s3-talloc Change TALLOC_ARRAY() to talloc_array()
[samba.git] / source3 / locking / locking.c
index b6e2ced3360fee4febac39b080d5b8947ccb99f2..55412ec8b2aa8c702db98b739d6e6fe2f64b276c 100644 (file)
@@ -2,21 +2,21 @@
    Unix SMB/CIFS implementation.
    Locking functions
    Copyright (C) Andrew Tridgell 1992-2000
-   Copyright (C) Jeremy Allison 1992-2000
-   
+   Copyright (C) Jeremy Allison 1992-2006
+   Copyright (C) Volker Lendecke 2005
+
    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
+   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, write to the Free Software
-   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
    Revision History:
 
    rewrtten completely to use new tdb code. Tridge, Dec '99
 
    Added POSIX locking support. Jeremy Allison (jeremy@valinux.com), Apr. 2000.
+   Added Unix Extensions POSIX locking support. Jeremy Allison Mar 2006.
 */
 
 #include "includes.h"
-uint16 global_smbpid;
+#include "system/filesys.h"
+#include "locking/proto.h"
+#include "smbd/globals.h"
+#include "dbwrap.h"
+#include "../libcli/security/security.h"
+#include "serverid.h"
+#include "messages.h"
+#include "util_tdb.h"
+
+#undef DBGC_CLASS
+#define DBGC_CLASS DBGC_LOCKING
+
+#define NO_LOCKING_COUNT (-1)
 
 /* the locking database handle */
-static TDB_CONTEXT *tdb;
-
-struct locking_data {
-        union {
-                int num_share_mode_entries;
-                share_mode_entry dummy; /* Needed for alignment. */
-        } u;
-        /* the following two entries are implicit
-           share_mode_entry modes[num_share_mode_entries];
-           char file_name[];
-        */
-};
+static struct db_context *lock_db;
 
 /****************************************************************************
- Debugging aid :-).
+ Debugging aids :-).
 ****************************************************************************/
 
-static const char *lock_type_name(enum brl_type lock_type)
+const char *lock_type_name(enum brl_type lock_type)
 {
-       return (lock_type == READ_LOCK) ? "READ" : "WRITE";
+       switch (lock_type) {
+               case READ_LOCK:
+                       return "READ";
+               case WRITE_LOCK:
+                       return "WRITE";
+               case PENDING_READ_LOCK:
+                       return "PENDING_READ";
+               case PENDING_WRITE_LOCK:
+                       return "PENDING_WRITE";
+               default:
+                       return "other";
+       }
+}
+
+const char *lock_flav_name(enum brl_flavour lock_flav)
+{
+       return (lock_flav == WINDOWS_LOCK) ? "WINDOWS_LOCK" : "POSIX_LOCK";
 }
 
 /****************************************************************************
  Utility function called to see if a file region is locked.
+ Called in the read/write codepath.
 ****************************************************************************/
 
-BOOL is_locked(files_struct *fsp,connection_struct *conn,
-              SMB_BIG_UINT count,SMB_BIG_UINT offset, 
-              enum brl_type lock_type)
+void init_strict_lock_struct(files_struct *fsp,
+                               uint64_t smblctx,
+                               br_off start,
+                               br_off size,
+                               enum brl_type lock_type,
+                               struct lock_struct *plock)
 {
-       int snum = SNUM(conn);
-       BOOL ret;
-       
-       if (count == 0)
-               return(False);
+       SMB_ASSERT(lock_type == READ_LOCK || lock_type == WRITE_LOCK);
+
+       plock->context.smblctx = smblctx;
+        plock->context.tid = fsp->conn->cnum;
+        plock->context.pid = sconn_server_id(fsp->conn->sconn);
+        plock->start = start;
+        plock->size = size;
+        plock->fnum = fsp->fnum;
+        plock->lock_type = lock_type;
+        plock->lock_flav = lp_posix_cifsu_locktype(fsp);
+}
 
-       if (!lp_locking(snum) || !lp_strict_locking(snum))
-               return(False);
+bool strict_lock_default(files_struct *fsp, struct lock_struct *plock)
+{
+       int strict_locking = lp_strict_locking(fsp->conn->params);
+       bool ret = False;
 
-       ret = !brl_locktest(fsp->dev, fsp->inode, fsp->fnum,
-                            global_smbpid, sys_getpid(), conn->cnum, 
-                            offset, count, lock_type);
+       if (plock->size == 0) {
+               return True;
+       }
 
-       DEBUG(10,("is_locked: brl start=%.0f len=%.0f %s for file %s\n",
-                       (double)offset, (double)count, ret ? "locked" : "unlocked",
-                       fsp->fsp_name ));
+       if (!lp_locking(fsp->conn->params) || !strict_locking) {
+               return True;
+       }
 
-       /*
-        * There is no lock held by an SMB daemon, check to
-        * see if there is a POSIX lock from a UNIX or NFS process.
-        */
+       if (strict_locking == Auto) {
+               if  (EXCLUSIVE_OPLOCK_TYPE(fsp->oplock_type) && (plock->lock_type == READ_LOCK || plock->lock_type == WRITE_LOCK)) {
+                       DEBUG(10,("is_locked: optimisation - exclusive oplock on file %s\n", fsp_str_dbg(fsp)));
+                       ret = True;
+               } else if ((fsp->oplock_type == LEVEL_II_OPLOCK) &&
+                          (plock->lock_type == READ_LOCK)) {
+                       DEBUG(10,("is_locked: optimisation - level II oplock on file %s\n", fsp_str_dbg(fsp)));
+                       ret = True;
+               } else {
+                       struct byte_range_lock *br_lck;
 
-       if(!ret && lp_posix_locking(snum)) {
-               ret = is_posix_locked(fsp, offset, count, lock_type);
+                       br_lck = brl_get_locks_readonly(fsp);
+                       if (!br_lck) {
+                               return True;
+                       }
+                       ret = brl_locktest(br_lck,
+                                       plock->context.smblctx,
+                                       plock->context.pid,
+                                       plock->start,
+                                       plock->size,
+                                       plock->lock_type,
+                                       plock->lock_flav);
+               }
+       } else {
+               struct byte_range_lock *br_lck;
 
-               DEBUG(10,("is_locked: posix start=%.0f len=%.0f %s for file %s\n",
-                               (double)offset, (double)count, ret ? "locked" : "unlocked",
-                               fsp->fsp_name ));
+               br_lck = brl_get_locks_readonly(fsp);
+               if (!br_lck) {
+                       return True;
+               }
+               ret = brl_locktest(br_lck,
+                               plock->context.smblctx,
+                               plock->context.pid,
+                               plock->start,
+                               plock->size,
+                               plock->lock_type,
+                               plock->lock_flav);
        }
 
+       DEBUG(10,("strict_lock_default: flavour = %s brl start=%.0f "
+                       "len=%.0f %s for fnum %d file %s\n",
+                       lock_flav_name(plock->lock_flav),
+                       (double)plock->start, (double)plock->size,
+                       ret ? "unlocked" : "locked",
+                       plock->fnum, fsp_str_dbg(fsp)));
+
        return ret;
 }
 
+void strict_unlock_default(files_struct *fsp, struct lock_struct *plock)
+{
+}
+
 /****************************************************************************
Utility function called by locking requests.
Find out if a lock could be granted - return who is blocking us if we can't.
 ****************************************************************************/
 
-static NTSTATUS do_lock(files_struct *fsp,connection_struct *conn, uint16 lock_pid,
-                SMB_BIG_UINT count,SMB_BIG_UINT offset,enum brl_type lock_type, BOOL *my_lock_ctx)
+NTSTATUS query_lock(files_struct *fsp,
+                       uint64_t *psmblctx,
+                       uint64_t *pcount,
+                       uint64_t *poffset,
+                       enum brl_type *plock_type,
+                       enum brl_flavour lock_flav)
 {
-       NTSTATUS status = NT_STATUS_LOCK_NOT_GRANTED;
+       struct byte_range_lock *br_lck = NULL;
+
+       if (!fsp->can_lock) {
+               return fsp->is_directory ? NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
+       }
 
-       if (!lp_locking(SNUM(conn)))
+       if (!lp_locking(fsp->conn->params)) {
                return NT_STATUS_OK;
+       }
 
-       /* NOTE! 0 byte long ranges ARE allowed and should be stored  */
+       br_lck = brl_get_locks_readonly(fsp);
+       if (!br_lck) {
+               return NT_STATUS_NO_MEMORY;
+       }
 
-       DEBUG(10,("do_lock: lock type %s start=%.0f len=%.0f requested for file %s\n",
-                 lock_type_name(lock_type), (double)offset, (double)count, fsp->fsp_name ));
-
-       if (OPEN_FSP(fsp) && fsp->can_lock && (fsp->conn == conn)) {
-               status = brl_lock(fsp->dev, fsp->inode, fsp->fnum,
-                                 lock_pid, sys_getpid(), conn->cnum, 
-                                 offset, count, 
-                                 lock_type, my_lock_ctx);
-
-               if (NT_STATUS_IS_OK(status) && lp_posix_locking(SNUM(conn))) {
-
-                       /*
-                        * Try and get a POSIX lock on this range.
-                        * Note that this is ok if it is a read lock
-                        * overlapping on a different fd. JRA.
-                        */
-
-                       if (!set_posix_lock(fsp, offset, count, lock_type)) {
-                               if (errno == EACCES || errno == EAGAIN)
-                                       status = NT_STATUS_FILE_LOCK_CONFLICT;
-                               else
-                                       status = map_nt_error_from_unix(errno);
-
-                               /*
-                                * We failed to map - we must now remove the brl
-                                * lock entry.
-                                */
-                               (void)brl_unlock(fsp->dev, fsp->inode, fsp->fnum,
-                                                               lock_pid, sys_getpid(), conn->cnum, 
-                                                               offset, count, False,
-                                                               NULL, NULL);
-                       }
-               }
+       return brl_lockquery(br_lck,
+                       psmblctx,
+                       sconn_server_id(fsp->conn->sconn),
+                       poffset,
+                       pcount,
+                       plock_type,
+                       lock_flav);
+}
+
+static void increment_current_lock_count(files_struct *fsp,
+    enum brl_flavour lock_flav)
+{
+       if (lock_flav == WINDOWS_LOCK &&
+           fsp->current_lock_count != NO_LOCKING_COUNT) {
+               /* blocking ie. pending, locks also count here,
+                * as this is an efficiency counter to avoid checking
+                * the lock db. on close. JRA. */
+
+               fsp->current_lock_count++;
+       } else {
+               /* Notice that this has had a POSIX lock request.
+                * We can't count locks after this so forget them.
+                */
+               fsp->current_lock_count = NO_LOCKING_COUNT;
        }
+}
 
-       return status;
+static void decrement_current_lock_count(files_struct *fsp,
+    enum brl_flavour lock_flav)
+{
+       if (lock_flav == WINDOWS_LOCK &&
+           fsp->current_lock_count != NO_LOCKING_COUNT) {
+               SMB_ASSERT(fsp->current_lock_count > 0);
+               fsp->current_lock_count--;
+       }
 }
 
 /****************************************************************************
- Utility function called by locking requests. This is *DISGUSTING*. It also
- appears to be "What Windows Does" (tm). Andrew, ever wonder why Windows 2000
- is so slow on the locking tests...... ? This is the reason. Much though I hate
- it, we need this. JRA.
+ Utility function called by locking requests.
 ****************************************************************************/
 
-NTSTATUS do_lock_spin(files_struct *fsp,connection_struct *conn, uint16 lock_pid,
-                SMB_BIG_UINT count,SMB_BIG_UINT offset,enum brl_type lock_type, BOOL *my_lock_ctx)
+struct byte_range_lock *do_lock(struct messaging_context *msg_ctx,
+                       files_struct *fsp,
+                       uint64_t smblctx,
+                       uint64_t count,
+                       uint64_t offset,
+                       enum brl_type lock_type,
+                       enum brl_flavour lock_flav,
+                       bool blocking_lock,
+                       NTSTATUS *perr,
+                       uint64_t *psmblctx,
+                       struct blocking_lock_record *blr)
 {
-       int j, maxj = lp_lock_spin_count();
-       int sleeptime = lp_lock_sleep_time();
-       NTSTATUS status, ret;
+       struct byte_range_lock *br_lck = NULL;
 
-       if (maxj <= 0)
-               maxj = 1;
+       /* silently return ok on print files as we don't do locking there */
+       if (fsp->print_file) {
+               *perr = NT_STATUS_OK;
+               return NULL;
+       }
 
-       ret = NT_STATUS_OK; /* to keep dumb compilers happy */
+       if (!fsp->can_lock) {
+               *perr = fsp->is_directory ? NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
+               return NULL;
+       }
 
-       for (j = 0; j < maxj; j++) {
-               status = do_lock(fsp, conn, lock_pid, count, offset, lock_type, my_lock_ctx);
-               if (!NT_STATUS_EQUAL(status, NT_STATUS_LOCK_NOT_GRANTED) &&
-                   !NT_STATUS_EQUAL(status, NT_STATUS_FILE_LOCK_CONFLICT)) {
-                       return status;
-               }
-               /* if we do fail then return the first error code we got */
-               if (j == 0) {
-                       ret = status;
-                       /* Don't spin if we blocked ourselves. */
-                       if (*my_lock_ctx)
-                               return ret;
-               }
-               if (sleeptime)
-                       sys_usleep(sleeptime);
+       if (!lp_locking(fsp->conn->params)) {
+               *perr = NT_STATUS_OK;
+               return NULL;
        }
-       return ret;
-}
 
-/* Struct passed to brl_unlock. */
-struct posix_unlock_data_struct {
-       files_struct *fsp;
-       SMB_BIG_UINT offset;
-       SMB_BIG_UINT count;
-};
+       /* NOTE! 0 byte long ranges ARE allowed and should be stored  */
 
-/****************************************************************************
- Function passed to brl_unlock to allow POSIX unlock to be done first.
-****************************************************************************/
+       DEBUG(10,("do_lock: lock flavour %s lock type %s start=%.0f len=%.0f "
+               "blocking_lock=%s requested for fnum %d file %s\n",
+               lock_flav_name(lock_flav), lock_type_name(lock_type),
+               (double)offset, (double)count, blocking_lock ? "true" :
+               "false", fsp->fnum, fsp_str_dbg(fsp)));
 
-static void posix_unlock(void *pre_data)
-{
-       struct posix_unlock_data_struct *pdata = (struct posix_unlock_data_struct *)pre_data;
+       br_lck = brl_get_locks(talloc_tos(), fsp);
+       if (!br_lck) {
+               *perr = NT_STATUS_NO_MEMORY;
+               return NULL;
+       }
 
-       if (lp_posix_locking(SNUM(pdata->fsp->conn)))
-               release_posix_lock(pdata->fsp, pdata->offset, pdata->count);
+       *perr = brl_lock(msg_ctx,
+                       br_lck,
+                       smblctx,
+                       sconn_server_id(fsp->conn->sconn),
+                       offset,
+                       count,
+                       lock_type,
+                       lock_flav,
+                       blocking_lock,
+                       psmblctx,
+                       blr);
+
+       DEBUG(10, ("do_lock: returning status=%s\n", nt_errstr(*perr)));
+
+       increment_current_lock_count(fsp, lock_flav);
+       return br_lck;
 }
 
 /****************************************************************************
  Utility function called by unlocking requests.
 ****************************************************************************/
 
-NTSTATUS do_unlock(files_struct *fsp,connection_struct *conn, uint16 lock_pid,
-                  SMB_BIG_UINT count,SMB_BIG_UINT offset)
+NTSTATUS do_unlock(struct messaging_context *msg_ctx,
+                       files_struct *fsp,
+                       uint64_t smblctx,
+                       uint64_t count,
+                       uint64_t offset,
+                       enum brl_flavour lock_flav)
 {
-       BOOL ok = False;
-       struct posix_unlock_data_struct posix_data;
-       
-       if (!lp_locking(SNUM(conn)))
+       bool ok = False;
+       struct byte_range_lock *br_lck = NULL;
+
+       if (!fsp->can_lock) {
+               return fsp->is_directory ? NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
+       }
+
+       if (!lp_locking(fsp->conn->params)) {
                return NT_STATUS_OK;
-       
-       if (!OPEN_FSP(fsp) || !fsp->can_lock || (fsp->conn != conn)) {
-               return NT_STATUS_INVALID_HANDLE;
        }
-       
-       DEBUG(10,("do_unlock: unlock start=%.0f len=%.0f requested for file %s\n",
-                 (double)offset, (double)count, fsp->fsp_name ));
 
-       /*
-        * Remove the existing lock record from the tdb lockdb
-        * before looking at POSIX locks. If this record doesn't
-        * match then don't bother looking to remove POSIX locks.
-        */
+       DEBUG(10,("do_unlock: unlock start=%.0f len=%.0f requested for fnum %d file %s\n",
+                 (double)offset, (double)count, fsp->fnum,
+                 fsp_str_dbg(fsp)));
+
+       br_lck = brl_get_locks(talloc_tos(), fsp);
+       if (!br_lck) {
+               return NT_STATUS_NO_MEMORY;
+       }
+
+       ok = brl_unlock(msg_ctx,
+                       br_lck,
+                       smblctx,
+                       sconn_server_id(fsp->conn->sconn),
+                       offset,
+                       count,
+                       lock_flav);
 
-       posix_data.fsp = fsp;
-       posix_data.offset = offset;
-       posix_data.count = count;
+       TALLOC_FREE(br_lck);
 
-       ok = brl_unlock(fsp->dev, fsp->inode, fsp->fnum,
-                       lock_pid, sys_getpid(), conn->cnum, offset, count,
-                       False, posix_unlock, (void *)&posix_data);
-   
        if (!ok) {
                DEBUG(10,("do_unlock: returning ERRlock.\n" ));
                return NT_STATUS_RANGE_NOT_LOCKED;
        }
+
+       decrement_current_lock_count(fsp, lock_flav);
        return NT_STATUS_OK;
 }
 
 /****************************************************************************
Remove any locks on this fd. Called from file_close().
Cancel any pending blocked locks.
 ****************************************************************************/
 
-void locking_close_file(files_struct *fsp)
+NTSTATUS do_lock_cancel(files_struct *fsp,
+                       uint64 smblctx,
+                       uint64_t count,
+                       uint64_t offset,
+                       enum brl_flavour lock_flav,
+                       struct blocking_lock_record *blr)
 {
-       pid_t pid = sys_getpid();
+       bool ok = False;
+       struct byte_range_lock *br_lck = NULL;
 
-       if (!lp_locking(SNUM(fsp->conn)))
-               return;
+       if (!fsp->can_lock) {
+               return fsp->is_directory ?
+                       NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
+       }
 
-       /*
-        * Just release all the brl locks, no need to release individually.
-        */
+       if (!lp_locking(fsp->conn->params)) {
+               return NT_STATUS_DOS(ERRDOS, ERRcancelviolation);
+       }
 
-       brl_close(fsp->dev, fsp->inode, pid, fsp->conn->cnum, fsp->fnum);
+       DEBUG(10,("do_lock_cancel: cancel start=%.0f len=%.0f requested for fnum %d file %s\n",
+                 (double)offset, (double)count, fsp->fnum,
+                 fsp_str_dbg(fsp)));
 
-       if(lp_posix_locking(SNUM(fsp->conn))) {
+       br_lck = brl_get_locks(talloc_tos(), fsp);
+       if (!br_lck) {
+               return NT_STATUS_NO_MEMORY;
+       }
 
-               /* 
-                * Release all the POSIX locks.
-                */
-               posix_locking_close_file(fsp);
+       ok = brl_lock_cancel(br_lck,
+                       smblctx,
+                       sconn_server_id(fsp->conn->sconn),
+                       offset,
+                       count,
+                       lock_flav,
+                       blr);
+
+       TALLOC_FREE(br_lck);
 
+       if (!ok) {
+               DEBUG(10,("do_lock_cancel: returning ERRcancelviolation.\n" ));
+               return NT_STATUS_DOS(ERRDOS, ERRcancelviolation);
        }
+
+       decrement_current_lock_count(fsp, lock_flav);
+       return NT_STATUS_OK;
 }
 
 /****************************************************************************
Initialise the locking functions.
Remove any locks on this fd. Called from file_close().
 ****************************************************************************/
 
-static int open_read_only;
-
-BOOL locking_init(int read_only)
+void locking_close_file(struct messaging_context *msg_ctx,
+                       files_struct *fsp,
+                       enum file_close_type close_type)
 {
-       brl_init(read_only);
+       struct byte_range_lock *br_lck;
 
-       if (tdb)
-               return True;
+       if (!lp_locking(fsp->conn->params)) {
+               return;
+       }
 
-       tdb = tdb_open_log(lock_path("locking.tdb"), 
-                      0, TDB_DEFAULT|(read_only?0x0:TDB_CLEAR_IF_FIRST), 
-                      read_only?O_RDONLY:O_RDWR|O_CREAT,
-                      0644);
+       /* If we have not outstanding locks or pending
+        * locks then we don't need to look in the lock db.
+        */
 
-       if (!tdb) {
-               DEBUG(0,("ERROR: Failed to initialise locking database\n"));
-               return False;
+       if (fsp->current_lock_count == 0) {
+               return;
        }
-       
-       if (!posix_locking_init(read_only))
-               return False;
 
-       open_read_only = read_only;
+       br_lck = brl_get_locks(talloc_tos(),fsp);
 
-       return True;
+       if (br_lck) {
+               cancel_pending_lock_requests_by_fid(fsp, br_lck, close_type);
+               brl_close_fnum(msg_ctx, br_lck);
+               TALLOC_FREE(br_lck);
+       }
 }
 
-/*******************************************************************
Deinitialize the share_mode management.
-******************************************************************/
+/****************************************************************************
Initialise the locking functions.
+****************************************************************************/
 
-BOOL locking_end(void)
+static bool locking_init_internal(bool read_only)
 {
+       brl_init(read_only);
 
-       brl_shutdown(open_read_only);
-       if (tdb) {
-
-               if (tdb_close(tdb) != 0)
-                       return False;
-       }
+       if (lock_db)
+               return True;
 
-       return True;
-}
+       lock_db = db_open(NULL, lock_path("locking.tdb"),
+                         lp_open_files_db_hash_size(),
+                         TDB_DEFAULT|TDB_VOLATILE|TDB_CLEAR_IF_FIRST|TDB_INCOMPATIBLE_HASH,
+                         read_only?O_RDONLY:O_RDWR|O_CREAT, 0644);
 
-/*******************************************************************
- Form a static locking key for a dev/inode pair.
-******************************************************************/
+       if (!lock_db) {
+               DEBUG(0,("ERROR: Failed to initialise locking database\n"));
+               return False;
+       }
 
-static TDB_DATA locking_key(SMB_DEV_T dev, SMB_INO_T inode)
-{
-       static struct locking_key key;
-       TDB_DATA kbuf;
-
-       memset(&key, '\0', sizeof(key));
-       key.dev = dev;
-       key.inode = inode;
-       kbuf.dptr = (char *)&key;
-       kbuf.dsize = sizeof(key);
-       return kbuf;
-}
+       if (!posix_locking_init(read_only))
+               return False;
 
-static TDB_DATA locking_key_fsp(files_struct *fsp)
-{
-       return locking_key(fsp->dev, fsp->inode);
+       return True;
 }
 
-/*******************************************************************
- Lock a hash bucket entry.
-******************************************************************/
-
-BOOL lock_share_entry(connection_struct *conn,
-                     SMB_DEV_T dev, SMB_INO_T inode)
+bool locking_init(void)
 {
-       return tdb_chainlock(tdb, locking_key(dev, inode)) == 0;
+       return locking_init_internal(false);
 }
 
-/*******************************************************************
- Unlock a hash bucket entry.
-******************************************************************/
-
-void unlock_share_entry(connection_struct *conn,
-                       SMB_DEV_T dev, SMB_INO_T inode)
+bool locking_init_readonly(void)
 {
-       tdb_chainunlock(tdb, locking_key(dev, inode));
+       return locking_init_internal(true);
 }
 
 /*******************************************************************
- Lock a hash bucket entry. use a fsp for convenience
+ Deinitialize the share_mode management.
 ******************************************************************/
 
-BOOL lock_share_entry_fsp(files_struct *fsp)
+bool locking_end(void)
 {
-       return tdb_chainlock(tdb, locking_key(fsp->dev, fsp->inode)) == 0;
+       brl_shutdown();
+       TALLOC_FREE(lock_db);
+       return true;
 }
 
 /*******************************************************************
Unlock a hash bucket entry.
Form a static locking key for a dev/inode pair.
 ******************************************************************/
 
-void unlock_share_entry_fsp(files_struct *fsp)
+static TDB_DATA locking_key(const struct file_id *id, struct file_id *tmp)
 {
-       tdb_chainunlock(tdb, locking_key(fsp->dev, fsp->inode));
+       *tmp = *id;
+       return make_tdb_data((const uint8_t *)tmp, sizeof(*tmp));
 }
 
 /*******************************************************************
  Print out a share mode.
 ********************************************************************/
 
-char *share_mode_str(int num, share_mode_entry *e)
+char *share_mode_str(TALLOC_CTX *ctx, int num, const struct share_mode_entry *e)
 {
-       static pstring share_str;
-
-       slprintf(share_str, sizeof(share_str)-1, "share_mode_entry[%d]: \
-pid = %lu, share_mode = 0x%x, desired_access = 0x%x, port = 0x%x, type= 0x%x, file_id = %lu, dev = 0x%x, inode = %.0f",
-       num, (unsigned long)e->pid, e->share_mode, (unsigned int)e->desired_access, e->op_port, e->op_type, e->share_file_id,
-       (unsigned int)e->dev, (double)e->inode );
-
-       return share_str;
+       return talloc_asprintf(ctx, "share_mode_entry[%d]: %s "
+                "pid = %s, share_access = 0x%x, private_options = 0x%x, "
+                "access_mask = 0x%x, mid = 0x%llx, type= 0x%x, gen_id = %lu, "
+                "uid = %u, flags = %u, file_id %s, name_hash = 0x%x",
+                num,
+                e->op_type == UNUSED_SHARE_MODE_ENTRY ? "UNUSED" : "",
+                procid_str_static(&e->pid),
+                e->share_access, e->private_options,
+                e->access_mask, (unsigned long long)e->op_mid,
+                e->op_type, e->share_file_id,
+                (unsigned int)e->uid, (unsigned int)e->flags,
+                file_id_string_tos(&e->id),
+                (unsigned int)e->name_hash);
 }
 
 /*******************************************************************
@@ -408,832 +514,1218 @@ pid = %lu, share_mode = 0x%x, desired_access = 0x%x, port = 0x%x, type= 0x%x, fi
 
 static void print_share_mode_table(struct locking_data *data)
 {
-       int num_share_modes = data->u.num_share_mode_entries;
-       share_mode_entry *shares = (share_mode_entry *)(data + 1);
+       int num_share_modes = data->u.s.num_share_mode_entries;
+       struct share_mode_entry *shares =
+               (struct share_mode_entry *)(data + 1);
        int i;
 
        for (i = 0; i < num_share_modes; i++) {
-               share_mode_entry *entry_p = &shares[i];
-               DEBUG(10,("print_share_mode_table: %s\n", share_mode_str(i, entry_p) ));
+               struct share_mode_entry entry;
+               char *str;
+
+               /*
+                * We need to memcpy the entry here due to alignment
+                * restrictions that are not met when directly accessing
+                * shares[i]
+                */
+
+               memcpy(&entry, &shares[i], sizeof(struct share_mode_entry));
+               str = share_mode_str(talloc_tos(), i, &entry);
+
+               DEBUG(10,("print_share_mode_table: %s\n", str ? str : ""));
+               TALLOC_FREE(str);
        }
 }
 
-/*******************************************************************
- Get all share mode entries for a dev/inode pair.
-********************************************************************/
-
-int get_share_modes(connection_struct *conn, 
-                   SMB_DEV_T dev, SMB_INO_T inode, 
-                   share_mode_entry **pp_shares)
+static int parse_delete_tokens_list(struct share_mode_lock *lck,
+               struct locking_data *pdata,
+               const TDB_DATA dbuf)
 {
-       TDB_DATA dbuf;
-       struct locking_data *data;
-       int num_share_modes;
-       share_mode_entry *shares = NULL;
-       TDB_DATA key = locking_key(dev, inode);
-       *pp_shares = NULL;
-
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
-               return 0;
+       uint8_t *p = dbuf.dptr + sizeof(struct locking_data) +
+                       (lck->num_share_modes *
+                       sizeof(struct share_mode_entry));
+       uint8_t *end_ptr = dbuf.dptr + (dbuf.dsize - 2);
+       int delete_tokens_size = 0;
+       int i;
 
-       data = (struct locking_data *)dbuf.dptr;
-       num_share_modes = data->u.num_share_mode_entries;
-       if(num_share_modes) {
-               pstring fname;
-               int i;
-               int del_count = 0;
+       lck->delete_tokens = NULL;
 
-               shares = (share_mode_entry *)memdup(dbuf.dptr + sizeof(*data),  
-                                               num_share_modes * sizeof(share_mode_entry));
+       for (i = 0; i < pdata->u.s.num_delete_token_entries; i++) {
+               uint32_t token_len;
+               struct delete_token_list *pdtl;
 
-               if (!shares) {
-                       SAFE_FREE(dbuf.dptr);
-                       return 0;
+               if (end_ptr - p < (sizeof(uint32_t) + sizeof(uint32_t) +
+                                       sizeof(uid_t) + sizeof(gid_t))) {
+                       DEBUG(0,("parse_delete_tokens_list: "
+                               "corrupt token list (%u)",
+                               (unsigned int)(end_ptr - p)));
+                       smb_panic("corrupt token list");
+                       return -1;
                }
 
-               /* Save off the associated filename. */
-               pstrcpy(fname, dbuf.dptr + sizeof(*data) + num_share_modes * sizeof(share_mode_entry));
+               memcpy(&token_len, p, sizeof(token_len));
+               delete_tokens_size += token_len;
+
+               if (p + token_len > end_ptr || token_len < sizeof(token_len) +
+                                               sizeof(pdtl->name_hash) +
+                                               sizeof(uid_t) +
+                                               sizeof(gid_t)) {
+                       DEBUG(0,("parse_delete_tokens_list: "
+                               "invalid token length (%u)\n",
+                               (unsigned int)token_len ));
+                       smb_panic("invalid token length");
+                       return -1;
+               }
 
-               /*
-                * Ensure that each entry has a real process attached.
-                */
+               p += sizeof(token_len);
 
-               for (i = 0; i < num_share_modes; ) {
-                       share_mode_entry *entry_p = &shares[i];
-                       if (process_exists(entry_p->pid)) {
-                               DEBUG(10,("get_share_modes: %s\n", share_mode_str(i, entry_p) ));
-                               i++;
-                       } else {
-                               DEBUG(10,("get_share_modes: deleted %s\n", share_mode_str(i, entry_p) ));
-                               if (num_share_modes - i - 1 > 0) {
-                                       memcpy( &shares[i], &shares[i+1],
-                                               sizeof(share_mode_entry) * (num_share_modes - i - 1));
-                               }
-                               num_share_modes--;
-                               del_count++;
-                       }
+               pdtl = TALLOC_ZERO_P(lck, struct delete_token_list);
+               if (pdtl == NULL) {
+                       DEBUG(0,("parse_delete_tokens_list: talloc failed"));
+                       return -1;
+               }
+               /* Copy out the name_hash. */
+               memcpy(&pdtl->name_hash, p, sizeof(pdtl->name_hash));
+               p += sizeof(pdtl->name_hash);
+
+               pdtl->delete_token = TALLOC_ZERO_P(pdtl, struct security_unix_token);
+               if (pdtl->delete_token == NULL) {
+                       DEBUG(0,("parse_delete_tokens_list: talloc failed"));
+                       return -1;
                }
 
-               /* Did we delete any ? If so, re-store in tdb. */
-               if (del_count) {
-                       data->u.num_share_mode_entries = num_share_modes;
-                       
-                       if (num_share_modes) {
-                               memcpy(dbuf.dptr + sizeof(*data), shares,
-                                               num_share_modes * sizeof(share_mode_entry));
-                               /* Append the filename. */
-                               pstrcpy(dbuf.dptr + sizeof(*data) + num_share_modes * sizeof(share_mode_entry), fname);
+               /* Copy out the uid and gid. */
+               memcpy(&pdtl->delete_token->uid, p, sizeof(uid_t));
+               p += sizeof(uid_t);
+               memcpy(&pdtl->delete_token->gid, p, sizeof(gid_t));
+               p += sizeof(gid_t);
+
+               token_len -= (sizeof(token_len) + sizeof(pdtl->name_hash) +
+                               sizeof(uid_t) + sizeof(gid_t));
+
+               /* Any supplementary groups ? */
+               if (token_len) {
+                       int j;
+
+                       if (token_len % sizeof(gid_t) != 0) {
+                               DEBUG(0,("parse_delete_tokens_list: "
+                                       "corrupt group list (%u)",
+                                       (unsigned int)(token_len % sizeof(gid_t)) ));
+                               smb_panic("corrupt group list");
+                               return -1;
                        }
 
-                       /* The record has shrunk a bit */
-                       dbuf.dsize -= del_count * sizeof(share_mode_entry);
+                       pdtl->delete_token->ngroups = token_len / sizeof(gid_t);
+                       pdtl->delete_token->groups = talloc_array(pdtl->delete_token, gid_t,
+                                               pdtl->delete_token->ngroups);
+                       if (pdtl->delete_token->groups == NULL) {
+                               DEBUG(0,("parse_delete_tokens_list: talloc failed"));
+                               return -1;
+                       }
 
-                       if (data->u.num_share_mode_entries == 0) {
-                               if (tdb_delete(tdb, key) == -1) {
-                                       SAFE_FREE(shares);
-                                       SAFE_FREE(dbuf.dptr);
-                                       return 0;
-                               }
-                       } else {
-                               if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1) {
-                                       SAFE_FREE(shares);
-                                       SAFE_FREE(dbuf.dptr);
-                                       return 0;
-                               }
+                       for (j = 0; j < pdtl->delete_token->ngroups; j++) {
+                               memcpy(&pdtl->delete_token->groups[j], p, sizeof(gid_t));
+                               p += sizeof(gid_t);
                        }
                }
+               /* Add to the list. */
+               DLIST_ADD(lck->delete_tokens, pdtl);
        }
 
-       SAFE_FREE(dbuf.dptr);
-       *pp_shares = shares;
-       return num_share_modes;
+       return delete_tokens_size;
 }
 
 /*******************************************************************
Fill a share mode entry.
Get all share mode entries for a dev/inode pair.
 ********************************************************************/
 
-static void fill_share_mode(char *p, files_struct *fsp, uint16 port, uint16 op_type)
+static bool parse_share_modes(const TDB_DATA dbuf, struct share_mode_lock *lck)
 {
-       share_mode_entry *e = (share_mode_entry *)p;
-       void *x = &e->time; /* Needed to force alignment. p may not be aligned.... */
-
-       memset(e, '\0', sizeof(share_mode_entry));
-       e->pid = sys_getpid();
-       e->share_mode = fsp->share_mode;
-       e->desired_access = fsp->desired_access;
-       e->op_port = port;
-       e->op_type = op_type;
-       memcpy(x, &fsp->open_time, sizeof(struct timeval));
-       e->share_file_id = fsp->file_id;
-       e->dev = fsp->dev;
-       e->inode = fsp->inode;
-}
+       struct locking_data data;
+       int delete_tokens_size;
+       int i;
 
-/*******************************************************************
- Check if two share mode entries are identical, ignoring oplock 
- and port info and desired_access.
-********************************************************************/
+       if (dbuf.dsize < sizeof(struct locking_data)) {
+               smb_panic("parse_share_modes: buffer too short");
+       }
 
-BOOL share_modes_identical( share_mode_entry *e1, share_mode_entry *e2)
-{
-#if 1 /* JRA PARANOIA TEST - REMOVE LATER */
-       if (e1->pid == e2->pid &&
-               e1->share_file_id == e2->share_file_id &&
-               e1->dev == e2->dev &&
-               e1->inode == e2->inode &&
-               (e1->share_mode & ~DELETE_ON_CLOSE_FLAG) != (e2->share_mode & ~DELETE_ON_CLOSE_FLAG)) {
-                       DEBUG(0,("PANIC: share_modes_identical: share_mode missmatch (e1 = %u, e2 = %u). Logic error.\n",
-                               (unsigned int)(e1->share_mode & ~DELETE_ON_CLOSE_FLAG),
-                               (unsigned int)(e2->share_mode & ~DELETE_ON_CLOSE_FLAG) ));
-               smb_panic("PANIC: share_modes_identical logic error.\n");
-       }
-#endif
-
-       return (e1->pid == e2->pid &&
-               (e1->share_mode & ~DELETE_ON_CLOSE_FLAG) == (e2->share_mode & ~DELETE_ON_CLOSE_FLAG) &&
-               e1->dev == e2->dev &&
-               e1->inode == e2->inode &&
-               e1->share_file_id == e2->share_file_id );
-}
+       memcpy(&data, dbuf.dptr, sizeof(data));
+
+       lck->old_write_time = data.u.s.old_write_time;
+       lck->changed_write_time = data.u.s.changed_write_time;
+       lck->num_share_modes = data.u.s.num_share_mode_entries;
+
+       DEBUG(10, ("parse_share_modes: owrt: %s, "
+                  "cwrt: %s, ntok: %u, num_share_modes: %d\n",
+                  timestring(talloc_tos(),
+                             convert_timespec_to_time_t(lck->old_write_time)),
+                  timestring(talloc_tos(),
+                             convert_timespec_to_time_t(
+                                     lck->changed_write_time)),
+                  (unsigned int)data.u.s.num_delete_token_entries,
+                  lck->num_share_modes));
+
+       if ((lck->num_share_modes < 0) || (lck->num_share_modes > 1000000)) {
+               DEBUG(0, ("invalid number of share modes: %d\n",
+                         lck->num_share_modes));
+               smb_panic("parse_share_modes: invalid number of share modes");
+       }
 
-/*******************************************************************
- Delete a specific share mode. Return the number
- of entries left, and a memdup'ed copy of the entry deleted (if required).
- Ignore if no entry deleted.
-********************************************************************/
+       lck->share_modes = NULL;
 
-ssize_t del_share_entry( SMB_DEV_T dev, SMB_INO_T inode,
-                       share_mode_entry *entry, share_mode_entry **ppse)
-{
-       TDB_DATA dbuf;
-       struct locking_data *data;
-       int i, del_count=0;
-       share_mode_entry *shares;
-       ssize_t count = 0;
-       TDB_DATA key = locking_key(dev, inode);
+       if (lck->num_share_modes != 0) {
+
+               if (dbuf.dsize < (sizeof(struct locking_data) +
+                                 (lck->num_share_modes *
+                                  sizeof(struct share_mode_entry)))) {
+                       smb_panic("parse_share_modes: buffer too short");
+               }
 
-       if (ppse)
-               *ppse = NULL;
+               lck->share_modes = (struct share_mode_entry *)
+                       TALLOC_MEMDUP(lck,
+                                     dbuf.dptr+sizeof(struct locking_data),
+                                     lck->num_share_modes *
+                                     sizeof(struct share_mode_entry));
 
-       /* read in the existing share modes */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
-               return -1;
+               if (lck->share_modes == NULL) {
+                       smb_panic("parse_share_modes: talloc failed");
+               }
+       }
+
+       /* Get any delete tokens. */
+       delete_tokens_size = parse_delete_tokens_list(lck, &data, dbuf);
+       if (delete_tokens_size < 0) {
+               smb_panic("parse_share_modes: parse_delete_tokens_list failed");
+       }
 
-       data = (struct locking_data *)dbuf.dptr;
-       shares = (share_mode_entry *)(dbuf.dptr + sizeof(*data));
+       /* Save off the associated service path and filename. */
+       lck->servicepath = (const char *)dbuf.dptr + sizeof(struct locking_data) +
+               (lck->num_share_modes * sizeof(struct share_mode_entry)) +
+               delete_tokens_size;
+
+       lck->base_name = (const char *)dbuf.dptr + sizeof(struct locking_data) +
+               (lck->num_share_modes * sizeof(struct share_mode_entry)) +
+               delete_tokens_size +
+               strlen(lck->servicepath) + 1;
+
+       lck->stream_name = (const char *)dbuf.dptr + sizeof(struct locking_data) +
+               (lck->num_share_modes * sizeof(struct share_mode_entry)) +
+               delete_tokens_size +
+               strlen(lck->servicepath) + 1 +
+               strlen(lck->base_name) + 1;
 
        /*
-        * Find any with this pid and delete it
-        * by overwriting with the rest of the data 
-        * from the record.
+        * Ensure that each entry has a real process attached.
         */
 
-       DEBUG(10,("del_share_entry: num_share_modes = %d\n", data->u.num_share_mode_entries ));
-
-       for (i=0;i<data->u.num_share_mode_entries;) {
-               if (share_modes_identical(&shares[i], entry)) {
-                       DEBUG(10,("del_share_entry: deleted %s\n",
-                               share_mode_str(i, &shares[i]) ));
-                       if (ppse)
-                               *ppse = memdup(&shares[i], sizeof(*shares));
-                       data->u.num_share_mode_entries--;
-                       if ((dbuf.dsize - (sizeof(*data) + (i+1)*sizeof(*shares))) > 0) {
-                               memmove(&shares[i], &shares[i+1], 
-                                       dbuf.dsize - (sizeof(*data) + (i+1)*sizeof(*shares)));
-                       }
-                       del_count++;
+       for (i = 0; i < lck->num_share_modes; i++) {
+               struct share_mode_entry *entry_p = &lck->share_modes[i];
+               char *str = NULL;
+               if (DEBUGLEVEL >= 10) {
+                       str = share_mode_str(NULL, i, entry_p);
+               }
+               DEBUG(10,("parse_share_modes: %s\n",
+                       str ? str : ""));
+               if (!serverid_exists(&entry_p->pid)) {
+                       DEBUG(10,("parse_share_modes: deleted %s\n",
+                               str ? str : ""));
+                       entry_p->op_type = UNUSED_SHARE_MODE_ENTRY;
+                       lck->modified = True;
+               }
+               TALLOC_FREE(str);
+       }
 
-                       DEBUG(10,("del_share_entry: deleting entry %d\n", i ));
+       return True;
+}
 
-               } else {
-                       i++;
+static TDB_DATA unparse_share_modes(const struct share_mode_lock *lck)
+{
+       TDB_DATA result;
+       int num_valid = 0;
+       int i;
+       struct locking_data *data;
+       ssize_t offset;
+       ssize_t sp_len, bn_len, sn_len;
+       uint32_t delete_tokens_size = 0;
+       struct delete_token_list *pdtl = NULL;
+       uint32_t num_delete_token_entries = 0;
+
+       result.dptr = NULL;
+       result.dsize = 0;
+
+       for (i=0; i<lck->num_share_modes; i++) {
+               if (!is_unused_share_mode_entry(&lck->share_modes[i])) {
+                       num_valid += 1;
                }
        }
 
-       if (del_count) {
-               /* the record may have shrunk a bit */
-               dbuf.dsize -= del_count * sizeof(*shares);
+       if (num_valid == 0) {
+               return result;
+       }
 
-               count = (ssize_t)data->u.num_share_mode_entries;
+       sp_len = strlen(lck->servicepath);
+       bn_len = strlen(lck->base_name);
+       sn_len = lck->stream_name != NULL ? strlen(lck->stream_name) : 0;
+
+       for (pdtl = lck->delete_tokens; pdtl; pdtl = pdtl->next) {
+               num_delete_token_entries++;
+               delete_tokens_size += (sizeof(uint32_t) +
+                               sizeof(uint32_t) +
+                               sizeof(uid_t) +
+                               sizeof(gid_t) +
+                               pdtl->delete_token->ngroups*sizeof(gid_t));
+       }
 
-               /* store it back in the database */
-               if (data->u.num_share_mode_entries == 0) {
-                       if (tdb_delete(tdb, key) == -1)
-                               count = -1;
-               } else {
-                       if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-                               count = -1;
+       result.dsize = sizeof(*data) +
+               lck->num_share_modes * sizeof(struct share_mode_entry) +
+               delete_tokens_size +
+               sp_len + 1 +
+               bn_len + 1 +
+               sn_len + 1;
+       result.dptr = talloc_array(lck, uint8, result.dsize);
+
+       if (result.dptr == NULL) {
+               smb_panic("talloc failed");
+       }
+
+       data = (struct locking_data *)result.dptr;
+       ZERO_STRUCTP(data);
+       data->u.s.num_share_mode_entries = lck->num_share_modes;
+       data->u.s.old_write_time = lck->old_write_time;
+       data->u.s.changed_write_time = lck->changed_write_time;
+       data->u.s.num_delete_token_entries = num_delete_token_entries;
+
+       DEBUG(10,("unparse_share_modes: owrt: %s cwrt: %s, ntok: %u, "
+                 "num: %d\n",
+                 timestring(talloc_tos(),
+                            convert_timespec_to_time_t(lck->old_write_time)),
+                 timestring(talloc_tos(),
+                            convert_timespec_to_time_t(
+                                    lck->changed_write_time)),
+                 (unsigned int)data->u.s.num_delete_token_entries,
+                 data->u.s.num_share_mode_entries));
+
+       memcpy(result.dptr + sizeof(*data), lck->share_modes,
+              sizeof(struct share_mode_entry)*lck->num_share_modes);
+       offset = sizeof(*data) +
+               sizeof(struct share_mode_entry)*lck->num_share_modes;
+
+       /* Store any delete on close tokens. */
+       for (pdtl = lck->delete_tokens; pdtl; pdtl = pdtl->next) {
+               struct security_unix_token *pdt = pdtl->delete_token;
+               uint32_t token_size = sizeof(uint32_t) +
+                                       sizeof(uint32_t) +
+                                       sizeof(uid_t) +
+                                       sizeof(gid_t) +
+                                       (pdt->ngroups * sizeof(gid_t));
+               uint8_t *p = result.dptr + offset;
+
+               memcpy(p, &token_size, sizeof(uint32_t));
+               p += sizeof(uint32_t);
+
+               memcpy(p, &pdtl->name_hash, sizeof(uint32_t));
+               p += sizeof(uint32_t);
+
+               memcpy(p, &pdt->uid, sizeof(uid_t));
+               p += sizeof(uid_t);
+
+               memcpy(p, &pdt->gid, sizeof(gid_t));
+               p += sizeof(gid_t);
+
+               for (i = 0; i < pdt->ngroups; i++) {
+                       memcpy(p, &pdt->groups[i], sizeof(gid_t));
+                       p += sizeof(gid_t);
                }
+               offset += token_size;
        }
-       DEBUG(10,("del_share_entry: Remaining table.\n"));
-       print_share_mode_table((struct locking_data *)dbuf.dptr);
-       SAFE_FREE(dbuf.dptr);
-       return count;
+
+       strlcpy((char *)result.dptr + offset,
+               lck->servicepath ? lck->servicepath : "",
+               result.dsize - offset);
+       offset += sp_len + 1;
+       strlcpy((char *)result.dptr + offset,
+               lck->base_name ? lck->base_name : "",
+               result.dsize - offset);
+       offset += bn_len + 1;
+       strlcpy((char *)result.dptr + offset,
+               lck->stream_name ? lck->stream_name : "",
+               result.dsize - offset);
+
+       if (DEBUGLEVEL >= 10) {
+               print_share_mode_table(data);
+       }
+
+       return result;
 }
 
-/*******************************************************************
- Del the share mode of a file for this process. Return the number
- of entries left, and a memdup'ed copy of the entry deleted.
-********************************************************************/
+static int share_mode_lock_destructor(struct share_mode_lock *lck)
+{
+       NTSTATUS status;
+       TDB_DATA data;
+
+       if (!lck->modified) {
+               return 0;
+       }
+
+       data = unparse_share_modes(lck);
+
+       if (data.dptr == NULL) {
+               if (!lck->fresh) {
+                       /* There has been an entry before, delete it */
 
-ssize_t del_share_mode(files_struct *fsp, share_mode_entry **ppse)
+                       status = lck->record->delete_rec(lck->record);
+                       if (!NT_STATUS_IS_OK(status)) {
+                               char *errmsg;
+
+                               DEBUG(0, ("delete_rec returned %s\n",
+                                         nt_errstr(status)));
+
+                               if (asprintf(&errmsg, "could not delete share "
+                                            "entry: %s\n",
+                                            nt_errstr(status)) == -1) {
+                                       smb_panic("could not delete share"
+                                                 "entry");
+                               }
+                               smb_panic(errmsg);
+                       }
+               }
+               goto done;
+       }
+
+       status = lck->record->store(lck->record, data, TDB_REPLACE);
+       if (!NT_STATUS_IS_OK(status)) {
+               char *errmsg;
+
+               DEBUG(0, ("store returned %s\n", nt_errstr(status)));
+
+               if (asprintf(&errmsg, "could not store share mode entry: %s",
+                            nt_errstr(status)) == -1) {
+                       smb_panic("could not store share mode entry");
+               }
+               smb_panic(errmsg);
+       }
+
+ done:
+
+       return 0;
+}
+
+static bool fill_share_mode_lock(struct share_mode_lock *lck,
+                                struct file_id id,
+                                const char *servicepath,
+                                const struct smb_filename *smb_fname,
+                                TDB_DATA share_mode_data,
+                                const struct timespec *old_write_time)
 {
-       share_mode_entry entry;
+       /* Ensure we set every field here as the destructor must be
+          valid even if parse_share_modes fails. */
+
+       lck->servicepath = NULL;
+       lck->base_name = NULL;
+       lck->stream_name = NULL;
+       lck->id = id;
+       lck->num_share_modes = 0;
+       lck->share_modes = NULL;
+       lck->delete_tokens = NULL;
+       ZERO_STRUCT(lck->old_write_time);
+       ZERO_STRUCT(lck->changed_write_time);
+       lck->fresh = False;
+       lck->modified = False;
+
+       lck->fresh = (share_mode_data.dptr == NULL);
+
+       if (lck->fresh) {
+               bool has_stream;
+               if (smb_fname == NULL || servicepath == NULL
+                   || old_write_time == NULL) {
+                       return False;
+               }
 
-       /*
-        * Fake up a share_mode_entry for comparisons.
-        */
+               has_stream = smb_fname->stream_name != NULL;
+
+               lck->base_name = talloc_strdup(lck, smb_fname->base_name);
+               lck->stream_name = talloc_strdup(lck, smb_fname->stream_name);
+               lck->servicepath = talloc_strdup(lck, servicepath);
+               if (lck->base_name == NULL ||
+                   (has_stream && lck->stream_name == NULL) ||
+                   lck->servicepath == NULL) {
+                       DEBUG(0, ("talloc failed\n"));
+                       return False;
+               }
+               lck->old_write_time = *old_write_time;
+       } else {
+               if (!parse_share_modes(share_mode_data, lck)) {
+                       DEBUG(0, ("Could not parse share modes\n"));
+                       return False;
+               }
+       }
 
-       fill_share_mode((char *)&entry, fsp, 0, 0);
-       return del_share_entry(fsp->dev, fsp->inode, &entry, ppse);
+       return True;
 }
 
-/*******************************************************************
- Set the share mode of a file. Return False on fail, True on success.
-********************************************************************/
+struct share_mode_lock *get_share_mode_lock(TALLOC_CTX *mem_ctx,
+                                           const struct file_id id,
+                                           const char *servicepath,
+                                           const struct smb_filename *smb_fname,
+                                           const struct timespec *old_write_time)
+{
+       struct share_mode_lock *lck;
+       struct file_id tmp;
+       TDB_DATA key = locking_key(&id, &tmp);
+
+       if (!(lck = TALLOC_P(mem_ctx, struct share_mode_lock))) {
+               DEBUG(0, ("talloc failed\n"));
+               return NULL;
+       }
+
+       if (!(lck->record = lock_db->fetch_locked(lock_db, lck, key))) {
+               DEBUG(3, ("Could not lock share entry\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       if (!fill_share_mode_lock(lck, id, servicepath, smb_fname,
+                                 lck->record->value, old_write_time)) {
+               DEBUG(3, ("fill_share_mode_lock failed\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       talloc_set_destructor(lck, share_mode_lock_destructor);
+
+       return lck;
+}
 
-BOOL set_share_mode(files_struct *fsp, uint16 port, uint16 op_type)
+struct share_mode_lock *fetch_share_mode_unlocked(TALLOC_CTX *mem_ctx,
+                                                 const struct file_id id)
 {
-       TDB_DATA dbuf;
-       struct locking_data *data;
-       char *p=NULL;
-       int size;
-       TDB_DATA key = locking_key_fsp(fsp);
-       BOOL ret = True;
-               
-       /* read in the existing share modes if any */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr) {
-               size_t offset;
-               /* we'll need to create a new record */
-               pstring fname;
-
-               pstrcpy(fname, fsp->conn->connectpath);
-               pstrcat(fname, "/");
-               pstrcat(fname, fsp->fsp_name);
-
-               size = sizeof(*data) + sizeof(share_mode_entry) + strlen(fname) + 1;
-               p = (char *)SMB_MALLOC(size);
-               if (!p)
-                       return False;
-               data = (struct locking_data *)p;
-               data->u.num_share_mode_entries = 1;
-       
-               DEBUG(10,("set_share_mode: creating entry for file %s. num_share_modes = 1\n",
-                       fsp->fsp_name ));
-
-               offset = sizeof(*data) + sizeof(share_mode_entry);
-               safe_strcpy(p + offset, fname, size - offset - 1);
-               fill_share_mode(p + sizeof(*data), fsp, port, op_type);
-               dbuf.dptr = p;
-               dbuf.dsize = size;
-               if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-                       ret = False;
-
-               print_share_mode_table((struct locking_data *)p);
-
-               SAFE_FREE(p);
-               return ret;
-       }
-
-       /* we're adding to an existing entry - this is a bit fiddly */
-       data = (struct locking_data *)dbuf.dptr;
-
-       data->u.num_share_mode_entries++;
-       
-       DEBUG(10,("set_share_mode: adding entry for file %s. new num_share_modes = %d\n",
-               fsp->fsp_name, data->u.num_share_mode_entries ));
-
-       size = dbuf.dsize + sizeof(share_mode_entry);
-       p = SMB_MALLOC(size);
-       if (!p) {
-               SAFE_FREE(dbuf.dptr);
-               return False;
+       struct share_mode_lock *lck;
+       struct file_id tmp;
+       TDB_DATA key = locking_key(&id, &tmp);
+       TDB_DATA data;
+
+       if (!(lck = TALLOC_P(mem_ctx, struct share_mode_lock))) {
+               DEBUG(0, ("talloc failed\n"));
+               return NULL;
        }
-       memcpy(p, dbuf.dptr, sizeof(*data));
-       fill_share_mode(p + sizeof(*data), fsp, port, op_type);
-       memcpy(p + sizeof(*data) + sizeof(share_mode_entry), dbuf.dptr + sizeof(*data),
-              dbuf.dsize - sizeof(*data));
-       SAFE_FREE(dbuf.dptr);
-       dbuf.dptr = p;
-       dbuf.dsize = size;
-       if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-               ret = False;
-       print_share_mode_table((struct locking_data *)p);
-       SAFE_FREE(p);
-       return ret;
+
+       if (lock_db->fetch(lock_db, lck, key, &data) == -1) {
+               DEBUG(3, ("Could not fetch share entry\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       if (!fill_share_mode_lock(lck, id, NULL, NULL, data, NULL)) {
+               DEBUG(10, ("fetch_share_mode_unlocked: no share_mode record "
+                          "around (file not open)\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       return lck;
 }
 
 /*******************************************************************
- A generic in-place modification call for share mode entries.
+ Sets the service name and filename for rename.
+ At this point we emit "file renamed" messages to all
+ process id's that have this file open.
+ Based on an initial code idea from SATOH Fumiyasu <fumiya@samba.gr.jp>
 ********************************************************************/
 
-static BOOL mod_share_mode( SMB_DEV_T dev, SMB_INO_T inode, share_mode_entry *entry,
-                          void (*mod_fn)(share_mode_entry *, SMB_DEV_T, SMB_INO_T, void *),
-                          void *param)
+bool rename_share_filename(struct messaging_context *msg_ctx,
+                       struct share_mode_lock *lck,
+                       const char *servicepath,
+                       uint32_t orig_name_hash,
+                       uint32_t new_name_hash,
+                       const struct smb_filename *smb_fname_dst)
 {
-       TDB_DATA dbuf;
-       struct locking_data *data;
+       size_t sp_len;
+       size_t bn_len;
+       size_t sn_len;
+       size_t msg_len;
+       char *frm = NULL;
        int i;
-       share_mode_entry *shares;
-       BOOL need_store=False;
-       BOOL ret = True;
-       TDB_DATA key = locking_key(dev, inode);
-
-       /* read in the existing share modes */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
+       bool strip_two_chars = false;
+       bool has_stream = smb_fname_dst->stream_name != NULL;
+
+       DEBUG(10, ("rename_share_filename: servicepath %s newname %s\n",
+                  servicepath, smb_fname_dst->base_name));
+
+       /*
+        * rename_internal_fsp() and rename_internals() add './' to
+        * head of newname if newname does not contain a '/'.
+        */
+       if (smb_fname_dst->base_name[0] &&
+           smb_fname_dst->base_name[1] &&
+           smb_fname_dst->base_name[0] == '.' &&
+           smb_fname_dst->base_name[1] == '/') {
+               strip_two_chars = true;
+       }
+
+       lck->servicepath = talloc_strdup(lck, servicepath);
+       lck->base_name = talloc_strdup(lck, smb_fname_dst->base_name +
+                                      (strip_two_chars ? 2 : 0));
+       lck->stream_name = talloc_strdup(lck, smb_fname_dst->stream_name);
+       if (lck->base_name == NULL ||
+           (has_stream && lck->stream_name == NULL) ||
+           lck->servicepath == NULL) {
+               DEBUG(0, ("rename_share_filename: talloc failed\n"));
+               return False;
+       }
+       lck->modified = True;
+
+       sp_len = strlen(lck->servicepath);
+       bn_len = strlen(lck->base_name);
+       sn_len = has_stream ? strlen(lck->stream_name) : 0;
+
+       msg_len = MSG_FILE_RENAMED_MIN_SIZE + sp_len + 1 + bn_len + 1 +
+           sn_len + 1;
+
+       /* Set up the name changed message. */
+       frm = talloc_array(lck, char, msg_len);
+       if (!frm) {
                return False;
+       }
+
+       push_file_id_24(frm, &lck->id);
+
+       DEBUG(10,("rename_share_filename: msg_len = %u\n", (unsigned int)msg_len ));
+
+       strlcpy(&frm[24],
+               lck->servicepath ? lck->servicepath : "",
+               sp_len+1);
+       strlcpy(&frm[24 + sp_len + 1],
+               lck->base_name ? lck->base_name : "",
+               bn_len+1);
+       strlcpy(&frm[24 + sp_len + 1 + bn_len + 1],
+               lck->stream_name ? lck->stream_name : "",
+               sn_len+1);
+
+       /* Send the messages. */
+       for (i=0; i<lck->num_share_modes; i++) {
+               struct share_mode_entry *se = &lck->share_modes[i];
+               if (!is_valid_share_mode_entry(se)) {
+                       continue;
+               }
+
+               /* If this is a hardlink to the inode
+                  with a different name, skip this. */
+               if (se->name_hash != orig_name_hash) {
+                       continue;
+               }
 
-       data = (struct locking_data *)dbuf.dptr;
-       shares = (share_mode_entry *)(dbuf.dptr + sizeof(*data));
+               se->name_hash = new_name_hash;
 
-       /* find any with our pid and call the supplied function */
-       for (i=0;i<data->u.num_share_mode_entries;i++) {
-               if (share_modes_identical(entry, &shares[i])) {
-                       mod_fn(&shares[i], dev, inode, param);
-                       need_store=True;
+               /* But not to ourselves... */
+               if (procid_is_me(&se->pid)) {
+                       continue;
                }
+
+               DEBUG(10,("rename_share_filename: sending rename message to "
+                         "pid %s file_id %s sharepath %s base_name %s "
+                         "stream_name %s\n",
+                         procid_str_static(&se->pid),
+                         file_id_string_tos(&lck->id),
+                         lck->servicepath, lck->base_name,
+                       has_stream ? lck->stream_name : ""));
+
+               messaging_send_buf(msg_ctx, se->pid, MSG_SMB_FILE_RENAME,
+                                  (uint8 *)frm, msg_len);
        }
 
-       /* if the mod fn was called then store it back */
-       if (need_store) {
-               if (data->u.num_share_mode_entries == 0) {
-                       if (tdb_delete(tdb, key) == -1)
-                               ret = False;
-               } else {
-                       if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-                               ret = False;
+       return True;
+}
+
+void get_file_infos(struct file_id id,
+                   uint32_t name_hash,
+                   bool *delete_on_close,
+                   struct timespec *write_time)
+{
+       struct share_mode_lock *lck;
+
+       if (delete_on_close) {
+               *delete_on_close = false;
+       }
+
+       if (write_time) {
+               ZERO_STRUCTP(write_time);
+       }
+
+       if (!(lck = fetch_share_mode_unlocked(talloc_tos(), id))) {
+               return;
+       }
+
+       if (delete_on_close) {
+               *delete_on_close = is_delete_on_close_set(lck, name_hash);
+       }
+
+       if (write_time) {
+               struct timespec wt;
+
+               wt = lck->changed_write_time;
+               if (null_timespec(wt)) {
+                       wt = lck->old_write_time;
                }
+
+               *write_time = wt;
        }
 
-       SAFE_FREE(dbuf.dptr);
-       return ret;
+       TALLOC_FREE(lck);
 }
 
-/*******************************************************************
- Static function that actually does the work for the generic function
- below.
-********************************************************************/
+bool is_valid_share_mode_entry(const struct share_mode_entry *e)
+{
+       int num_props = 0;
 
-static void remove_share_oplock_fn(share_mode_entry *entry, SMB_DEV_T dev, SMB_INO_T inode, 
-                                   void *param)
+       if (e->op_type == UNUSED_SHARE_MODE_ENTRY) {
+               /* cope with dead entries from the process not
+                  existing. These should not be considered valid,
+                  otherwise we end up doing zero timeout sharing
+                  violation */
+               return False;
+       }
+
+       num_props += ((e->op_type == NO_OPLOCK) ? 1 : 0);
+       num_props += (EXCLUSIVE_OPLOCK_TYPE(e->op_type) ? 1 : 0);
+       num_props += (LEVEL_II_OPLOCK_TYPE(e->op_type) ? 1 : 0);
+
+       SMB_ASSERT(num_props <= 1);
+       return (num_props != 0);
+}
+
+bool is_deferred_open_entry(const struct share_mode_entry *e)
+{
+       return (e->op_type == DEFERRED_OPEN_ENTRY);
+}
+
+bool is_unused_share_mode_entry(const struct share_mode_entry *e)
 {
-       DEBUG(10,("remove_share_oplock_fn: removing oplock info for entry dev=%x ino=%.0f\n",
-                 (unsigned int)dev, (double)inode ));
-       /* Delete the oplock info. */
-       entry->op_port = 0;
-       entry->op_type = NO_OPLOCK;
+       return (e->op_type == UNUSED_SHARE_MODE_ENTRY);
 }
 
 /*******************************************************************
Remove an oplock port and mode entry from a share mode.
Fill a share mode entry.
 ********************************************************************/
 
-BOOL remove_share_oplock(files_struct *fsp)
+static void fill_share_mode_entry(struct share_mode_entry *e,
+                                 files_struct *fsp,
+                                 uid_t uid, uint64_t mid, uint16 op_type)
 {
-       share_mode_entry entry;
-       /*
-        * Fake up an entry for comparisons...
-        */
-       fill_share_mode((char *)&entry, fsp, 0, 0);
-       return mod_share_mode(fsp->dev, fsp->inode, &entry, remove_share_oplock_fn, NULL);
+       ZERO_STRUCTP(e);
+       e->pid = sconn_server_id(fsp->conn->sconn);
+       e->share_access = fsp->share_access;
+       e->private_options = fsp->fh->private_options;
+       e->access_mask = fsp->access_mask;
+       e->op_mid = mid;
+       e->op_type = op_type;
+       e->time.tv_sec = fsp->open_time.tv_sec;
+       e->time.tv_usec = fsp->open_time.tv_usec;
+       e->id = fsp->file_id;
+       e->share_file_id = fsp->fh->gen_id;
+       e->uid = (uint32)uid;
+       e->flags = fsp->posix_open ? SHARE_MODE_FLAG_POSIX_OPEN : 0;
+       e->name_hash = fsp->name_hash;
 }
 
-/*******************************************************************
- Static function that actually does the work for the generic function
- below.
-********************************************************************/
+static void fill_deferred_open_entry(struct share_mode_entry *e,
+                                    const struct timeval request_time,
+                                    struct file_id id,
+                                    struct server_id pid,
+                                    uint64_t mid)
+{
+       ZERO_STRUCTP(e);
+       e->pid = pid;
+       e->op_mid = mid;
+       e->op_type = DEFERRED_OPEN_ENTRY;
+       e->time.tv_sec = request_time.tv_sec;
+       e->time.tv_usec = request_time.tv_usec;
+       e->id = id;
+       e->uid = (uint32)-1;
+       e->flags = 0;
+}
 
-static void downgrade_share_oplock_fn(share_mode_entry *entry, SMB_DEV_T dev, SMB_INO_T inode, 
-                                   void *param)
+static void add_share_mode_entry(struct share_mode_lock *lck,
+                                const struct share_mode_entry *entry)
 {
-       DEBUG(10,("downgrade_share_oplock_fn: downgrading oplock info for entry dev=%x ino=%.0f\n",
-                 (unsigned int)dev, (double)inode ));
-       entry->op_type = LEVEL_II_OPLOCK;
+       int i;
+
+       for (i=0; i<lck->num_share_modes; i++) {
+               struct share_mode_entry *e = &lck->share_modes[i];
+               if (is_unused_share_mode_entry(e)) {
+                       *e = *entry;
+                       break;
+               }
+       }
+
+       if (i == lck->num_share_modes) {
+               /* No unused entry found */
+               ADD_TO_ARRAY(lck, struct share_mode_entry, *entry,
+                            &lck->share_modes, &lck->num_share_modes);
+       }
+       lck->modified = True;
 }
 
-/*******************************************************************
- Downgrade a oplock type from exclusive to level II.
-********************************************************************/
+void set_share_mode(struct share_mode_lock *lck, files_struct *fsp,
+                   uid_t uid, uint64_t mid, uint16 op_type)
+{
+       struct share_mode_entry entry;
+       fill_share_mode_entry(&entry, fsp, uid, mid, op_type);
+       add_share_mode_entry(lck, &entry);
+}
 
-BOOL downgrade_share_oplock(files_struct *fsp)
+void add_deferred_open(struct share_mode_lock *lck, uint64_t mid,
+                      struct timeval request_time,
+                      struct server_id pid, struct file_id id)
 {
-       share_mode_entry entry;
-       /*
-        * Fake up an entry for comparisons...
-        */
-       fill_share_mode((char *)&entry, fsp, 0, 0);
-       return mod_share_mode(fsp->dev, fsp->inode, &entry, downgrade_share_oplock_fn, NULL);
+       struct share_mode_entry entry;
+       fill_deferred_open_entry(&entry, request_time, id, pid, mid);
+       add_share_mode_entry(lck, &entry);
 }
 
 /*******************************************************************
- Get/Set the delete on close flag in a set of share modes.
- Return False on fail, True on success.
+ Check if two share mode entries are identical, ignoring oplock 
+ and mid info and desired_access. (Removed paranoia test - it's
+ not automatically a logic error if they are identical. JRA.)
 ********************************************************************/
 
-BOOL modify_delete_flag( SMB_DEV_T dev, SMB_INO_T inode, BOOL delete_on_close)
+static bool share_modes_identical(struct share_mode_entry *e1,
+                                 struct share_mode_entry *e2)
 {
-       TDB_DATA dbuf;
-       struct locking_data *data;
-       int i;
-       share_mode_entry *shares;
-       TDB_DATA key = locking_key(dev, inode);
+       /* We used to check for e1->share_access == e2->share_access here
+          as well as the other fields but 2 different DOS or FCB opens
+          sharing the same share mode entry may validly differ in
+          fsp->share_access field. */
 
-       /* read in the existing share modes */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
-               return False;
+       return (procid_equal(&e1->pid, &e2->pid) &&
+               file_id_equal(&e1->id, &e2->id) &&
+               e1->share_file_id == e2->share_file_id );
+}
 
-       data = (struct locking_data *)dbuf.dptr;
-       shares = (share_mode_entry *)(dbuf.dptr + sizeof(*data));
+static bool deferred_open_identical(struct share_mode_entry *e1,
+                                   struct share_mode_entry *e2)
+{
+       return (procid_equal(&e1->pid, &e2->pid) &&
+               (e1->op_mid == e2->op_mid) &&
+               file_id_equal(&e1->id, &e2->id));
+}
 
-       /* Set/Unset the delete on close element. */
-       for (i=0;i<data->u.num_share_mode_entries;i++,shares++) {
-               shares->share_mode = (delete_on_close ?
-                            (shares->share_mode | DELETE_ON_CLOSE_FLAG) :
-                            (shares->share_mode & ~DELETE_ON_CLOSE_FLAG) );
-       }
+static struct share_mode_entry *find_share_mode_entry(struct share_mode_lock *lck,
+                                                     struct share_mode_entry *entry)
+{
+       int i;
 
-       /* store it back */
-       if (data->u.num_share_mode_entries) {
-               if (tdb_store(tdb, key, dbuf, TDB_REPLACE)==-1) {
-                       SAFE_FREE(dbuf.dptr);
-                       return False;
+       for (i=0; i<lck->num_share_modes; i++) {
+               struct share_mode_entry *e = &lck->share_modes[i];
+               if (is_valid_share_mode_entry(entry) &&
+                   is_valid_share_mode_entry(e) &&
+                   share_modes_identical(e, entry)) {
+                       return e;
+               }
+               if (is_deferred_open_entry(entry) &&
+                   is_deferred_open_entry(e) &&
+                   deferred_open_identical(e, entry)) {
+                       return e;
                }
        }
-
-       SAFE_FREE(dbuf.dptr);
-       return True;
+       return NULL;
 }
 
 /*******************************************************************
- Print out a deferred open entry.
+ Del the share mode of a file for this process. Return the number of
+ entries left.
 ********************************************************************/
 
-char *deferred_open_str(int num, deferred_open_entry *e)
+bool del_share_mode(struct share_mode_lock *lck, files_struct *fsp)
 {
-       static pstring de_str;
+       struct share_mode_entry entry, *e;
+
+       /* Don't care about the pid owner being correct here - just a search. */
+       fill_share_mode_entry(&entry, fsp, (uid_t)-1, 0, NO_OPLOCK);
 
-       slprintf(de_str, sizeof(de_str)-1, "deferred_open_entry[%d]: \
-pid = %lu, mid = %u, dev = 0x%x, inode = %.0f, port = %u, time = [%u.%06u]",
-               num, (unsigned long)e->pid, (unsigned int)e->mid, (unsigned int)e->dev, (double)e->inode,
-               (unsigned int)e->port,
-               (unsigned int)e->time.tv_sec, (unsigned int)e->time.tv_usec );
+       e = find_share_mode_entry(lck, &entry);
+       if (e == NULL) {
+               return False;
+       }
 
-       return de_str;
+       e->op_type = UNUSED_SHARE_MODE_ENTRY;
+       lck->modified = True;
+       return True;
 }
 
-/* Internal data structures for deferred opens... */
+void del_deferred_open_entry(struct share_mode_lock *lck, uint64_t mid,
+                            struct server_id pid)
+{
+       struct share_mode_entry entry, *e;
 
-struct de_locking_key {
-       char name[4];
-       SMB_DEV_T dev;
-       SMB_INO_T inode;
-};
+       fill_deferred_open_entry(&entry, timeval_zero(),
+                                lck->id, pid, mid);
 
-struct deferred_open_data {
-        union {
-                int num_deferred_open_entries;
-                deferred_open_entry dummy; /* Needed for alignment. */
-        } u;
-        /* the following two entries are implicit
-           deferred_open_entry de_entries[num_deferred_open_entries];
-           char file_name[];
-        */
-};
+       e = find_share_mode_entry(lck, &entry);
+       if (e == NULL) {
+               return;
+       }
+
+       e->op_type = UNUSED_SHARE_MODE_ENTRY;
+       lck->modified = True;
+}
 
 /*******************************************************************
Print out a deferred open table.
Remove an oplock mid and mode entry from a share mode.
 ********************************************************************/
 
-static void print_deferred_open_table(struct deferred_open_data *data)
+bool remove_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
 {
-       int num_de_entries = data->u.num_deferred_open_entries;
-       deferred_open_entry *de_entries = (deferred_open_entry *)(data + 1);
-       int i;
+       struct share_mode_entry entry, *e;
+
+       /* Don't care about the pid owner being correct here - just a search. */
+       fill_share_mode_entry(&entry, fsp, (uid_t)-1, 0, NO_OPLOCK);
 
-       for (i = 0; i < num_de_entries; i++) {
-               deferred_open_entry *entry_p = &de_entries[i];
-               DEBUG(10,("print_deferred_open_table: %s\n", deferred_open_str(i, entry_p) ));
+       e = find_share_mode_entry(lck, &entry);
+       if (e == NULL) {
+               return False;
        }
-}
 
+       if (EXCLUSIVE_OPLOCK_TYPE(e->op_type)) {
+               /*
+                * Going from exclusive or batch,
+                * we always go through FAKE_LEVEL_II
+                * first.
+                */
+               if (!EXCLUSIVE_OPLOCK_TYPE(fsp->oplock_type)) {
+                       smb_panic("remove_share_oplock: logic error");
+               }
+               e->op_type = FAKE_LEVEL_II_OPLOCK;
+       } else {
+               e->op_type = NO_OPLOCK;
+       }
+       lck->modified = True;
+       return True;
+}
 
 /*******************************************************************
Form a static deferred open locking key for a dev/inode pair.
-******************************************************************/
Downgrade a oplock type from exclusive to level II.
+********************************************************************/
 
-static TDB_DATA deferred_open_locking_key(SMB_DEV_T dev, SMB_INO_T inode)
+bool downgrade_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
 {
-       static struct de_locking_key key;
-       TDB_DATA kbuf;
-
-       memset(&key, '\0', sizeof(key));
-       memcpy(&key.name[0], "DOE", 4);
-       key.dev = dev;
-       key.inode = inode;
-       kbuf.dptr = (char *)&key;
-       kbuf.dsize = sizeof(key);
-       return kbuf;
+       struct share_mode_entry entry, *e;
+
+       /* Don't care about the pid owner being correct here - just a search. */
+       fill_share_mode_entry(&entry, fsp, (uid_t)-1, 0, NO_OPLOCK);
+
+       e = find_share_mode_entry(lck, &entry);
+       if (e == NULL) {
+               return False;
+       }
+
+       e->op_type = LEVEL_II_OPLOCK;
+       lck->modified = True;
+       return True;
 }
 
-/*******************************************************************
Get all deferred open entries for a dev/inode pair.
-********************************************************************/
+/****************************************************************************
Check if setting delete on close is allowed on this fsp.
+****************************************************************************/
 
-int get_deferred_opens(connection_struct *conn, 
-                   SMB_DEV_T dev, SMB_INO_T inode, 
-                   deferred_open_entry **pp_de_entries)
+NTSTATUS can_set_delete_on_close(files_struct *fsp, uint32 dosmode)
 {
-       TDB_DATA dbuf;
-       struct deferred_open_data *data;
-       int num_de_entries;
-       deferred_open_entry *de_entries = NULL;
-       TDB_DATA key = deferred_open_locking_key(dev, inode);
+       /*
+        * Only allow delete on close for writable files.
+        */
 
-       *pp_de_entries = NULL;
+       if ((dosmode & FILE_ATTRIBUTE_READONLY) &&
+           !lp_delete_readonly(SNUM(fsp->conn))) {
+               DEBUG(10,("can_set_delete_on_close: file %s delete on close "
+                         "flag set but file attribute is readonly.\n",
+                         fsp_str_dbg(fsp)));
+               return NT_STATUS_CANNOT_DELETE;
+       }
 
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
-               return 0;
+       /*
+        * Only allow delete on close for writable shares.
+        */
 
-       data = (struct deferred_open_data *)dbuf.dptr;
-       num_de_entries = data->u.num_deferred_open_entries;
-       if(num_de_entries) {
-               pstring fname;
-               int i;
-               int del_count = 0;
+       if (!CAN_WRITE(fsp->conn)) {
+               DEBUG(10,("can_set_delete_on_close: file %s delete on "
+                         "close flag set but write access denied on share.\n",
+                         fsp_str_dbg(fsp)));
+               return NT_STATUS_ACCESS_DENIED;
+       }
 
-               de_entries = (deferred_open_entry *)memdup(dbuf.dptr + sizeof(*data),   
-                                               num_de_entries * sizeof(deferred_open_entry));
+       /*
+        * Only allow delete on close for files/directories opened with delete
+        * intent.
+        */
 
-               if (!de_entries) {
-                       SAFE_FREE(dbuf.dptr);
-                       return 0;
-               }
+       if (!(fsp->access_mask & DELETE_ACCESS)) {
+               DEBUG(10,("can_set_delete_on_close: file %s delete on "
+                         "close flag set but delete access denied.\n",
+                         fsp_str_dbg(fsp)));
+               return NT_STATUS_ACCESS_DENIED;
+       }
 
-               /* Save off the associated filename. */
-               pstrcpy(fname, dbuf.dptr + sizeof(*data) + num_de_entries * sizeof(deferred_open_entry));
+       /* Don't allow delete on close for non-empty directories. */
+       if (fsp->is_directory) {
+               SMB_ASSERT(!is_ntfs_stream_smb_fname(fsp->fsp_name));
+               return can_delete_directory(fsp->conn,
+                                           fsp->fsp_name->base_name);
+       }
 
-               /*
-                * Ensure that each entry has a real process attached.
-                */
+       return NT_STATUS_OK;
+}
 
-               for (i = 0; i < num_de_entries; ) {
-                       deferred_open_entry *entry_p = &de_entries[i];
-                       if (process_exists(entry_p->pid)) {
-                               DEBUG(10,("get_deferred_opens: %s\n", deferred_open_str(i, entry_p) ));
-                               i++;
-                       } else {
-                               DEBUG(10,("get_deferred_opens: deleted %s\n", deferred_open_str(i, entry_p) ));
-                               if (num_de_entries - i - 1 > 0) {
-                                       memcpy( &de_entries[i], &de_entries[i+1],
-                                               sizeof(deferred_open_entry) * (num_de_entries - i - 1));
-                               }
-                               num_de_entries--;
-                               del_count++;
-                       }
-               }
+/*************************************************************************
+ Return a talloced copy of a struct security_unix_token. NULL on fail.
+ (Should this be in locking.c.... ?).
+*************************************************************************/
 
-               /* Did we delete any ? If so, re-store in tdb. */
-               if (del_count) {
-                       data->u.num_deferred_open_entries = num_de_entries;
-                       
-                       if (num_de_entries) {
-                               memcpy(dbuf.dptr + sizeof(*data), de_entries,
-                                               num_de_entries * sizeof(deferred_open_entry));
-                               /* Append the filename. */
-                               pstrcpy(dbuf.dptr + sizeof(*data) + num_de_entries * sizeof(deferred_open_entry), fname);
-                       }
+static struct security_unix_token *copy_unix_token(TALLOC_CTX *ctx, const struct security_unix_token *tok)
+{
+       struct security_unix_token *cpy;
 
-                       /* The record has shrunk a bit */
-                       dbuf.dsize -= del_count * sizeof(deferred_open_entry);
+       cpy = TALLOC_P(ctx, struct security_unix_token);
+       if (!cpy) {
+               return NULL;
+       }
 
-                       if (data->u.num_deferred_open_entries == 0) {
-                               if (tdb_delete(tdb, key) == -1) {
-                                       SAFE_FREE(de_entries);
-                                       SAFE_FREE(dbuf.dptr);
-                                       return 0;
-                               }
-                       } else {
-                               if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1) {
-                                       SAFE_FREE(de_entries);
-                                       SAFE_FREE(dbuf.dptr);
-                                       return 0;
-                               }
-                       }
+       cpy->uid = tok->uid;
+       cpy->gid = tok->gid;
+       cpy->ngroups = tok->ngroups;
+       if (tok->ngroups) {
+               /* Make this a talloc child of cpy. */
+               cpy->groups = (gid_t *)talloc_memdup(
+                       cpy, tok->groups, tok->ngroups * sizeof(gid_t));
+               if (!cpy->groups) {
+                       TALLOC_FREE(cpy);
+                       return NULL;
                }
        }
-
-       SAFE_FREE(dbuf.dptr);
-       *pp_de_entries = de_entries;
-       return num_de_entries;
+       return cpy;
 }
 
-/*******************************************************************
Check if two deferred open entries are identical.
-********************************************************************/
+/****************************************************************************
Adds a delete on close token.
+****************************************************************************/
 
-static BOOL deferred_open_entries_identical( deferred_open_entry *e1, deferred_open_entry *e2)
+static bool add_delete_on_close_token(struct share_mode_lock *lck,
+                       uint32_t name_hash,
+                       const struct security_unix_token *tok)
 {
-       return (e1->pid == e2->pid &&
-               e1->mid == e2->mid &&
-               e1->port == e2->port &&
-               e1->dev == e2->dev &&
-               e1->inode == e2->inode &&
-               e1->time.tv_sec == e2->time.tv_sec &&
-               e1->time.tv_usec == e2->time.tv_usec);
+       struct delete_token_list *dtl;
+
+       dtl = TALLOC_ZERO_P(lck, struct delete_token_list);
+       if (dtl == NULL) {
+               return false;
+       }
+
+       dtl->name_hash = name_hash;
+       dtl->delete_token = copy_unix_token(lck, tok);
+       if (dtl->delete_token == NULL) {
+               TALLOC_FREE(dtl);
+               return false;
+       }
+       DLIST_ADD(lck->delete_tokens, dtl);
+       lck->modified = true;
+       return true;
 }
 
-/*******************************************************************
- Delete a specific deferred open entry.
- Ignore if no entry deleted.
-********************************************************************/
+/****************************************************************************
+ Sets the delete on close flag over all share modes on this file.
+ Modify the share mode entry for all files open
+ on this device and inode to tell other smbds we have
+ changed the delete on close flag. This will be noticed
+ in the close code, the last closer will delete the file
+ if flag is set.
+ This makes a copy of any struct security_unix_token into the
+ lck entry. This function is used when the lock is already granted.
+****************************************************************************/
 
-BOOL delete_deferred_open_entry(deferred_open_entry *entry)
+void set_delete_on_close_lck(files_struct *fsp,
+                       struct share_mode_lock *lck,
+                       bool delete_on_close,
+                       const struct security_unix_token *tok)
 {
-       TDB_DATA dbuf;
-       struct deferred_open_data *data;
-       int i, del_count=0;
-       deferred_open_entry *de_entries;
-       BOOL ret = True;
-       TDB_DATA key = deferred_open_locking_key(entry->dev, entry->inode);
+       struct delete_token_list *dtl;
+       bool ret;
 
-       /* read in the existing share modes */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr)
-               return -1;
+       if (delete_on_close) {
+               SMB_ASSERT(tok != NULL);
+       } else {
+               SMB_ASSERT(tok == NULL);
+       }
 
-       data = (struct deferred_open_data *)dbuf.dptr;
-       de_entries = (deferred_open_entry *)(dbuf.dptr + sizeof(*data));
+       for (dtl = lck->delete_tokens; dtl; dtl = dtl->next) {
+               if (dtl->name_hash == fsp->name_hash) {
+                       lck->modified = true;
+                       if (delete_on_close == false) {
+                               /* Delete this entry. */
+                               DLIST_REMOVE(lck->delete_tokens, dtl);
+                               TALLOC_FREE(dtl);
+                               return;
+                       }
+                       /* Replace this token with the
+                          given tok. */
+                       TALLOC_FREE(dtl->delete_token);
+                       dtl->delete_token = copy_unix_token(dtl, tok);
+                       SMB_ASSERT(dtl->delete_token != NULL);
+               }
+       }
 
-       /*
-        * Find any with this pid and delete it
-        * by overwriting with the rest of the data 
-        * from the record.
-        */
+       if (!delete_on_close) {
+               /* Nothing to delete - not found. */
+               return;
+       }
 
-       DEBUG(10,("delete_deferred_open_entry: num_deferred_open_entries = %d\n",
-               data->u.num_deferred_open_entries ));
+       ret = add_delete_on_close_token(lck, fsp->name_hash, tok);
+       SMB_ASSERT(ret);
+}
 
-       for (i=0;i<data->u.num_deferred_open_entries;) {
-               if (deferred_open_entries_identical(&de_entries[i], entry)) {
-                       DEBUG(10,("delete_deferred_open_entry: deleted %s\n",
-                               deferred_open_str(i, &de_entries[i]) ));
+bool set_delete_on_close(files_struct *fsp, bool delete_on_close, const struct security_unix_token *tok)
+{
+       struct share_mode_lock *lck;
 
-                       data->u.num_deferred_open_entries--;
-                       if ((dbuf.dsize - (sizeof(*data) + (i+1)*sizeof(*de_entries))) > 0) {
-                               memmove(&de_entries[i], &de_entries[i+1], 
-                                       dbuf.dsize - (sizeof(*data) + (i+1)*sizeof(*de_entries)));
-                       }
-                       del_count++;
+       DEBUG(10,("set_delete_on_close: %s delete on close flag for "
+                 "fnum = %d, file %s\n",
+                 delete_on_close ? "Adding" : "Removing", fsp->fnum,
+                 fsp_str_dbg(fsp)));
 
-                       DEBUG(10,("delete_deferred_open_entry: deleting entry %d\n", i ));
+       lck = get_share_mode_lock(talloc_tos(), fsp->file_id, NULL, NULL,
+                                 NULL);
+       if (lck == NULL) {
+               return False;
+       }
 
-               } else {
-                       i++;
-               }
+       set_delete_on_close_lck(fsp, lck, delete_on_close,
+                       delete_on_close ? tok : NULL);
+
+       if (fsp->is_directory) {
+               SMB_ASSERT(!is_ntfs_stream_smb_fname(fsp->fsp_name));
+               send_stat_cache_delete_message(fsp->conn->sconn->msg_ctx,
+                                              fsp->fsp_name->base_name);
        }
 
-       SMB_ASSERT(del_count == 0 || del_count == 1);
+       TALLOC_FREE(lck);
 
-       if (del_count) {
-               /* the record may have shrunk a bit */
-               dbuf.dsize -= del_count * sizeof(*de_entries);
+       fsp->delete_on_close = delete_on_close;
 
-               /* store it back in the database */
-               if (data->u.num_deferred_open_entries == 0) {
-                       if (tdb_delete(tdb, key) == -1)
-                               ret = False;
-               } else {
-                       if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-                               ret = False;
+       return True;
+}
+
+const struct security_unix_token *get_delete_on_close_token(struct share_mode_lock *lck, uint32_t name_hash)
+{
+       struct delete_token_list *dtl;
+
+       DEBUG(10,("get_delete_on_close_token: name_hash = 0x%x\n",
+                       (unsigned int)name_hash ));
+
+       for (dtl = lck->delete_tokens; dtl; dtl = dtl->next) {
+               DEBUG(10,("get_delete_on_close_token: dtl->name_hash = 0x%x\n",
+                               (unsigned int)dtl->name_hash ));
+               if (dtl->name_hash == name_hash) {
+                       return dtl->delete_token;
                }
        }
-       DEBUG(10,("delete_deferred_open_entry: Remaining table.\n"));
-       print_deferred_open_table((struct deferred_open_data*)dbuf.dptr);
-       SAFE_FREE(dbuf.dptr);
-       return ret;
+       return NULL;
 }
 
-/*******************************************************************
- Fill a deferred open entry.
-********************************************************************/
-
-static void fill_deferred_open(char *p, uint16 mid, struct timeval *ptv, SMB_DEV_T dev, SMB_INO_T inode, uint16 port)
+bool is_delete_on_close_set(struct share_mode_lock *lck, uint32_t name_hash)
 {
-       deferred_open_entry *e = (deferred_open_entry *)p;
-       void *x = &e->time; /* Needed to force alignment. p may not be aligned.... */
-
-       memset(e, '\0', sizeof(deferred_open_entry));
-       e->mid = mid;
-       e->pid = sys_getpid();
-       memcpy(x, ptv, sizeof(struct timeval));
-       e->dev = dev;
-       e->inode = inode;
-       e->port = port;
+       return (get_delete_on_close_token(lck, name_hash) != NULL);
 }
 
-/*******************************************************************
- Add a deferred open record. Return False on fail, True on success.
-********************************************************************/
+bool set_sticky_write_time(struct file_id fileid, struct timespec write_time)
+{
+       struct share_mode_lock *lck;
+
+       DEBUG(5,("set_sticky_write_time: %s id=%s\n",
+                timestring(talloc_tos(),
+                           convert_timespec_to_time_t(write_time)),
+                file_id_string_tos(&fileid)));
+
+       lck = get_share_mode_lock(NULL, fileid, NULL, NULL, NULL);
+       if (lck == NULL) {
+               return False;
+       }
+
+       if (timespec_compare(&lck->changed_write_time, &write_time) != 0) {
+               lck->modified = True;
+               lck->changed_write_time = write_time;
+       }
 
-BOOL add_deferred_open(uint16 mid, struct timeval *ptv, SMB_DEV_T dev, SMB_INO_T inode, uint16 port, const char *fname)
+       TALLOC_FREE(lck);
+       return True;
+}
+
+bool set_write_time(struct file_id fileid, struct timespec write_time)
 {
-       TDB_DATA dbuf;
-       struct deferred_open_data *data;
-       char *p=NULL;
-       int size;
-       TDB_DATA key = deferred_open_locking_key(dev, inode);
-       BOOL ret = True;
-               
-       /* read in the existing deferred open records if any */
-       dbuf = tdb_fetch(tdb, key);
-       if (!dbuf.dptr) {
-               size_t offset;
-               /* we'll need to create a new record */
-
-               size = sizeof(*data) + sizeof(deferred_open_entry) + strlen(fname) + 1;
-               p = (char *)SMB_MALLOC(size);
-               if (!p)
-                       return False;
-               data = (struct deferred_open_data *)p;
-               data->u.num_deferred_open_entries = 1;
-       
-               DEBUG(10,("add_deferred_open: creating entry for file %s. num_deferred_open_entries = 1\n",
-                       fname ));
-
-               offset = sizeof(*data) + sizeof(deferred_open_entry);
-               safe_strcpy(p + offset, fname, size - offset - 1);
-               fill_deferred_open(p + sizeof(*data), mid, ptv, dev, inode, port);
-               dbuf.dptr = p;
-               dbuf.dsize = size;
-               if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-                       ret = False;
-
-               print_deferred_open_table((struct deferred_open_data *)p);
-
-               SAFE_FREE(p);
-               return ret;
-       }
-
-       /* we're adding to an existing entry - this is a bit fiddly */
-       data = (struct deferred_open_data *)dbuf.dptr;
-
-       data->u.num_deferred_open_entries++;
-       
-       DEBUG(10,("add_deferred_open: adding entry for file %s. new num_deferred_open_entries = %d\n",
-               fname, data->u.num_deferred_open_entries ));
-
-       size = dbuf.dsize + sizeof(deferred_open_entry);
-       p = SMB_MALLOC(size);
-       if (!p) {
-               SAFE_FREE(dbuf.dptr);
+       struct share_mode_lock *lck;
+
+       DEBUG(5,("set_write_time: %s id=%s\n",
+                timestring(talloc_tos(),
+                           convert_timespec_to_time_t(write_time)),
+                file_id_string_tos(&fileid)));
+
+       lck = get_share_mode_lock(NULL, fileid, NULL, NULL, NULL);
+       if (lck == NULL) {
                return False;
        }
-       memcpy(p, dbuf.dptr, sizeof(*data));
-       fill_deferred_open(p + sizeof(*data), mid, ptv, dev, inode, port);
-       memcpy(p + sizeof(*data) + sizeof(deferred_open_entry), dbuf.dptr + sizeof(*data),
-              dbuf.dsize - sizeof(*data));
-       SAFE_FREE(dbuf.dptr);
-       dbuf.dptr = p;
-       dbuf.dsize = size;
-       if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1)
-               ret = False;
-       print_deferred_open_table((struct deferred_open_data *)p);
-       SAFE_FREE(p);
-       return ret;
+
+       if (timespec_compare(&lck->old_write_time, &write_time) != 0) {
+               lck->modified = True;
+               lck->old_write_time = write_time;
+       }
+
+       TALLOC_FREE(lck);
+       return True;
 }
 
-/****************************************************************************
- Traverse the whole database with this function, calling traverse_callback
- on each share mode
-****************************************************************************/
 
-static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA kbuf, TDB_DATA dbuf, 
-                       void* state)
+struct forall_state {
+       void (*fn)(const struct share_mode_entry *entry,
+                  const char *sharepath,
+                  const char *fname,
+                  void *private_data);
+       void *private_data;
+};
+
+static int traverse_fn(struct db_record *rec, void *_state)
 {
+       struct forall_state *state = (struct forall_state *)_state;
        struct locking_data *data;
-       share_mode_entry *shares;
-       char *name;
+       struct share_mode_entry *shares;
+       const char *sharepath;
+       const char *fname;
+       const char *del_tokens;
+       uint32_t total_del_token_size = 0;
        int i;
 
-       SHAREMODE_FN(traverse_callback) = (SHAREMODE_FN_CAST())state;
-
        /* Ensure this is a locking_key record. */
-       if (kbuf.dsize != sizeof(struct locking_key))
+       if (rec->key.dsize != sizeof(struct file_id))
                return 0;
 
-       data = (struct locking_data *)dbuf.dptr;
-       shares = (share_mode_entry *)(dbuf.dptr + sizeof(*data));
-       name = dbuf.dptr + sizeof(*data) + data->u.num_share_mode_entries*sizeof(*shares);
+       data = (struct locking_data *)rec->value.dptr;
+       shares = (struct share_mode_entry *)(rec->value.dptr + sizeof(*data));
+       del_tokens = (const char *)rec->value.dptr + sizeof(*data) +
+               data->u.s.num_share_mode_entries*sizeof(*shares);
+
+       for (i = 0; i < data->u.s.num_delete_token_entries; i++) {
+               uint32_t del_token_size;
+               memcpy(&del_token_size, del_tokens, sizeof(uint32_t));
+               total_del_token_size += del_token_size;
+               del_tokens += del_token_size;
+       }
 
-       for (i=0;i<data->u.num_share_mode_entries;i++) {
-               traverse_callback(&shares[i], name);
+       sharepath = (const char *)rec->value.dptr + sizeof(*data) +
+               data->u.s.num_share_mode_entries*sizeof(*shares) +
+               total_del_token_size;
+       fname = (const char *)rec->value.dptr + sizeof(*data) +
+               data->u.s.num_share_mode_entries*sizeof(*shares) +
+               total_del_token_size +
+               strlen(sharepath) + 1;
+
+       for (i=0;i<data->u.s.num_share_mode_entries;i++) {
+               state->fn(&shares[i], sharepath, fname,
+                         state->private_data);
        }
        return 0;
 }
@@ -1243,9 +1735,17 @@ static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA kbuf, TDB_DATA dbuf,
  share mode system.
 ********************************************************************/
 
-int share_mode_forall(SHAREMODE_FN(fn))
+int share_mode_forall(void (*fn)(const struct share_mode_entry *, const char *,
+                                const char *, void *),
+                     void *private_data)
 {
-       if (!tdb)
+       struct forall_state state;
+
+       if (lock_db == NULL)
                return 0;
-       return tdb_traverse(tdb, traverse_fn, (void*)fn);
+
+       state.fn = fn;
+       state.private_data = private_data;
+
+       return lock_db->traverse_read(lock_db, traverse_fn, (void *)&state);
 }