r23784: use the GPLv3 boilerplate as recommended by the FSF and the license text
[tprouty/samba.git] / source / locking / locking.c
index 0ffb937432629eba4045971f2dfa195f10e44682..c7fb572b1ddcb53a4f78ae33f5e64a0caba1d5e3 100644 (file)
@@ -7,7 +7,7 @@
    
    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,
@@ -16,8 +16,7 @@
    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:
 
 */
 
 #include "includes.h"
-uint16 global_smbpid;
 
 #undef DBGC_CLASS
 #define DBGC_CLASS DBGC_LOCKING
 
 /* the locking database handle */
-static TDB_CONTEXT *tdb;
+static struct db_context *lock_db;
 
 /****************************************************************************
  Debugging aids :-).
@@ -56,8 +54,10 @@ const char *lock_type_name(enum brl_type lock_type)
                        return "READ";
                case WRITE_LOCK:
                        return "WRITE";
-               case PENDING_LOCK:
-                       return "PENDING";
+               case PENDING_READ_LOCK:
+                       return "PENDING_READ";
+               case PENDING_WRITE_LOCK:
+                       return "PENDING_WRITE";
                default:
                        return "other";
        }
@@ -74,20 +74,20 @@ const char *lock_flav_name(enum brl_flavour lock_flav)
 ****************************************************************************/
 
 BOOL is_locked(files_struct *fsp,
+               uint32 smbpid,
                SMB_BIG_UINT count,
                SMB_BIG_UINT offset, 
                enum brl_type lock_type)
 {
-       int snum = SNUM(fsp->conn);
-       int strict_locking = lp_strict_locking(snum);
-       enum brl_flavour lock_flav = lp_posix_cifsu_locktype();
+       int strict_locking = lp_strict_locking(fsp->conn->params);
+       enum brl_flavour lock_flav = lp_posix_cifsu_locktype(fsp);
        BOOL ret = True;
        
        if (count == 0) {
                return False;
        }
 
-       if (!lp_locking(snum) || !strict_locking) {
+       if (!lp_locking(fsp->conn->params) || !strict_locking) {
                return False;
        }
 
@@ -100,32 +100,32 @@ BOOL is_locked(files_struct *fsp,
                        DEBUG(10,("is_locked: optimisation - level II oplock on file %s\n", fsp->fsp_name ));
                        ret = False;
                } else {
-                       struct byte_range_lock *br_lck = brl_get_locks(fsp);
+                       struct byte_range_lock *br_lck = brl_get_locks_readonly(NULL, fsp);
                        if (!br_lck) {
                                return False;
                        }
                        ret = !brl_locktest(br_lck,
-                                       global_smbpid,
+                                       smbpid,
                                        procid_self(),
                                        offset,
                                        count,
                                        lock_type,
                                        lock_flav);
-                       byte_range_lock_destructor(br_lck);
+                       TALLOC_FREE(br_lck);
                }
        } else {
-               struct byte_range_lock *br_lck = brl_get_locks(fsp);
+               struct byte_range_lock *br_lck = brl_get_locks_readonly(NULL, fsp);
                if (!br_lck) {
                        return False;
                }
                ret = !brl_locktest(br_lck,
-                               global_smbpid,
+                               smbpid,
                                procid_self(),
                                offset,
                                count,
                                lock_type,
                                lock_flav);
-               byte_range_lock_destructor(br_lck);
+               TALLOC_FREE(br_lck);
        }
 
        DEBUG(10,("is_locked: flavour = %s brl start=%.0f len=%.0f %s for fnum %d file %s\n",
@@ -141,7 +141,7 @@ BOOL is_locked(files_struct *fsp,
 ****************************************************************************/
 
 NTSTATUS query_lock(files_struct *fsp,
-                       uint16 *psmbpid,
+                       uint32 *psmbpid,
                        SMB_BIG_UINT *pcount,
                        SMB_BIG_UINT *poffset,
                        enum brl_type *plock_type,
@@ -150,15 +150,15 @@ NTSTATUS query_lock(files_struct *fsp,
        struct byte_range_lock *br_lck = NULL;
        NTSTATUS status = NT_STATUS_LOCK_NOT_GRANTED;
 
-       if (!OPEN_FSP(fsp) || !fsp->can_lock) {
-               return NT_STATUS_INVALID_HANDLE;
+       if (!fsp->can_lock) {
+               return fsp->is_directory ? NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
        }
 
-       if (!lp_locking(SNUM(fsp->conn))) {
+       if (!lp_locking(fsp->conn->params)) {
                return NT_STATUS_OK;
        }
 
-       br_lck = brl_get_locks(fsp);
+       br_lck = brl_get_locks_readonly(NULL, fsp);
        if (!br_lck) {
                return NT_STATUS_NO_MEMORY;
        }
@@ -171,7 +171,7 @@ NTSTATUS query_lock(files_struct *fsp,
                        plock_type,
                        lock_flav);
 
-       byte_range_lock_destructor(br_lck);
+       TALLOC_FREE(br_lck);
        return status;
 }
 
@@ -179,23 +179,27 @@ NTSTATUS query_lock(files_struct *fsp,
  Utility function called by locking requests.
 ****************************************************************************/
 
-NTSTATUS do_lock(files_struct *fsp,
-                       uint16 lock_pid,
+struct byte_range_lock *do_lock(struct messaging_context *msg_ctx,
+                       files_struct *fsp,
+                       uint32 lock_pid,
                        SMB_BIG_UINT count,
                        SMB_BIG_UINT offset,
                        enum brl_type lock_type,
                        enum brl_flavour lock_flav,
-                       BOOL *my_lock_ctx)
+                       BOOL blocking_lock,
+                       NTSTATUS *perr,
+                       uint32 *plock_pid)
 {
        struct byte_range_lock *br_lck = NULL;
-       NTSTATUS status = NT_STATUS_LOCK_NOT_GRANTED;
 
-       if (!OPEN_FSP(fsp) || !fsp->can_lock) {
-               return NT_STATUS_INVALID_HANDLE;
+       if (!fsp->can_lock) {
+               *perr = fsp->is_directory ? NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
+               return NULL;
        }
 
-       if (!lp_locking(SNUM(fsp->conn))) {
-               return NT_STATUS_OK;
+       if (!lp_locking(fsp->conn->params)) {
+               *perr = NT_STATUS_OK;
+               return NULL;
        }
 
        /* NOTE! 0 byte long ranges ARE allowed and should be stored  */
@@ -204,89 +208,89 @@ NTSTATUS do_lock(files_struct *fsp,
                lock_flav_name(lock_flav), lock_type_name(lock_type),
                (double)offset, (double)count, fsp->fnum, fsp->fsp_name ));
 
-       br_lck = brl_get_locks(fsp);
+       br_lck = brl_get_locks(NULL, fsp);
        if (!br_lck) {
-               return NT_STATUS_NO_MEMORY;
+               *perr = NT_STATUS_NO_MEMORY;
+               return NULL;
        }
 
-       status = brl_lock(br_lck,
+       *perr = brl_lock(msg_ctx,
+                       br_lck,
                        lock_pid,
                        procid_self(),
                        offset,
                        count, 
                        lock_type,
                        lock_flav,
-                       my_lock_ctx);
+                       blocking_lock,
+                       plock_pid);
 
-       byte_range_lock_destructor(br_lck);
-       return status;
+       /* 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++;
+
+       return br_lck;
 }
 
 /****************************************************************************
- 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 unlocking requests.
 ****************************************************************************/
 
-NTSTATUS do_lock_spin(files_struct *fsp,
-                       uint16 lock_pid,
+NTSTATUS do_unlock(struct messaging_context *msg_ctx,
+                       files_struct *fsp,
+                       uint32 lock_pid,
                        SMB_BIG_UINT count,
                        SMB_BIG_UINT offset,
-                       enum brl_type lock_type,
-                       enum brl_flavour lock_flav,
-                       BOOL *my_lock_ctx)
+                       enum brl_flavour lock_flav)
 {
-       int j, maxj = lp_lock_spin_count();
-       int sleeptime = lp_lock_sleep_time();
-       NTSTATUS status, ret;
-
-       if (maxj <= 0) {
-               maxj = 1;
+       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;
        }
+       
+       DEBUG(10,("do_unlock: unlock start=%.0f len=%.0f requested for fnum %d file %s\n",
+                 (double)offset, (double)count, fsp->fnum, fsp->fsp_name ));
 
-       ret = NT_STATUS_OK; /* to keep dumb compilers happy */
+       br_lck = brl_get_locks(NULL, fsp);
+       if (!br_lck) {
+               return NT_STATUS_NO_MEMORY;
+       }
 
-       for (j = 0; j < maxj; j++) {
-               status = do_lock(fsp,
-                               lock_pid,
-                               count,
-                               offset,
-                               lock_type,
-                               lock_flav,
-                               my_lock_ctx);
+       ok = brl_unlock(msg_ctx,
+                       br_lck,
+                       lock_pid,
+                       procid_self(),
+                       offset,
+                       count,
+                       lock_flav);
+   
+       TALLOC_FREE(br_lck);
 
-               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 (!ok) {
+               DEBUG(10,("do_unlock: returning ERRlock.\n" ));
+               return NT_STATUS_RANGE_NOT_LOCKED;
+       }
 
-                       /* Only spin for Windows locks. */
-                       if (lock_flav == POSIX_LOCK) {
-                               return ret;
-                       }
-               }
+       SMB_ASSERT(fsp->current_lock_count > 0);
+       fsp->current_lock_count--;
 
-               if (sleeptime) {
-                       sys_usleep(sleeptime);
-               }
-       }
-       return ret;
+       return NT_STATUS_OK;
 }
 
 /****************************************************************************
Utility function called by unlocking requests.
Cancel any pending blocked locks.
 ****************************************************************************/
 
-NTSTATUS do_unlock(files_struct *fsp,
-                       uint16 lock_pid,
+NTSTATUS do_lock_cancel(files_struct *fsp,
+                       uint32 lock_pid,
                        SMB_BIG_UINT count,
                        SMB_BIG_UINT offset,
                        enum brl_flavour lock_flav)
@@ -294,36 +298,40 @@ NTSTATUS do_unlock(files_struct *fsp,
        BOOL ok = False;
        struct byte_range_lock *br_lck = NULL;
        
-       if (!lp_locking(SNUM(fsp->conn))) {
-               return NT_STATUS_OK;
+       if (!fsp->can_lock) {
+               return fsp->is_directory ?
+                       NT_STATUS_INVALID_DEVICE_REQUEST : NT_STATUS_INVALID_HANDLE;
        }
        
-       if (!OPEN_FSP(fsp) || !fsp->can_lock) {
-               return NT_STATUS_INVALID_HANDLE;
+       if (!lp_locking(fsp->conn->params)) {
+               return NT_STATUS_DOS(ERRDOS, ERRcancelviolation);
        }
-       
-       DEBUG(10,("do_unlock: unlock start=%.0f len=%.0f requested for fnum %d file %s\n",
+
+       DEBUG(10,("do_lock_cancel: cancel start=%.0f len=%.0f requested for fnum %d file %s\n",
                  (double)offset, (double)count, fsp->fnum, fsp->fsp_name ));
 
-       br_lck = brl_get_locks(fsp);
+       br_lck = brl_get_locks(NULL, fsp);
        if (!br_lck) {
                return NT_STATUS_NO_MEMORY;
        }
 
-       ok = brl_unlock(br_lck,
+       ok = brl_lock_cancel(br_lck,
                        lock_pid,
                        procid_self(),
                        offset,
                        count,
                        lock_flav);
    
-       byte_range_lock_destructor(br_lck);
+       TALLOC_FREE(br_lck);
 
        if (!ok) {
-               DEBUG(10,("do_unlock: returning ERRlock.\n" ));
-               return NT_STATUS_RANGE_NOT_LOCKED;
+               DEBUG(10,("do_lock_cancel: returning ERRcancelviolation.\n" ));
+               return NT_STATUS_DOS(ERRDOS, ERRcancelviolation);
        }
 
+       SMB_ASSERT(fsp->current_lock_count > 0);
+       fsp->current_lock_count--;
+
        return NT_STATUS_OK;
 }
 
@@ -331,28 +339,29 @@ NTSTATUS do_unlock(files_struct *fsp,
  Remove any locks on this fd. Called from file_close().
 ****************************************************************************/
 
-void locking_close_file(files_struct *fsp)
+void locking_close_file(struct messaging_context *msg_ctx,
+                       files_struct *fsp)
 {
        struct byte_range_lock *br_lck;
-       struct process_id pid = procid_self();
 
-       if (!lp_locking(SNUM(fsp->conn)))
+       if (!lp_locking(fsp->conn->params)) {
                return;
+       }
 
-       /*
-        * Just release all the brl locks, no need to release individually.
+       /* If we have not outstanding locks or pending
+        * locks then we don't need to look in the lock db.
         */
 
-       br_lck = brl_get_locks(fsp);
-       if (br_lck) {
-               brl_close_fnum(br_lck, pid);
-               byte_range_lock_destructor(br_lck);
+       if (fsp->current_lock_count == 0) {
+               return;
        }
 
-       if(lp_posix_locking(SNUM(fsp->conn))) {
-               /* Release all the POSIX locks.*/
-               posix_locking_close_file(fsp);
+       br_lck = brl_get_locks(NULL,fsp);
 
+       if (br_lck) {
+               cancel_pending_lock_requests_by_fid(fsp, br_lck);
+               brl_close_fnum(msg_ctx, br_lck);
+               TALLOC_FREE(br_lck);
        }
 }
 
@@ -366,16 +375,17 @@ BOOL locking_init(int read_only)
 {
        brl_init(read_only);
 
-       if (tdb)
+       if (lock_db)
                return True;
 
-       tdb = tdb_open_log(lock_path("locking.tdb"), 
-                       lp_open_files_db_hash_size(),
-                       TDB_DEFAULT|(read_only?0x0:TDB_CLEAR_IF_FIRST), 
-                       read_only?O_RDONLY:O_RDWR|O_CREAT,
-                       0644);
+       lock_db = db_open(NULL, lock_path("locking.tdb"),
+                         lp_open_files_db_hash_size(),
+                         TDB_DEFAULT
+                         |TDB_VOLATILE
+                         |(read_only?0x0:TDB_CLEAR_IF_FIRST),
+                         read_only?O_RDONLY:O_RDWR|O_CREAT, 0644);
 
-       if (!tdb) {
+       if (!lock_db) {
                DEBUG(0,("ERROR: Failed to initialise locking database\n"));
                return False;
        }
@@ -394,40 +404,23 @@ BOOL locking_init(int read_only)
 
 BOOL locking_end(void)
 {
-       BOOL ret = True;
-
        brl_shutdown(open_read_only);
-       if (tdb) {
-               if (tdb_close(tdb) != 0)
-                       ret = False;
+       if (lock_db) {
+               TALLOC_FREE(lock_db);
        }
-
-       return ret;
+       return True;
 }
 
 /*******************************************************************
  Form a static locking key for a dev/inode pair.
 ******************************************************************/
 
-/* key and data records in the tdb locking database */
-struct locking_key {
-       SMB_DEV_T dev;
-       SMB_INO_T ino;
-};
-
-/*******************************************************************
- Form a static locking key for a dev/inode pair.
-******************************************************************/
-
-static TDB_DATA locking_key(SMB_DEV_T dev, SMB_INO_T inode)
+static TDB_DATA locking_key(struct file_id id)
 {
-       static struct locking_key key;
+       static struct file_id key;      
        TDB_DATA kbuf;
-
-       memset(&key, '\0', sizeof(key));
-       key.dev = dev;
-       key.ino = inode;
-       kbuf.dptr = (char *)&key;
+       key = id;
+       kbuf.dptr = (uint8 *)&key;
        kbuf.dsize = sizeof(key);
        return kbuf;
 }
@@ -442,14 +435,15 @@ char *share_mode_str(int num, struct share_mode_entry *e)
 
        slprintf(share_str, sizeof(share_str)-1, "share_mode_entry[%d]: %s "
                 "pid = %s, share_access = 0x%x, private_options = 0x%x, "
-                "access_mask = 0x%x, mid = 0x%x, type= 0x%x, file_id = %lu, "
-                "dev = 0x%x, inode = %.0f",
+                "access_mask = 0x%x, mid = 0x%x, type= 0x%x, gen_id = %lu, "
+                "uid = %u, flags = %u, file_id %s",
                 num,
                 e->op_type == UNUSED_SHARE_MODE_ENTRY ? "UNUSED" : "",
                 procid_str_static(&e->pid),
                 e->share_access, e->private_options,
                 e->access_mask, e->op_mid, e->op_type, e->share_file_id,
-                (unsigned int)e->dev, (double)e->inode );
+                (unsigned int)e->uid, (unsigned int)e->flags,
+                file_id_static_string(&e->id));
 
        return share_str;
 }
@@ -484,26 +478,23 @@ static BOOL parse_share_modes(TDB_DATA dbuf, struct share_mode_lock *lck)
        int i;
 
        if (dbuf.dsize < sizeof(struct locking_data)) {
-               smb_panic("PANIC: parse_share_modes: buffer too short.\n");
+               smb_panic("parse_share_modes: buffer too short");
        }
 
        data = (struct locking_data *)dbuf.dptr;
 
        lck->delete_on_close = data->u.s.delete_on_close;
-       lck->initial_delete_on_close = data->u.s.initial_delete_on_close;
        lck->num_share_modes = data->u.s.num_share_mode_entries;
 
        DEBUG(10, ("parse_share_modes: delete_on_close: %d, "
-                  "initial_delete_on_close: %d, "
                   "num_share_modes: %d\n",
                lck->delete_on_close,
-               lck->initial_delete_on_close,
                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("PANIC: invalid number of share modes");
+               smb_panic("parse_share_modes: invalid number of share modes");
        }
 
        lck->share_modes = NULL;
@@ -513,21 +504,22 @@ static BOOL parse_share_modes(TDB_DATA dbuf, struct share_mode_lock *lck)
                if (dbuf.dsize < (sizeof(struct locking_data) +
                                  (lck->num_share_modes *
                                   sizeof(struct share_mode_entry)))) {
-                       smb_panic("PANIC: parse_share_modes: buffer too short.\n");
+                       smb_panic("parse_share_modes: buffer too short");
                }
                                  
-               lck->share_modes = talloc_memdup(lck, dbuf.dptr+sizeof(*data),
-                                                lck->num_share_modes *
-                                                sizeof(struct share_mode_entry));
+               lck->share_modes = (struct share_mode_entry *)
+                       TALLOC_MEMDUP(lck, dbuf.dptr+sizeof(*data),
+                                     lck->num_share_modes *
+                                     sizeof(struct share_mode_entry));
 
                if (lck->share_modes == NULL) {
-                       smb_panic("talloc failed\n");
+                       smb_panic("parse_share_modes: talloc failed");
                }
        }
 
        /* Get any delete token. */
        if (data->u.s.delete_token_size) {
-               char *p = dbuf.dptr + sizeof(*data) +
+               uint8 *p = dbuf.dptr + sizeof(*data) +
                                (lck->num_share_modes *
                                sizeof(struct share_mode_entry));
 
@@ -535,12 +527,12 @@ static BOOL parse_share_modes(TDB_DATA dbuf, struct share_mode_lock *lck)
                                ((data->u.s.delete_token_size - sizeof(uid_t)) % sizeof(gid_t)) != 0) {
                        DEBUG(0, ("parse_share_modes: invalid token size %d\n",
                                data->u.s.delete_token_size));
-                       smb_panic("parse_share_modes: invalid token size\n");
+                       smb_panic("parse_share_modes: invalid token size");
                }
 
                lck->delete_token = TALLOC_P(lck, UNIX_USER_TOKEN);
                if (!lck->delete_token) {
-                       smb_panic("talloc failed\n");
+                       smb_panic("parse_share_modes: talloc failed");
                }
 
                /* Copy out the uid and gid. */
@@ -559,7 +551,7 @@ static BOOL parse_share_modes(TDB_DATA dbuf, struct share_mode_lock *lck)
                        lck->delete_token->groups = TALLOC_ARRAY(lck->delete_token, gid_t,
                                                        lck->delete_token->ngroups);
                        if (!lck->delete_token) {
-                               smb_panic("talloc failed\n");
+                               smb_panic("parse_share_modes: talloc failed");
                        }
 
                        for (i = 0; i < lck->delete_token->ngroups; i++) {
@@ -573,16 +565,22 @@ static BOOL parse_share_modes(TDB_DATA dbuf, struct share_mode_lock *lck)
        }
 
        /* Save off the associated service path and filename. */
-       lck->servicepath = talloc_strdup(lck, dbuf.dptr + sizeof(*data) +
+       lck->servicepath = talloc_strdup(lck, (const char *)dbuf.dptr + sizeof(*data) +
                                        (lck->num_share_modes *
                                        sizeof(struct share_mode_entry)) +
                                        data->u.s.delete_token_size );
+       if (lck->servicepath == NULL) {
+               smb_panic("parse_share_modes: talloc_strdup failed");
+       }
 
-       lck->filename = talloc_strdup(lck, dbuf.dptr + sizeof(*data) +
+       lck->filename = talloc_strdup(lck, (const char *)dbuf.dptr + sizeof(*data) +
                                        (lck->num_share_modes *
                                        sizeof(struct share_mode_entry)) +
                                        data->u.s.delete_token_size +
                                        strlen(lck->servicepath) + 1 );
+       if (lck->filename == NULL) {
+               smb_panic("parse_share_modes: talloc_strdup failed");
+       }
 
        /*
         * Ensure that each entry has a real process attached.
@@ -635,21 +633,19 @@ static TDB_DATA unparse_share_modes(struct share_mode_lock *lck)
                delete_token_size +
                sp_len + 1 +
                strlen(lck->filename) + 1;
-       result.dptr = talloc_size(lck, result.dsize);
+       result.dptr = TALLOC_ARRAY(lck, uint8, result.dsize);
 
        if (result.dptr == NULL) {
-               smb_panic("talloc failed\n");
+               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.delete_on_close = lck->delete_on_close;
-       data->u.s.initial_delete_on_close = lck->initial_delete_on_close;
        data->u.s.delete_token_size = delete_token_size;
-       DEBUG(10, ("unparse_share_modes: del: %d, initial del %d, tok = %u, num: %d\n",
+       DEBUG(10, ("unparse_share_modes: del: %d, tok = %u, num: %d\n",
                data->u.s.delete_on_close,
-               data->u.s.initial_delete_on_close,
                (unsigned int)data->u.s.delete_token_size,
                data->u.s.num_share_mode_entries));
        memcpy(result.dptr + sizeof(*data), lck->share_modes,
@@ -659,7 +655,7 @@ static TDB_DATA unparse_share_modes(struct share_mode_lock *lck)
 
        /* Store any delete on close token. */
        if (lck->delete_token) {
-               char *p = result.dptr + offset;
+               uint8 *p = result.dptr + offset;
 
                memcpy(p, &lck->delete_token->uid, sizeof(uid_t));
                p += sizeof(uid_t);
@@ -674,10 +670,10 @@ static TDB_DATA unparse_share_modes(struct share_mode_lock *lck)
                offset = p - result.dptr;
        }
 
-       safe_strcpy(result.dptr + offset, lck->servicepath,
+       safe_strcpy((char *)result.dptr + offset, lck->servicepath,
                    result.dsize - offset - 1);
        offset += sp_len + 1;
-       safe_strcpy(result.dptr + offset, lck->filename,
+       safe_strcpy((char *)result.dptr + offset, lck->filename,
                    result.dsize - offset - 1);
 
        if (DEBUGLEVEL >= 10) {
@@ -687,15 +683,13 @@ static TDB_DATA unparse_share_modes(struct share_mode_lock *lck)
        return result;
 }
 
-static int share_mode_lock_destructor(void *p)
+static int share_mode_lock_destructor(struct share_mode_lock *lck)
 {
-       struct share_mode_lock *lck =
-               talloc_get_type_abort(p, struct share_mode_lock);
-       TDB_DATA key = locking_key(lck->dev, lck->ino);
+       NTSTATUS status;
        TDB_DATA data;
 
        if (!lck->modified) {
-               goto done;
+               return 0;
        }
 
        data = unparse_share_modes(lck);
@@ -703,91 +697,130 @@ static int share_mode_lock_destructor(void *p)
        if (data.dptr == NULL) {
                if (!lck->fresh) {
                        /* There has been an entry before, delete it */
-                       if (tdb_delete(tdb, key) == -1) {
-                               smb_panic("Could not delete share entry\n");
+
+                       status = lck->record->delete_rec(lck->record);
+                       if (!NT_STATUS_IS_OK(status)) {
+                               DEBUG(0, ("delete_rec returned %s\n",
+                                         nt_errstr(status)));
+                               smb_panic("could not delete share entry");
                        }
                }
                goto done;
        }
 
-       if (tdb_store(tdb, key, data, TDB_REPLACE) == -1) {
-               smb_panic("Could not store share mode entry\n");
+       status = lck->record->store(lck->record, data, TDB_REPLACE);
+       if (!NT_STATUS_IS_OK(status)) {
+               DEBUG(0, ("store returned %s\n", nt_errstr(status)));
+               smb_panic("could not store share mode entry");
        }
 
  done:
-       tdb_chainunlock(tdb, key);
 
        return 0;
 }
 
-struct share_mode_lock *get_share_mode_lock(TALLOC_CTX *mem_ctx,
-                                               SMB_DEV_T dev, SMB_INO_T ino,
-                                               const char *servicepath,
-                                               const char *fname)
+static BOOL fill_share_mode_lock(struct share_mode_lock *lck,
+                                struct file_id id,
+                                const char *servicepath,
+                                const char *fname,
+                                TDB_DATA share_mode_data)
 {
-       struct share_mode_lock *lck;
-       TDB_DATA key = locking_key(dev, ino);
-       TDB_DATA data;
-
-       lck = TALLOC_P(mem_ctx, struct share_mode_lock);
-       if (lck == NULL) {
-               DEBUG(0, ("talloc failed\n"));
-               return NULL;
-       }
-
        /* Ensure we set every field here as the destructor must be
           valid even if parse_share_modes fails. */
 
        lck->servicepath = NULL;
        lck->filename = NULL;
-       lck->dev = dev;
-       lck->ino = ino;
+       lck->id = id;
        lck->num_share_modes = 0;
        lck->share_modes = NULL;
        lck->delete_token = NULL;
        lck->delete_on_close = False;
-       lck->initial_delete_on_close = False;
        lck->fresh = False;
        lck->modified = False;
 
-       if (tdb_chainlock(tdb, key) != 0) {
-               DEBUG(3, ("Could not lock share entry\n"));
-               TALLOC_FREE(lck);
-               return NULL;
-       }
-
-       /* We must set the destructor immediately after the chainlock
-          ensure the lock is cleaned up on any of the error return
-          paths below. */
-
-       talloc_set_destructor(lck, share_mode_lock_destructor);
-
-       data = tdb_fetch(tdb, key);
-       lck->fresh = (data.dptr == NULL);
+       lck->fresh = (share_mode_data.dptr == NULL);
 
        if (lck->fresh) {
-
                if (fname == NULL || servicepath == NULL) {
-                       TALLOC_FREE(lck);
-                       return NULL;
+                       return False;
                }
                lck->filename = talloc_strdup(lck, fname);
                lck->servicepath = talloc_strdup(lck, servicepath);
                if (lck->filename == NULL || lck->servicepath == NULL) {
                        DEBUG(0, ("talloc failed\n"));
-                       TALLOC_FREE(lck);
-                       return NULL;
+                       return False;
                }
        } else {
-               if (!parse_share_modes(data, lck)) {
+               if (!parse_share_modes(share_mode_data, lck)) {
                        DEBUG(0, ("Could not parse share modes\n"));
-                       TALLOC_FREE(lck);
-                       SAFE_FREE(data.dptr);
-                       return NULL;
+                       return False;
                }
        }
 
-       SAFE_FREE(data.dptr);
+       return True;
+}
+
+struct share_mode_lock *get_share_mode_lock(TALLOC_CTX *mem_ctx,
+                                           struct file_id id,
+                                           const char *servicepath,
+                                           const char *fname)
+{
+       struct share_mode_lock *lck;
+       TDB_DATA key;
+
+       key.dptr = (unsigned char *)&id;
+       key.dsize = sizeof(id);
+
+       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, fname,
+                                 lck->record->value)) {
+               DEBUG(3, ("fill_share_mode_lock failed\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       talloc_set_destructor(lck, share_mode_lock_destructor);
+
+       return lck;
+}
+
+struct share_mode_lock *fetch_share_mode_unlocked(TALLOC_CTX *mem_ctx,
+                                                 struct file_id id,
+                                                 const char *servicepath,
+                                                 const char *fname)
+{
+       struct share_mode_lock *lck;
+       TDB_DATA key = locking_key(id);
+       TDB_DATA data;
+
+       if (!(lck = TALLOC_P(mem_ctx, struct share_mode_lock))) {
+               DEBUG(0, ("talloc failed\n"));
+               return NULL;
+       }
+
+       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, servicepath, fname, data)) {
+               DEBUG(3, ("fill_share_mode_lock failed\n"));
+               TALLOC_FREE(lck);
+               return NULL;
+       }
+
+       TALLOC_FREE(data.dptr);
 
        return lck;
 }
@@ -799,7 +832,8 @@ struct share_mode_lock *get_share_mode_lock(TALLOC_CTX *mem_ctx,
  Based on an initial code idea from SATOH Fumiyasu <fumiya@samba.gr.jp>
 ********************************************************************/
 
-BOOL rename_share_filename(struct share_mode_lock *lck,
+BOOL rename_share_filename(struct messaging_context *msg_ctx,
+                       struct share_mode_lock *lck,
                        const char *servicepath,
                        const char *newname)
 {
@@ -809,10 +843,6 @@ BOOL rename_share_filename(struct share_mode_lock *lck,
        char *frm = NULL;
        int i;
 
-       if (!lck) {
-               return False;
-       }
-
        DEBUG(10, ("rename_share_filename: servicepath %s newname %s\n",
                servicepath, newname));
 
@@ -838,13 +868,12 @@ BOOL rename_share_filename(struct share_mode_lock *lck,
        msg_len = MSG_FILE_RENAMED_MIN_SIZE + sp_len + 1 + fn_len + 1;
 
        /* Set up the name changed message. */
-       frm = TALLOC(lck, msg_len);
+       frm = TALLOC_ARRAY(lck, char, msg_len);
        if (!frm) {
                return False;
        }
 
-       SDEV_T_VAL(frm,0,lck->dev);
-       SINO_T_VAL(frm,8,lck->ino);
+       push_file_id_16(frm, &lck->id);
 
        DEBUG(10,("rename_share_filename: msg_len = %u\n", (unsigned int)msg_len ));
 
@@ -862,26 +891,25 @@ BOOL rename_share_filename(struct share_mode_lock *lck,
                        continue;
                }
 
-               DEBUG(10,("rename_share_filename: sending rename message to pid %u "
-                       "dev %x, inode  %.0f sharepath %s newname %s\n",
-                       (unsigned int)procid_to_pid(&se->pid),
-                       (unsigned int)lck->dev, (double)lck->ino,
-                       lck->servicepath, lck->filename ));
+               DEBUG(10,("rename_share_filename: sending rename message to pid %s "
+                         "file_id %s sharepath %s newname %s\n",
+                         procid_str_static(&se->pid),
+                         file_id_static_string(&lck->id),
+                         lck->servicepath, lck->filename ));
 
-               become_root();
-               message_send_pid(se->pid, MSG_SMB_FILE_RENAME,
-                               frm, msg_len, True);
-               unbecome_root();
+               messaging_send_buf(msg_ctx, se->pid, MSG_SMB_FILE_RENAME,
+                                  (uint8 *)frm, msg_len);
        }
 
        return True;
 }
 
-BOOL get_delete_on_close_flag(SMB_DEV_T dev, SMB_INO_T inode)
+BOOL get_delete_on_close_flag(struct file_id id)
 {
        BOOL result;
-       struct share_mode_lock *lck = get_share_mode_lock(NULL, dev, inode, NULL, NULL);
-       if (!lck) {
+       struct share_mode_lock *lck;
+  
+       if (!(lck = fetch_share_mode_unlocked(NULL, id, NULL, NULL))) {
                return False;
        }
        result = lck->delete_on_close;
@@ -917,7 +945,7 @@ BOOL is_unused_share_mode_entry(const struct share_mode_entry *e)
 
 static void fill_share_mode_entry(struct share_mode_entry *e,
                                  files_struct *fsp,
-                                 uint16 mid, uint16 op_type)
+                                 uid_t uid, uint16 mid, uint16 op_type)
 {
        ZERO_STRUCTP(e);
        e->pid = procid_self();
@@ -928,14 +956,15 @@ static void fill_share_mode_entry(struct share_mode_entry *e,
        e->op_type = op_type;
        e->time.tv_sec = fsp->open_time.tv_sec;
        e->time.tv_usec = fsp->open_time.tv_usec;
-       e->share_file_id = fsp->fh->file_id;
-       e->dev = fsp->dev;
-       e->inode = fsp->inode;
+       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;
 }
 
 static void fill_deferred_open_entry(struct share_mode_entry *e,
                                     const struct timeval request_time,
-                                    SMB_DEV_T dev, SMB_INO_T ino, uint16 mid)
+                                    struct file_id id, uint16 mid)
 {
        ZERO_STRUCTP(e);
        e->pid = procid_self();
@@ -943,8 +972,9 @@ static void fill_deferred_open_entry(struct share_mode_entry *e,
        e->op_type = DEFERRED_OPEN_ENTRY;
        e->time.tv_sec = request_time.tv_sec;
        e->time.tv_usec = request_time.tv_usec;
-       e->dev = dev;
-       e->inode = ino;
+       e->id = id;
+       e->uid = (uint32)-1;
+       e->flags = 0;
 }
 
 static void add_share_mode_entry(struct share_mode_lock *lck,
@@ -969,19 +999,22 @@ static void add_share_mode_entry(struct share_mode_lock *lck,
 }
 
 void set_share_mode(struct share_mode_lock *lck, files_struct *fsp,
-                   uint16 mid, uint16 op_type)
+                       uid_t uid, uint16 mid, uint16 op_type, BOOL initial_delete_on_close_allowed)
 {
        struct share_mode_entry entry;
-       fill_share_mode_entry(&entry, fsp, mid, op_type);
+       fill_share_mode_entry(&entry, fsp, uid, mid, op_type);
+       if (initial_delete_on_close_allowed) {
+               entry.flags |= SHARE_MODE_ALLOW_INITIAL_DELETE_ON_CLOSE;
+       }
        add_share_mode_entry(lck, &entry);
 }
 
 void add_deferred_open(struct share_mode_lock *lck, uint16 mid,
                       struct timeval request_time,
-                      SMB_DEV_T dev, SMB_INO_T ino)
+                      struct file_id id)
 {
        struct share_mode_entry entry;
-       fill_deferred_open_entry(&entry, request_time, dev, ino, mid);
+       fill_deferred_open_entry(&entry, request_time, id, mid);
        add_share_mode_entry(lck, &entry);
 }
 
@@ -1000,8 +1033,7 @@ static BOOL share_modes_identical(struct share_mode_entry *e1,
           fsp->share_access field. */
 
        return (procid_equal(&e1->pid, &e2->pid) &&
-               e1->dev == e2->dev &&
-               e1->inode == e2->inode &&
+               file_id_equal(&e1->id, &e2->id) &&
                e1->share_file_id == e2->share_file_id );
 }
 
@@ -1010,8 +1042,7 @@ static BOOL deferred_open_identical(struct share_mode_entry *e1,
 {
        return (procid_equal(&e1->pid, &e2->pid) &&
                (e1->op_mid == e2->op_mid) &&
-               (e1->dev == e2->dev) &&
-               (e1->inode == e2->inode));
+               file_id_equal(&e1->id, &e2->id));
 }
 
 static struct share_mode_entry *find_share_mode_entry(struct share_mode_lock *lck,
@@ -1044,7 +1075,8 @@ BOOL del_share_mode(struct share_mode_lock *lck, files_struct *fsp)
 {
        struct share_mode_entry entry, *e;
 
-       fill_share_mode_entry(&entry, fsp, 0, NO_OPLOCK);
+       /* 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) {
@@ -1061,7 +1093,7 @@ void del_deferred_open_entry(struct share_mode_lock *lck, uint16 mid)
        struct share_mode_entry entry, *e;
 
        fill_deferred_open_entry(&entry, timeval_zero(),
-                                lck->dev, lck->ino, mid);
+                                lck->id, mid);
 
        e = find_share_mode_entry(lck, &entry);
        if (e == NULL) {
@@ -1080,7 +1112,8 @@ BOOL remove_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
 {
        struct share_mode_entry entry, *e;
 
-       fill_share_mode_entry(&entry, fsp, 0, NO_OPLOCK);
+       /* 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) {
@@ -1101,7 +1134,8 @@ BOOL downgrade_share_oplock(struct share_mode_lock *lck, files_struct *fsp)
 {
        struct share_mode_entry entry, *e;
 
-       fill_share_mode_entry(&entry, fsp, 0, NO_OPLOCK);
+       /* 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) {
@@ -1161,9 +1195,30 @@ NTSTATUS can_set_delete_on_close(files_struct *fsp, BOOL delete_on_close,
                return NT_STATUS_ACCESS_DENIED;
        }
 
+       /* Don't allow delete on close for non-empty directories. */
+       if (fsp->is_directory) {
+               return can_delete_directory(fsp->conn, fsp->fsp_name);
+       }
+
        return NT_STATUS_OK;
 }
 
+/****************************************************************************
+ Do we have an open file handle that created this entry ?
+****************************************************************************/
+
+BOOL can_set_initial_delete_on_close(const struct share_mode_lock *lck)
+{
+       int i;
+
+       for (i=0; i<lck->num_share_modes; i++) {
+               if (lck->share_modes[i].flags & SHARE_MODE_ALLOW_INITIAL_DELETE_ON_CLOSE) {
+                       return True;
+               }
+       }
+       return False;
+}
+
 /*************************************************************************
  Return a talloced copy of a UNIX_USER_TOKEN. NULL on fail.
  (Should this be in locking.c.... ?).
@@ -1220,11 +1275,22 @@ void set_delete_on_close_token(struct share_mode_lock *lck, UNIX_USER_TOKEN *tok
  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.
- Note that setting this to any value clears the initial_delete_on_close flag.
- If delete_on_close is True this makes a copy of any UNIX_USER_TOKEN into the
- lck entry.
+ This makes a copy of any UNIX_USER_TOKEN into the
+ lck entry. This function is used when the lock is already granted.
 ****************************************************************************/
 
+void set_delete_on_close_lck(struct share_mode_lock *lck, BOOL delete_on_close, UNIX_USER_TOKEN *tok)
+{
+       if (lck->delete_on_close != delete_on_close) {
+               set_delete_on_close_token(lck, tok);
+               lck->delete_on_close = delete_on_close;
+               if (delete_on_close) {
+                       SMB_ASSERT(lck->delete_token != NULL);
+               }
+               lck->modified = True;
+       }
+}
+
 BOOL set_delete_on_close(files_struct *fsp, BOOL delete_on_close, UNIX_USER_TOKEN *tok)
 {
        struct share_mode_lock *lck;
@@ -1238,55 +1304,80 @@ BOOL set_delete_on_close(files_struct *fsp, BOOL delete_on_close, UNIX_USER_TOKE
                return True;
        }
 
-       lck = get_share_mode_lock(NULL, fsp->dev, fsp->inode, NULL, NULL);
+       lck = get_share_mode_lock(NULL, fsp->file_id, NULL, NULL);
        if (lck == NULL) {
                return False;
        }
 
-       if (lck->delete_on_close != delete_on_close) {
-               set_delete_on_close_token(lck, tok);
-               lck->delete_on_close = delete_on_close;
-               if (delete_on_close) {
-                       SMB_ASSERT(lck->delete_token != NULL);
-               }
-               lck->modified = True;
-       }
+       set_delete_on_close_lck(lck, delete_on_close, tok);
 
-       if (lck->initial_delete_on_close) {
-               lck->initial_delete_on_close = False;
-               lck->modified = True;
+       if (fsp->is_directory) {
+               send_stat_cache_delete_message(fsp->fsp_name);
        }
 
        TALLOC_FREE(lck);
        return True;
 }
 
-static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA kbuf, TDB_DATA dbuf, 
-                       void *state)
+/****************************************************************************
+ Sets the allow initial delete on close flag for this share mode.
+****************************************************************************/
+
+BOOL set_allow_initial_delete_on_close(struct share_mode_lock *lck, files_struct *fsp, BOOL delete_on_close)
 {
+       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;
+       }
+
+       if (delete_on_close) {
+               e->flags |= SHARE_MODE_ALLOW_INITIAL_DELETE_ON_CLOSE;
+       } else {
+               e->flags &= ~SHARE_MODE_ALLOW_INITIAL_DELETE_ON_CLOSE;
+       }
+       lck->modified = True;
+       return True;
+}
+
+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;
        struct share_mode_entry *shares;
        const char *sharepath;
        const char *fname;
        int i;
-       void (*traverse_callback)(struct share_mode_entry *, const char *, const char *) = 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 = (struct share_mode_entry *)(dbuf.dptr + sizeof(*data));
-       sharepath = dbuf.dptr + sizeof(*data) +
+       data = (struct locking_data *)rec->value.dptr;
+       shares = (struct share_mode_entry *)(rec->value.dptr + sizeof(*data));
+       sharepath = (const char *)rec->value.dptr + sizeof(*data) +
                data->u.s.num_share_mode_entries*sizeof(*shares) +
                data->u.s.delete_token_size;
-       fname = dbuf.dptr + sizeof(*data) +
+       fname = (const char *)rec->value.dptr + sizeof(*data) +
                data->u.s.num_share_mode_entries*sizeof(*shares) +
                data->u.s.delete_token_size +
                strlen(sharepath) + 1;
 
        for (i=0;i<data->u.s.num_share_mode_entries;i++) {
-               traverse_callback(&shares[i], sharepath, fname);
+               state->fn(&shares[i], sharepath, fname,
+                         state->private_data);
        }
        return 0;
 }
@@ -1296,9 +1387,17 @@ static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA kbuf, TDB_DATA dbuf,
  share mode system.
 ********************************************************************/
 
-int share_mode_forall(void (*fn)(const struct share_mode_entry *, const char *, const char *))
+int share_mode_forall(void (*fn)(const struct share_mode_entry *, const char *,
+                                const char *, void *),
+                     void *private_data)
 {
-       if (tdb == NULL)
+       struct forall_state state;
+
+       if (lock_db == NULL)
                return 0;
-       return tdb_traverse(tdb, traverse_fn, fn);
+
+       state.fn = fn;
+       state.private_data = private_data;
+
+       return lock_db->traverse_read(lock_db, traverse_fn, (void *)&state);
 }