*/
+static NTSTATUS idmap_tdb2_dump_data(struct idmap_domain *dom, struct id_map **maps, int *num_maps)
+{
+	DEBUG(0,("idmap_tdb2_dump_data not supported\n"));
+	return NT_STATUS_NOT_SUPPORTED;
+}
+
+static struct idmap_methods db_methods = {
+	.init            = idmap_tdb2_db_init,
+	.unixids_to_sids = idmap_tdb2_unixids_to_sids,
+	.sids_to_unixids = idmap_tdb2_sids_to_unixids,
+	.set_mapping     = idmap_tdb2_set_mapping,
+	.remove_mapping  = idmap_tdb2_remove_mapping,
+	.dump_data       = idmap_tdb2_dump_data,
+	.close_fn        = idmap_tdb2_close
+};
+
+static struct idmap_alloc_methods db_alloc_methods = {
+	.init        = idmap_tdb2_alloc_init,
+	.allocate_id = idmap_tdb2_allocate_id,
+	.get_id_hwm  = idmap_tdb2_get_hwm,
+	.set_id_hwm  = idmap_tdb2_set_hwm,
+	.close_fn    = idmap_tdb2_alloc_close
+};
+
+NTSTATUS idmap_tdb2_init(void)
+{
+	NTSTATUS ret;
+
+	/* register both backends */
+	ret = smb_register_idmap_alloc(SMB_IDMAP_INTERFACE_VERSION, "tdb2", &db_alloc_methods);
+	if (! NT_STATUS_IS_OK(ret)) {
+		DEBUG(0, ("Unable to register idmap alloc tdb2 module: %s\n", get_friendly_nt_error_msg(ret)));
+		return ret;
+	}
+
+	return smb_register_idmap(SMB_IDMAP_INTERFACE_VERSION, "tdb2", &db_methods);
+}