first public release of samba4 code

(This used to be commit b0510b5428b3461aeb9bbe3cc95f62fc73e2b97f)

15 files changed, 4965 insertions, 0 deletions

diff --git a/source4/tdb/.cvsignore b/source4/tdb/.cvsignore
new file mode 100644
index 0000000000..66445fe269
--- /dev/null
+++ b/source4/tdb/.cvsignore
@@ -0,0 +1,11 @@
+*.po
+*.po32
+tdbbackup
+tdbdump
+tdbtest
+tdbtool
+tdbtorture
+test.db
+test.gdbm
+test.tdb
+torture.tdb
diff --git a/source4/tdb/Makefile b/source4/tdb/Makefile
new file mode 100644
index 0000000000..59fbb079bd
--- /dev/null
+++ b/source4/tdb/Makefile
@@ -0,0 +1,29 @@
+#
+# Makefile for tdb directory
+#
+
+CFLAGS = -DSTANDALONE -DTDB_DEBUG -g -DHAVE_MMAP=1
+CC = gcc
+
+PROGS = tdbtest tdbtool tdbtorture
+TDB_OBJ = tdb.o spinlock.o
+
+default: $(PROGS)
+
+tdbtest: tdbtest.o $(TDB_OBJ)
+	$(CC) $(CFLAGS) -o tdbtest tdbtest.o $(TDB_OBJ) -lgdbm
+
+tdbtool: tdbtool.o $(TDB_OBJ)
+	$(CC) $(CFLAGS) -o tdbtool tdbtool.o $(TDB_OBJ)
+
+tdbtorture: tdbtorture.o $(TDB_OBJ)
+	$(CC) $(CFLAGS) -o tdbtorture tdbtorture.o $(TDB_OBJ)
+
+tdbdump: tdbdump.o $(TDB_OBJ)
+	$(CC) $(CFLAGS) -o tdbdump tdbdump.o $(TDB_OBJ)
+
+tdbbackup: tdbbackup.o $(TDB_OBJ)
+	$(CC) $(CFLAGS) -o tdbbackup tdbbackup.o $(TDB_OBJ)
+
+clean:
+	rm -f $(PROGS) *.o *~ *% core test.db test.tdb test.gdbm
diff --git a/source4/tdb/README b/source4/tdb/README
new file mode 100644
index 0000000000..fac3eacb4d
--- /dev/null
+++ b/source4/tdb/README
@@ -0,0 +1,167 @@
+tdb - a trivial database system
+tridge@linuxcare.com December 1999
+==================================
+
+This is a simple database API. It was inspired by the realisation that
+in Samba we have several ad-hoc bits of code that essentially
+implement small databases for sharing structures between parts of
+Samba. As I was about to add another I realised that a generic
+database module was called for to replace all the ad-hoc bits.
+
+I based the interface on gdbm. I couldn't use gdbm as we need to be
+able to have multiple writers to the databases at one time.
+
+Compilation
+-----------
+
+add HAVE_MMAP=1 to use mmap instead of read/write
+add TDB_DEBUG=1 for verbose debug info
+add NOLOCK=1 to disable locking code
+
+Testing
+-------
+
+Compile tdbtest.c and link with gdbm for testing. tdbtest will perform
+identical operations via tdb and gdbm then make sure the result is the
+same
+
+Also included is tdbtool, which allows simple database manipulation
+on the commandline.
+
+tdbtest and tdbtool are not built as part of Samba, but are included
+for completeness.
+
+Interface
+---------
+
+The interface is very similar to gdbm except for the following:
+
+- different open interface. The tdb_open call is more similar to a
+  traditional open()
+- no tdbm_reorganise() function
+- no tdbm_sync() function. No operations are cached in the library anyway
+- added a tdb_traverse() function for traversing the whole database
+
+A general rule for using tdb is that the caller frees any returned
+TDB_DATA structures. Just call free(p.dptr) to free a TDB_DATA
+return value called p. This is the same as gdbm.
+
+here is a full list of tdb functions with brief descriptions.
+
+
+----------------------------------------------------------------------
+TDB_CONTEXT *tdb_open(char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode)
+
+   open the database, creating it if necessary 
+
+   The open_flags and mode are passed straight to the open call on the database
+   file. A flags value of O_WRONLY is invalid
+
+   The hash size is advisory, use zero for a default value. 
+
+   return is NULL on error
+
+   possible tdb_flags are:
+    TDB_CLEAR_IF_FIRST - clear database if we are the only one with it open
+    TDB_INTERNAL - don't use a file, instaed store the data in
+                   memory. The filename is ignored in this case.
+    TDB_NOLOCK - don't do any locking
+    TDB_NOMMAP - don't use mmap
+
+----------------------------------------------------------------------
+char *tdb_error(TDB_CONTEXT *tdb);
+
+     return a error string for the last tdb error
+
+----------------------------------------------------------------------
+int tdb_close(TDB_CONTEXT *tdb);
+
+   close a database
+
+----------------------------------------------------------------------
+int tdb_update(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf);
+
+   update an entry in place - this only works if the new data size
+   is <= the old data size and the key exists.
+   on failure return -1
+
+----------------------------------------------------------------------
+TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   fetch an entry in the database given a key 
+   if the return value has a null dptr then a error occurred
+
+   caller must free the resulting data
+
+----------------------------------------------------------------------
+int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   check if an entry in the database exists 
+
+   note that 1 is returned if the key is found and 0 is returned if not found
+   this doesn't match the conventions in the rest of this module, but is
+   compatible with gdbm
+
+----------------------------------------------------------------------
+int tdb_traverse(TDB_CONTEXT *tdb, int (*fn)(TDB_CONTEXT *tdb,
+                 TDB_DATA key, TDB_DATA dbuf, void *state), void *state);
+
+   traverse the entire database - calling fn(tdb, key, data, state) on each 
+   element.
+
+   return -1 on error or the record count traversed
+
+   if fn is NULL then it is not called
+
+   a non-zero return value from fn() indicates that the traversal should stop
+
+----------------------------------------------------------------------
+TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb);
+
+   find the first entry in the database and return its key
+
+   the caller must free the returned data
+
+----------------------------------------------------------------------
+TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   find the next entry in the database, returning its key
+
+   the caller must free the returned data
+
+----------------------------------------------------------------------
+int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   delete an entry in the database given a key
+
+----------------------------------------------------------------------
+int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag);
+
+   store an element in the database, replacing any existing element
+   with the same key 
+
+   If flag==TDB_INSERT then don't overwrite an existing entry
+   If flag==TDB_MODIFY then don't create a new entry
+
+   return 0 on success, -1 on failure
+
+----------------------------------------------------------------------
+int tdb_writelock(TDB_CONTEXT *tdb);
+
+   lock the database. If we already have it locked then don't do anything
+
+----------------------------------------------------------------------
+int tdb_writeunlock(TDB_CONTEXT *tdb);
+   unlock the database
+
+----------------------------------------------------------------------
+int tdb_lockchain(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   lock one hash chain. This is meant to be used to reduce locking
+   contention - it cannot guarantee how many records will be locked
+
+----------------------------------------------------------------------
+int tdb_unlockchain(TDB_CONTEXT *tdb, TDB_DATA key);
+
+   unlock one hash chain
diff --git a/source4/tdb/spinlock.c b/source4/tdb/spinlock.c
new file mode 100644
index 0000000000..2370ce3bdd
--- /dev/null
+++ b/source4/tdb/spinlock.c
@@ -0,0 +1,430 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba database functions
+   Copyright (C) Anton Blanchard                   2001
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#if STANDALONE
+#include <stdlib.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <time.h>
+#include <signal.h>
+#include "tdb.h"
+#include "spinlock.h"
+
+#define DEBUG
+#else
+#include "includes.h"
+#endif
+
+#ifdef USE_SPINLOCKS
+
+/*
+ * ARCH SPECIFIC
+ */
+
+#if defined(SPARC_SPINLOCKS)
+
+static inline int __spin_trylock(spinlock_t *lock)
+{
+	unsigned int result;
+
+	asm volatile("ldstub    [%1], %0"
+		: "=r" (result)
+		: "r" (lock)
+		: "memory");
+
+	return (result == 0) ? 0 : EBUSY;
+}
+
+static inline void __spin_unlock(spinlock_t *lock)
+{
+	asm volatile("":::"memory");
+	*lock = 0;
+}
+
+static inline void __spin_lock_init(spinlock_t *lock)
+{
+	*lock = 0;
+}
+
+static inline int __spin_is_locked(spinlock_t *lock)
+{
+	return (*lock != 0);
+}
+
+#elif defined(POWERPC_SPINLOCKS) 
+
+static inline int __spin_trylock(spinlock_t *lock)
+{
+	unsigned int result;
+
+	__asm__ __volatile__(
+"1:	lwarx		%0,0,%1\n\
+	cmpwi		0,%0,0\n\
+	li		%0,0\n\
+	bne-		2f\n\
+	li		%0,1\n\
+	stwcx.		%0,0,%1\n\
+	bne-		1b\n\
+	isync\n\
+2:"	: "=&r"(result)
+	: "r"(lock)
+	: "cr0", "memory");
+
+	return (result == 1) ? 0 : EBUSY;
+}
+
+static inline void __spin_unlock(spinlock_t *lock)
+{
+	asm volatile("eieio":::"memory");
+	*lock = 0;
+}
+
+static inline void __spin_lock_init(spinlock_t *lock)
+{
+	*lock = 0;
+}
+
+static inline int __spin_is_locked(spinlock_t *lock)
+{
+	return (*lock != 0);
+}
+
+#elif defined(INTEL_SPINLOCKS) 
+
+static inline int __spin_trylock(spinlock_t *lock)
+{
+	int oldval;
+
+	asm volatile("xchgl %0,%1"
+		: "=r" (oldval), "=m" (*lock)
+		: "0" (0)
+		: "memory");
+
+	return oldval > 0 ? 0 : EBUSY;
+}
+
+static inline void __spin_unlock(spinlock_t *lock)
+{
+	asm volatile("":::"memory");
+	*lock = 1;
+}
+
+static inline void __spin_lock_init(spinlock_t *lock)
+{
+	*lock = 1;
+}
+
+static inline int __spin_is_locked(spinlock_t *lock)
+{
+	return (*lock != 1);
+}
+
+#elif defined(MIPS_SPINLOCKS) 
+
+static inline unsigned int load_linked(unsigned long addr)
+{
+	unsigned int res;
+
+	__asm__ __volatile__("ll\t%0,(%1)"
+		: "=r" (res)
+		: "r" (addr));
+
+	return res;
+}
+
+static inline unsigned int store_conditional(unsigned long addr, unsigned int value)
+{
+	unsigned int res;
+
+	__asm__ __volatile__("sc\t%0,(%2)"
+		: "=r" (res)
+		: "0" (value), "r" (addr));
+	return res;
+}
+
+static inline int __spin_trylock(spinlock_t *lock)
+{
+	unsigned int mw;
+
+	do {
+		mw = load_linked(lock);
+		if (mw) 
+			return EBUSY;
+	} while (!store_conditional(lock, 1));
+
+	asm volatile("":::"memory");
+
+	return 0;
+}
+
+static inline void __spin_unlock(spinlock_t *lock)
+{
+	asm volatile("":::"memory");
+	*lock = 0;
+}
+
+static inline void __spin_lock_init(spinlock_t *lock)
+{
+	*lock = 0;
+}
+
+static inline int __spin_is_locked(spinlock_t *lock)
+{
+	return (*lock != 0);
+}
+
+#else
+#error Need to implement spinlock code in spinlock.c
+#endif
+
+/*
+ * OS SPECIFIC
+ */
+
+static void yield_cpu(void)
+{
+	struct timespec tm;
+
+#ifdef USE_SCHED_YIELD
+	sched_yield();
+#else
+	/* Linux will busy loop for delays < 2ms on real time tasks */
+	tm.tv_sec = 0;
+	tm.tv_nsec = 2000000L + 1;
+	nanosleep(&tm, NULL);
+#endif
+}
+
+static int this_is_smp(void)
+{
+	return 0;
+}
+
+/*
+ * GENERIC
+ */
+
+static int smp_machine = 0;
+
+static inline void __spin_lock(spinlock_t *lock)
+{
+	int ntries = 0;
+
+	while(__spin_trylock(lock)) {
+		while(__spin_is_locked(lock)) {
+			if (smp_machine && ntries++ < MAX_BUSY_LOOPS)
+				continue;
+			yield_cpu();
+		}
+	}
+}
+
+static void __read_lock(tdb_rwlock_t *rwlock)
+{
+	int ntries = 0;
+
+	while(1) {
+		__spin_lock(&rwlock->lock);
+
+		if (!(rwlock->count & RWLOCK_BIAS)) {
+			rwlock->count++;
+			__spin_unlock(&rwlock->lock);
+			return;
+		}
+	
+		__spin_unlock(&rwlock->lock);
+
+		while(rwlock->count & RWLOCK_BIAS) {
+			if (smp_machine && ntries++ < MAX_BUSY_LOOPS)
+				continue;
+			yield_cpu();
+		}
+	}
+}
+
+static void __write_lock(tdb_rwlock_t *rwlock)
+{
+	int ntries = 0;
+
+	while(1) {
+		__spin_lock(&rwlock->lock);
+
+		if (rwlock->count == 0) {
+			rwlock->count |= RWLOCK_BIAS;
+			__spin_unlock(&rwlock->lock);
+			return;
+		}
+
+		__spin_unlock(&rwlock->lock);
+
+		while(rwlock->count != 0) {
+			if (smp_machine && ntries++ < MAX_BUSY_LOOPS)
+				continue;
+			yield_cpu();
+		}
+	}
+}
+
+static void __write_unlock(tdb_rwlock_t *rwlock)
+{
+	__spin_lock(&rwlock->lock);
+
+#ifdef DEBUG
+	if (!(rwlock->count & RWLOCK_BIAS))
+		fprintf(stderr, "bug: write_unlock\n");
+#endif
+
+	rwlock->count &= ~RWLOCK_BIAS;
+	__spin_unlock(&rwlock->lock);
+}
+
+static void __read_unlock(tdb_rwlock_t *rwlock)
+{
+	__spin_lock(&rwlock->lock);
+
+#ifdef DEBUG
+	if (!rwlock->count)
+		fprintf(stderr, "bug: read_unlock\n");
+
+	if (rwlock->count & RWLOCK_BIAS)
+		fprintf(stderr, "bug: read_unlock\n");
+#endif
+
+	rwlock->count--;
+	__spin_unlock(&rwlock->lock);
+}
+
+/* TDB SPECIFIC */
+
+/* lock a list in the database. list -1 is the alloc list */
+int tdb_spinlock(TDB_CONTEXT *tdb, int list, int rw_type)
+{
+	tdb_rwlock_t *rwlocks;
+
+	if (!tdb->map_ptr) return -1;
+	rwlocks = (tdb_rwlock_t *)((char *)tdb->map_ptr + tdb->header.rwlocks);
+
+	switch(rw_type) {
+	case F_RDLCK:
+		__read_lock(&rwlocks[list+1]);
+		break;
+
+	case F_WRLCK:
+		__write_lock(&rwlocks[list+1]);
+		break;
+
+	default:
+		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
+	}
+	return 0;
+}
+
+/* unlock the database. */
+int tdb_spinunlock(TDB_CONTEXT *tdb, int list, int rw_type)
+{
+	tdb_rwlock_t *rwlocks;
+
+	if (!tdb->map_ptr) return -1;
+	rwlocks = (tdb_rwlock_t *)((char *)tdb->map_ptr + tdb->header.rwlocks);
+
+	switch(rw_type) {
+	case F_RDLCK:
+		__read_unlock(&rwlocks[list+1]);
+		break;
+
+	case F_WRLCK:
+		__write_unlock(&rwlocks[list+1]);
+		break;
+
+	default:
+		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
+	}
+
+	return 0;
+}
+
+int tdb_create_rwlocks(int fd, unsigned int hash_size)
+{
+	unsigned size, i;
+	tdb_rwlock_t *rwlocks;
+
+	size = (hash_size + 1) * sizeof(tdb_rwlock_t);
+	rwlocks = malloc(size);
+	if (!rwlocks)
+		return -1;
+
+	for(i = 0; i < hash_size+1; i++) {
+		__spin_lock_init(&rwlocks[i].lock);
+		rwlocks[i].count = 0;
+	}
+
+	/* Write it out (appending to end) */
+	if (write(fd, rwlocks, size) != size) {
+		free(rwlocks);
+		return -1;
+	}
+	smp_machine = this_is_smp();
+	free(rwlocks);
+	return 0;
+}
+
+int tdb_clear_spinlocks(TDB_CONTEXT *tdb)
+{
+	tdb_rwlock_t *rwlocks;
+	unsigned i;
+
+	if (tdb->header.rwlocks == 0) return 0;
+	if (!tdb->map_ptr) return -1;
+
+	/* We're mmapped here */
+	rwlocks = (tdb_rwlock_t *)((char *)tdb->map_ptr + tdb->header.rwlocks);
+	for(i = 0; i < tdb->header.hash_size+1; i++) {
+		__spin_lock_init(&rwlocks[i].lock);
+		rwlocks[i].count = 0;
+	}
+	return 0;
+}
+#else
+int tdb_create_rwlocks(int fd, unsigned int hash_size) { return 0; }
+int tdb_spinlock(TDB_CONTEXT *tdb, int list, int rw_type) { return -1; }
+int tdb_spinunlock(TDB_CONTEXT *tdb, int list, int rw_type) { return -1; }
+
+/* Non-spinlock version: remove spinlock pointer */
+int tdb_clear_spinlocks(TDB_CONTEXT *tdb)
+{
+	tdb_off off = (tdb_off)((char *)&tdb->header.rwlocks
+				- (char *)&tdb->header);
+
+	tdb->header.rwlocks = 0;
+	if (lseek(tdb->fd, off, SEEK_SET) != off
+	    || write(tdb->fd, (void *)&tdb->header.rwlocks,
+		     sizeof(tdb->header.rwlocks)) 
+	    != sizeof(tdb->header.rwlocks))
+		return -1;
+	return 0;
+}
+#endif
diff --git a/source4/tdb/spinlock.h b/source4/tdb/spinlock.h
new file mode 100644
index 0000000000..d6a2ac6eb8
--- /dev/null
+++ b/source4/tdb/spinlock.h
@@ -0,0 +1,55 @@
+#ifndef __SPINLOCK_H__
+#define __SPINLOCK_H__
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "tdb.h"
+
+#ifdef USE_SPINLOCKS
+
+#define RWLOCK_BIAS 0x1000UL
+
+/* OS SPECIFIC */
+#define MAX_BUSY_LOOPS 1000
+#undef USE_SCHED_YIELD
+
+/* ARCH SPECIFIC */
+/* We should make sure these are padded to a cache line */
+#if defined(SPARC_SPINLOCKS)
+typedef volatile char spinlock_t;
+#elif defined(POWERPC_SPINLOCKS)
+typedef volatile unsigned long spinlock_t;
+#elif defined(INTEL_SPINLOCKS)
+typedef volatile int spinlock_t;
+#elif defined(MIPS_SPINLOCKS)
+typedef volatile unsigned long spinlock_t;
+#else
+#error Need to implement spinlock code in spinlock.h
+#endif
+
+typedef struct {
+	spinlock_t lock;
+	volatile int count;
+} tdb_rwlock_t;
+
+int tdb_spinlock(TDB_CONTEXT *tdb, int list, int rw_type);
+int tdb_spinunlock(TDB_CONTEXT *tdb, int list, int rw_type);
+int tdb_create_rwlocks(int fd, unsigned int hash_size);
+int tdb_clear_spinlocks(TDB_CONTEXT *tdb);
+
+#else /* !USE_SPINLOCKS */
+#if 0
+#define tdb_create_rwlocks(fd, hash_size) 0
+#define tdb_spinlock(tdb, list, rw_type) (-1)
+#define tdb_spinunlock(tdb, list, rw_type) (-1)
+#else
+int tdb_spinlock(TDB_CONTEXT *tdb, int list, int rw_type);
+int tdb_spinunlock(TDB_CONTEXT *tdb, int list, int rw_type);
+int tdb_create_rwlocks(int fd, unsigned int hash_size);
+#endif
+int tdb_clear_spinlocks(TDB_CONTEXT *tdb);
+#endif
+
+#endif
diff --git a/source4/tdb/tdb.c b/source4/tdb/tdb.c
new file mode 100644
index 0000000000..097209ff7a
--- /dev/null
+++ b/source4/tdb/tdb.c
@@ -0,0 +1,2020 @@
+ /* 
+   Unix SMB/CIFS implementation.
+   Samba database functions
+   Copyright (C) Andrew Tridgell              1999-2000
+   Copyright (C) Luke Kenneth Casson Leighton      2000
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+#ifdef STANDALONE
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <signal.h>
+#include "tdb.h"
+#include "spinlock.h"
+#else
+#include "includes.h"
+#endif
+
+#define TDB_MAGIC_FOOD "TDB file\n"
+#define TDB_VERSION (0x26011967 + 6)
+#define TDB_MAGIC (0x26011999U)
+#define TDB_FREE_MAGIC (~TDB_MAGIC)
+#define TDB_DEAD_MAGIC (0xFEE1DEAD)
+#define TDB_ALIGNMENT 4
+#define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
+#define DEFAULT_HASH_SIZE 131
+#define TDB_PAGE_SIZE 0x2000
+#define FREELIST_TOP (sizeof(struct tdb_header))
+#define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
+#define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
+#define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
+#define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
+#define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
+
+/* NB assumes there is a local variable called "tdb" that is the
+ * current context, also takes doubly-parenthesized print-style
+ * argument. */
+#define TDB_LOG(x) (tdb->log_fn?((tdb->log_fn x),0) : 0)
+
+/* lock offsets */
+#define GLOBAL_LOCK 0
+#define ACTIVE_LOCK 4
+
+#ifndef MAP_FILE
+#define MAP_FILE 0
+#endif
+
+#ifndef MAP_FAILED
+#define MAP_FAILED ((void *)-1)
+#endif
+
+/* free memory if the pointer is valid and zero the pointer */
+#ifndef SAFE_FREE
+#define SAFE_FREE(x) do { if ((x) != NULL) {free((x)); (x)=NULL;} } while(0)
+#endif
+
+#define BUCKET(hash) ((hash) % tdb->header.hash_size)
+TDB_DATA tdb_null;
+
+/* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
+static TDB_CONTEXT *tdbs = NULL;
+
+static int tdb_munmap(TDB_CONTEXT *tdb)
+{
+	if (tdb->flags & TDB_INTERNAL)
+		return 0;
+
+#ifdef HAVE_MMAP
+	if (tdb->map_ptr) {
+		int ret = munmap(tdb->map_ptr, tdb->map_size);
+		if (ret != 0)
+			return ret;
+	}
+#endif
+	tdb->map_ptr = NULL;
+	return 0;
+}
+
+static void tdb_mmap(TDB_CONTEXT *tdb)
+{
+	if (tdb->flags & TDB_INTERNAL)
+		return;
+
+#ifdef HAVE_MMAP
+	if (!(tdb->flags & TDB_NOMMAP)) {
+		tdb->map_ptr = mmap(NULL, tdb->map_size, 
+				    PROT_READ|(tdb->read_only? 0:PROT_WRITE), 
+				    MAP_SHARED|MAP_FILE, tdb->fd, 0);
+
+		/*
+		 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
+		 */
+
+		if (tdb->map_ptr == MAP_FAILED) {
+			tdb->map_ptr = NULL;
+			TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n", 
+				 tdb->map_size, strerror(errno)));
+		}
+	} else {
+		tdb->map_ptr = NULL;
+	}
+#else
+	tdb->map_ptr = NULL;
+#endif
+}
+
+/* Endian conversion: we only ever deal with 4 byte quantities */
+static void *convert(void *buf, u32 size)
+{
+	u32 i, *p = buf;
+	for (i = 0; i < size / 4; i++)
+		p[i] = TDB_BYTEREV(p[i]);
+	return buf;
+}
+#define DOCONV() (tdb->flags & TDB_CONVERT)
+#define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)
+
+/* the body of the database is made of one list_struct for the free space
+   plus a separate data list for each hash value */
+struct list_struct {
+	tdb_off next; /* offset of the next record in the list */
+	tdb_len rec_len; /* total byte length of record */
+	tdb_len key_len; /* byte length of key */
+	tdb_len data_len; /* byte length of data */
+	u32 full_hash; /* the full 32 bit hash of the key */
+	u32 magic;   /* try to catch errors */
+	/* the following union is implied:
+		union {
+			char record[rec_len];
+			struct {
+				char key[key_len];
+				char data[data_len];
+			}
+			u32 totalsize; (tailer)
+		}
+	*/
+};
+
+/***************************************************************
+ Allow a caller to set a "alarm" flag that tdb can check to abort
+ a blocking lock on SIGALRM.
+***************************************************************/
+
+static sig_atomic_t *palarm_fired;
+
+void tdb_set_lock_alarm(sig_atomic_t *palarm)
+{
+	palarm_fired = palarm;
+}
+
+/* a byte range locking function - return 0 on success
+   this functions locks/unlocks 1 byte at the specified offset.
+
+   On error, errno is also set so that errors are passed back properly
+   through tdb_open(). */
+static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset, 
+		      int rw_type, int lck_type, int probe)
+{
+	struct flock fl;
+	int ret;
+
+	if (tdb->flags & TDB_NOLOCK)
+		return 0;
+	if ((rw_type == F_WRLCK) && (tdb->read_only)) {
+		errno = EACCES;
+		return -1;
+	}
+
+	fl.l_type = rw_type;
+	fl.l_whence = SEEK_SET;
+	fl.l_start = offset;
+	fl.l_len = 1;
+	fl.l_pid = 0;
+
+	do {
+		ret = fcntl(tdb->fd,lck_type,&fl);
+		if (ret == -1 && errno == EINTR && palarm_fired && *palarm_fired)
+			break;
+	} while (ret == -1 && errno == EINTR);
+
+	if (ret == -1) {
+		if (!probe && lck_type != F_SETLK) {
+			/* Ensure error code is set for log fun to examine. */
+			if (errno == EINTR && palarm_fired && *palarm_fired)
+				tdb->ecode = TDB_ERR_LOCK_TIMEOUT;
+			else
+				tdb->ecode = TDB_ERR_LOCK;
+			TDB_LOG((tdb, 5,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d\n", 
+				 tdb->fd, offset, rw_type, lck_type));
+		}
+		/* Was it an alarm timeout ? */
+		if (errno == EINTR && palarm_fired && *palarm_fired)
+			return TDB_ERRCODE(TDB_ERR_LOCK_TIMEOUT, -1);
+		/* Otherwise - generic lock error. */
+		/* errno set by fcntl */
+		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
+	}
+	return 0;
+}
+
+/* lock a list in the database. list -1 is the alloc list */
+static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
+{
+	if (list < -1 || list >= (int)tdb->header.hash_size) {
+		TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n", 
+			   list, ltype));
+		return -1;
+	}
+	if (tdb->flags & TDB_NOLOCK)
+		return 0;
+
+	/* Since fcntl locks don't nest, we do a lock for the first one,
+	   and simply bump the count for future ones */
+	if (tdb->locked[list+1].count == 0) {
+		if (!tdb->read_only && tdb->header.rwlocks) {
+			if (tdb_spinlock(tdb, list, ltype)) {
+				TDB_LOG((tdb, 0, "tdb_lock spinlock failed on list ltype=%d\n", 
+					   list, ltype));
+				return -1;
+			}
+		} else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
+			TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n", 
+					   list, ltype, strerror(errno)));
+			return -1;
+		}
+		tdb->locked[list+1].ltype = ltype;
+	}
+	tdb->locked[list+1].count++;
+	return 0;
+}
+
+/* unlock the database: returns void because it's too late for errors. */
+	/* changed to return int it may be interesting to know there
+	   has been an error  --simo */
+static int tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
+{
+	int ret = -1;
+
+	if (tdb->flags & TDB_NOLOCK)
+		return 0;
+
+	/* Sanity checks */
+	if (list < -1 || list >= (int)tdb->header.hash_size) {
+		TDB_LOG((tdb, 0, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
+		return ret;
+	}
+
+	if (tdb->locked[list+1].count==0) {
+		TDB_LOG((tdb, 0, "tdb_unlock: count is 0\n"));
+		return ret;
+	}
+
+	if (tdb->locked[list+1].count == 1) {
+		/* Down to last nested lock: unlock underneath */
+		if (!tdb->read_only && tdb->header.rwlocks) {
+			ret = tdb_spinunlock(tdb, list, ltype);
+		} else {
+			ret = tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
+		}
+	} else {
+		ret = 0;
+	}
+	tdb->locked[list+1].count--;
+
+	if (ret)
+		TDB_LOG((tdb, 0,"tdb_unlock: An error occurred unlocking!\n")); 
+	return ret;
+}
+
+/* This is based on the hash algorithm from gdbm */
+static u32 tdb_hash(TDB_DATA *key)
+{
+	u32 value;	/* Used to compute the hash value.  */
+	u32   i;	/* Used to cycle through random values. */
+
+	/* Set the initial value from the key size. */
+	for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
+		value = (value + (key->dptr[i] << (i*5 % 24)));
+
+	return (1103515243 * value + 12345);  
+}
+
+/* check for an out of bounds access - if it is out of bounds then
+   see if the database has been expanded by someone else and expand
+   if necessary 
+   note that "len" is the minimum length needed for the db
+*/
+static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
+{
+	struct stat st;
+	if (len <= tdb->map_size)
+		return 0;
+	if (tdb->flags & TDB_INTERNAL) {
+		if (!probe) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, 0,"tdb_oob len %d beyond internal malloc size %d\n",
+				 (int)len, (int)tdb->map_size));
+		}
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+	}
+
+	if (fstat(tdb->fd, &st) == -1)
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+
+	if (st.st_size < (size_t)len) {
+		if (!probe) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
+				 (int)len, (int)st.st_size));
+		}
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+	}
+
+	/* Unmap, update size, remap */
+	if (tdb_munmap(tdb) == -1)
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+	tdb->map_size = st.st_size;
+	tdb_mmap(tdb);
+	return 0;
+}
+
+/* write a lump of data at a specified offset */
+static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
+{
+	if (tdb_oob(tdb, off + len, 0) != 0)
+		return -1;
+
+	if (tdb->map_ptr)
+		memcpy(off + (char *)tdb->map_ptr, buf, len);
+#ifdef HAVE_PWRITE
+	else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
+#else
+	else if (lseek(tdb->fd, off, SEEK_SET) != off
+		 || write(tdb->fd, buf, len) != (ssize_t)len) {
+#endif
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_IO;
+		TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
+			   off, len, strerror(errno)));
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+	}
+	return 0;
+}
+
+/* read a lump of data at a specified offset, maybe convert */
+static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
+{
+	if (tdb_oob(tdb, off + len, 0) != 0)
+		return -1;
+
+	if (tdb->map_ptr)
+		memcpy(buf, off + (char *)tdb->map_ptr, len);
+#ifdef HAVE_PREAD
+	else if (pread(tdb->fd, buf, len, off) != (ssize_t)len) {
+#else
+	else if (lseek(tdb->fd, off, SEEK_SET) != off
+		 || read(tdb->fd, buf, len) != (ssize_t)len) {
+#endif
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_IO;
+		TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
+			   off, len, strerror(errno)));
+		return TDB_ERRCODE(TDB_ERR_IO, -1);
+	}
+	if (cv)
+		convert(buf, len);
+	return 0;
+}
+
+/* read a lump of data, allocating the space for it */
+static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
+{
+	char *buf;
+
+	if (!(buf = malloc(len))) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_OOM;
+		TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
+			   len, strerror(errno)));
+		return TDB_ERRCODE(TDB_ERR_OOM, buf);
+	}
+	if (tdb_read(tdb, offset, buf, len, 0) == -1) {
+		SAFE_FREE(buf);
+		return NULL;
+	}
+	return buf;
+}
+
+/* read/write a tdb_off */
+static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
+{
+	return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
+}
+static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
+{
+	tdb_off off = *d;
+	return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
+}
+
+/* read/write a record */
+static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
+{
+	if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
+		return -1;
+	if (TDB_BAD_MAGIC(rec)) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_CORRUPT;
+		TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
+		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
+	}
+	return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
+}
+static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
+{
+	struct list_struct r = *rec;
+	return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
+}
+
+/* read a freelist record and check for simple errors */
+static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
+{
+	if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
+		return -1;
+
+	if (rec->magic == TDB_MAGIC) {
+		/* this happens when a app is showdown while deleting a record - we should
+		   not completely fail when this happens */
+		TDB_LOG((tdb, 0,"rec_free_read non-free magic at offset=%d - fixing\n", 
+			 rec->magic, off));
+		rec->magic = TDB_FREE_MAGIC;
+		if (tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
+			return -1;
+	}
+
+	if (rec->magic != TDB_FREE_MAGIC) {
+		/* Ensure ecode is set for log fn. */
+		tdb->ecode = TDB_ERR_CORRUPT;
+		TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n", 
+			   rec->magic, off));
+		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
+	}
+	if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
+		return -1;
+	return 0;
+}
+
+/* update a record tailer (must hold allocation lock) */
+static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
+			 const struct list_struct *rec)
+{
+	tdb_off totalsize;
+
+	/* Offset of tailer from record header */
+	totalsize = sizeof(*rec) + rec->rec_len;
+	return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
+			 &totalsize);
+}
+
+static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
+{
+	struct list_struct rec;
+	tdb_off tailer_ofs, tailer;
+
+	if (tdb_read(tdb, offset, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
+		printf("ERROR: failed to read record at %u\n", offset);
+		return 0;
+	}
+
+	printf(" rec: offset=%u next=%d rec_len=%d key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
+	       offset, rec.next, rec.rec_len, rec.key_len, rec.data_len, rec.full_hash, rec.magic);
+
+	tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off);
+	if (ofs_read(tdb, tailer_ofs, &tailer) == -1) {
+		printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
+		return rec.next;
+	}
+
+	if (tailer != rec.rec_len + sizeof(rec)) {
+		printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
+				(unsigned)tailer, (unsigned)(rec.rec_len + sizeof(rec)));
+	}
+	return rec.next;
+}
+
+static int tdb_dump_chain(TDB_CONTEXT *tdb, int i)
+{
+	tdb_off rec_ptr, top;
+
+	top = TDB_HASH_TOP(i);
+
+	if (tdb_lock(tdb, i, F_WRLCK) != 0)
+		return -1;
+
+	if (ofs_read(tdb, top, &rec_ptr) == -1)
+		return tdb_unlock(tdb, i, F_WRLCK);
+
+	if (rec_ptr)
+		printf("hash=%d\n", i);
+
+	while (rec_ptr) {
+		rec_ptr = tdb_dump_record(tdb, rec_ptr);
+	}
+
+	return tdb_unlock(tdb, i, F_WRLCK);
+}
+
+void tdb_dump_all(TDB_CONTEXT *tdb)
+{
+	int i;
+	for (i=0;i<tdb->header.hash_size;i++) {
+		tdb_dump_chain(tdb, i);
+	}
+	printf("freelist:\n");
+	tdb_dump_chain(tdb, -1);
+}
+
+int tdb_printfreelist(TDB_CONTEXT *tdb)
+{
+	int ret;
+	long total_free = 0;
+	tdb_off offset, rec_ptr;
+	struct list_struct rec;
+
+	if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
+		return ret;
+
+	offset = FREELIST_TOP;
+
+	/* read in the freelist top */
+	if (ofs_read(tdb, offset, &rec_ptr) == -1) {
+		tdb_unlock(tdb, -1, F_WRLCK);
+		return 0;
+	}
+
+	printf("freelist top=[0x%08x]\n", rec_ptr );
+	while (rec_ptr) {
+		if (tdb_read(tdb, rec_ptr, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return -1;
+		}
+
+		if (rec.magic != TDB_FREE_MAGIC) {
+			printf("bad magic 0x%08x in free list\n", rec.magic);
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return -1;
+		}
+
+		printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)]\n", rec.next, rec.rec_len, rec.rec_len );
+		total_free += rec.rec_len;
+
+		/* move to the next record */
+		rec_ptr = rec.next;
+	}
+	printf("total rec_len = [0x%08x (%d)]\n", (int)total_free, 
+               (int)total_free);
+
+	return tdb_unlock(tdb, -1, F_WRLCK);
+}
+
+/* Remove an element from the freelist.  Must have alloc lock. */
+static int remove_from_freelist(TDB_CONTEXT *tdb, tdb_off off, tdb_off next)
+{
+	tdb_off last_ptr, i;
+
+	/* read in the freelist top */
+	last_ptr = FREELIST_TOP;
+	while (ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
+		if (i == off) {
+			/* We've found it! */
+			return ofs_write(tdb, last_ptr, &next);
+		}
+		/* Follow chain (next offset is at start of record) */
+		last_ptr = i;
+	}
+	TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
+	return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
+}
+
+/* Add an element into the freelist. Merge adjacent records if
+   neccessary. */
+static int tdb_free(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
+{
+	tdb_off right, left;
+
+	/* Allocation and tailer lock */
+	if (tdb_lock(tdb, -1, F_WRLCK) != 0)
+		return -1;
+
+	/* set an initial tailer, so if we fail we don't leave a bogus record */
+	if (update_tailer(tdb, offset, rec) != 0) {
+		TDB_LOG((tdb, 0, "tdb_free: upfate_tailer failed!\n"));
+		goto fail;
+	}
+
+	/* Look right first (I'm an Australian, dammit) */
+	right = offset + sizeof(*rec) + rec->rec_len;
+	if (right + sizeof(*rec) <= tdb->map_size) {
+		struct list_struct r;
+
+		if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
+			TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
+			goto left;
+		}
+
+		/* If it's free, expand to include it. */
+		if (r.magic == TDB_FREE_MAGIC) {
+			if (remove_from_freelist(tdb, right, r.next) == -1) {
+				TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
+				goto left;
+			}
+			rec->rec_len += sizeof(r) + r.rec_len;
+		}
+	}
+
+left:
+	/* Look left */
+	left = offset - sizeof(tdb_off);
+	if (left > TDB_HASH_TOP(tdb->header.hash_size-1)) {
+		struct list_struct l;
+		tdb_off leftsize;
+
+		/* Read in tailer and jump back to header */
+		if (ofs_read(tdb, left, &leftsize) == -1) {
+			TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
+			goto update;
+		}
+		left = offset - leftsize;
+
+		/* Now read in record */
+		if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
+			TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
+			goto update;
+		}
+
+		/* If it's free, expand to include it. */
+		if (l.magic == TDB_FREE_MAGIC) {
+			if (remove_from_freelist(tdb, left, l.next) == -1) {
+				TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
+				goto update;
+			} else {
+				offset = left;
+				rec->rec_len += leftsize;
+			}
+		}
+	}
+
+update:
+	if (update_tailer(tdb, offset, rec) == -1) {
+		TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
+		goto fail;
+	}
+
+	/* Now, prepend to free list */
+	rec->magic = TDB_FREE_MAGIC;
+
+	if (ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
+	    rec_write(tdb, offset, rec) == -1 ||
+	    ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
+		TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
+		goto fail;
+	}
+
+	/* And we're done. */
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return -1;
+}
+
+
+/* expand a file.  we prefer to use ftruncate, as that is what posix
+  says to use for mmap expansion */
+static int expand_file(TDB_CONTEXT *tdb, tdb_off size, tdb_off addition)
+{
+	char buf[1024];
+#if HAVE_FTRUNCATE_EXTEND
+	if (ftruncate(tdb->fd, size+addition) != 0) {
+		TDB_LOG((tdb, 0, "expand_file ftruncate to %d failed (%s)\n", 
+			   size+addition, strerror(errno)));
+		return -1;
+	}
+#else
+	char b = 0;
+
+#ifdef HAVE_PWRITE
+	if (pwrite(tdb->fd,  &b, 1, (size+addition) - 1) != 1) {
+#else
+	if (lseek(tdb->fd, (size+addition) - 1, SEEK_SET) != (size+addition) - 1 || 
+	    write(tdb->fd, &b, 1) != 1) {
+#endif
+		TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n", 
+			   size+addition, strerror(errno)));
+		return -1;
+	}
+#endif
+
+	/* now fill the file with something. This ensures that the file isn't sparse, which would be
+	   very bad if we ran out of disk. This must be done with write, not via mmap */
+	memset(buf, 0x42, sizeof(buf));
+	while (addition) {
+		int n = addition>sizeof(buf)?sizeof(buf):addition;
+#ifdef HAVE_PWRITE
+		int ret = pwrite(tdb->fd, buf, n, size);
+#else
+		int ret;
+		if (lseek(tdb->fd, size, SEEK_SET) != size)
+			return -1;
+		ret = write(tdb->fd, buf, n);
+#endif
+		if (ret != n) {
+			TDB_LOG((tdb, 0, "expand_file write of %d failed (%s)\n", 
+				   n, strerror(errno)));
+			return -1;
+		}
+		addition -= n;
+		size += n;
+	}
+	return 0;
+}
+
+
+/* expand the database at least size bytes by expanding the underlying
+   file and doing the mmap again if necessary */
+static int tdb_expand(TDB_CONTEXT *tdb, tdb_off size)
+{
+	struct list_struct rec;
+	tdb_off offset;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		TDB_LOG((tdb, 0, "lock failed in tdb_expand\n"));
+		return -1;
+	}
+
+	/* must know about any previous expansions by another process */
+	tdb_oob(tdb, tdb->map_size + 1, 1);
+
+	/* always make room for at least 10 more records, and round
+           the database up to a multiple of TDB_PAGE_SIZE */
+	size = TDB_ALIGN(tdb->map_size + size*10, TDB_PAGE_SIZE) - tdb->map_size;
+
+	if (!(tdb->flags & TDB_INTERNAL))
+		tdb_munmap(tdb);
+
+	/*
+	 * We must ensure the file is unmapped before doing this
+	 * to ensure consistency with systems like OpenBSD where
+	 * writes and mmaps are not consistent.
+	 */
+
+	/* expand the file itself */
+	if (!(tdb->flags & TDB_INTERNAL)) {
+		if (expand_file(tdb, tdb->map_size, size) != 0)
+			goto fail;
+	}
+
+	tdb->map_size += size;
+
+	if (tdb->flags & TDB_INTERNAL)
+		tdb->map_ptr = realloc(tdb->map_ptr, tdb->map_size);
+	else {
+		/*
+		 * We must ensure the file is remapped before adding the space
+		 * to ensure consistency with systems like OpenBSD where
+		 * writes and mmaps are not consistent.
+		 */
+
+		/* We're ok if the mmap fails as we'll fallback to read/write */
+		tdb_mmap(tdb);
+	}
+
+	/* form a new freelist record */
+	memset(&rec,'\0',sizeof(rec));
+	rec.rec_len = size - sizeof(rec);
+
+	/* link it into the free list */
+	offset = tdb->map_size - size;
+	if (tdb_free(tdb, offset, &rec) == -1)
+		goto fail;
+
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return -1;
+}
+
+/* allocate some space from the free list. The offset returned points
+   to a unconnected list_struct within the database with room for at
+   least length bytes of total data
+
+   0 is returned if the space could not be allocated
+ */
+static tdb_off tdb_allocate(TDB_CONTEXT *tdb, tdb_len length,
+			    struct list_struct *rec)
+{
+	tdb_off rec_ptr, last_ptr, newrec_ptr;
+	struct list_struct newrec;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1)
+		return 0;
+
+	/* Extra bytes required for tailer */
+	length += sizeof(tdb_off);
+
+ again:
+	last_ptr = FREELIST_TOP;
+
+	/* read in the freelist top */
+	if (ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
+		goto fail;
+
+	/* keep looking until we find a freelist record big enough */
+	while (rec_ptr) {
+		if (rec_free_read(tdb, rec_ptr, rec) == -1)
+			goto fail;
+
+		if (rec->rec_len >= length) {
+			/* found it - now possibly split it up  */
+			if (rec->rec_len > length + MIN_REC_SIZE) {
+				/* Length of left piece */
+				length = TDB_ALIGN(length, TDB_ALIGNMENT);
+
+				/* Right piece to go on free list */
+				newrec.rec_len = rec->rec_len
+					- (sizeof(*rec) + length);
+				newrec_ptr = rec_ptr + sizeof(*rec) + length;
+
+				/* And left record is shortened */
+				rec->rec_len = length;
+			} else
+				newrec_ptr = 0;
+
+			/* Remove allocated record from the free list */
+			if (ofs_write(tdb, last_ptr, &rec->next) == -1)
+				goto fail;
+
+			/* Update header: do this before we drop alloc
+                           lock, otherwise tdb_free() might try to
+                           merge with us, thinking we're free.
+                           (Thanks Jeremy Allison). */
+			rec->magic = TDB_MAGIC;
+			if (rec_write(tdb, rec_ptr, rec) == -1)
+				goto fail;
+
+			/* Did we create new block? */
+			if (newrec_ptr) {
+				/* Update allocated record tailer (we
+                                   shortened it). */
+				if (update_tailer(tdb, rec_ptr, rec) == -1)
+					goto fail;
+
+				/* Free new record */
+				if (tdb_free(tdb, newrec_ptr, &newrec) == -1)
+					goto fail;
+			}
+
+			/* all done - return the new record offset */
+			tdb_unlock(tdb, -1, F_WRLCK);
+			return rec_ptr;
+		}
+		/* move to the next record */
+		last_ptr = rec_ptr;
+		rec_ptr = rec->next;
+	}
+	/* we didn't find enough space. See if we can expand the
+	   database and if we can then try again */
+	if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
+		goto again;
+ fail:
+	tdb_unlock(tdb, -1, F_WRLCK);
+	return 0;
+}
+
+/* initialise a new database with a specified hash size */
+static int tdb_new_database(TDB_CONTEXT *tdb, int hash_size)
+{
+	struct tdb_header *newdb;
+	int size, ret = -1;
+
+	/* We make it up in memory, then write it out if not internal */
+	size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off);
+	if (!(newdb = calloc(size, 1)))
+		return TDB_ERRCODE(TDB_ERR_OOM, -1);
+
+	/* Fill in the header */
+	newdb->version = TDB_VERSION;
+	newdb->hash_size = hash_size;
+#ifdef USE_SPINLOCKS
+	newdb->rwlocks = size;
+#endif
+	if (tdb->flags & TDB_INTERNAL) {
+		tdb->map_size = size;
+		tdb->map_ptr = (char *)newdb;
+		memcpy(&tdb->header, newdb, sizeof(tdb->header));
+		/* Convert the `ondisk' version if asked. */
+		CONVERT(*newdb);
+		return 0;
+	}
+	if (lseek(tdb->fd, 0, SEEK_SET) == -1)
+		goto fail;
+
+	if (ftruncate(tdb->fd, 0) == -1)
+		goto fail;
+
+	/* This creates an endian-converted header, as if read from disk */
+	CONVERT(*newdb);
+	memcpy(&tdb->header, newdb, sizeof(tdb->header));
+	/* Don't endian-convert the magic food! */
+	memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
+	if (write(tdb->fd, newdb, size) != size)
+		ret = -1;
+	else
+		ret = tdb_create_rwlocks(tdb->fd, hash_size);
+
+  fail:
+	SAFE_FREE(newdb);
+	return ret;
+}
+
+/* Returns 0 on fail.  On success, return offset of record, and fills
+   in rec */
+static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
+			struct list_struct *r)
+{
+	tdb_off rec_ptr;
+	
+	/* read in the hash top */
+	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+		return 0;
+
+	/* keep looking until we find the right record */
+	while (rec_ptr) {
+		if (rec_read(tdb, rec_ptr, r) == -1)
+			return 0;
+
+		if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
+			char *k;
+			/* a very likely hit - read the key */
+			k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r), 
+					   r->key_len);
+			if (!k)
+				return 0;
+
+			if (memcmp(key.dptr, k, key.dsize) == 0) {
+				SAFE_FREE(k);
+				return rec_ptr;
+			}
+			SAFE_FREE(k);
+		}
+		rec_ptr = r->next;
+	}
+	return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
+}
+
+/* If they do lockkeys, check that this hash is one they locked */
+static int tdb_keylocked(TDB_CONTEXT *tdb, u32 hash)
+{
+	u32 i;
+	if (!tdb->lockedkeys)
+		return 1;
+	for (i = 0; i < tdb->lockedkeys[0]; i++)
+		if (tdb->lockedkeys[i+1] == hash)
+			return 1;
+	return TDB_ERRCODE(TDB_ERR_NOLOCK, 0);
+}
+
+/* As tdb_find, but if you succeed, keep the lock */
+static tdb_off tdb_find_lock(TDB_CONTEXT *tdb, TDB_DATA key, int locktype,
+			     struct list_struct *rec)
+{
+	u32 hash, rec_ptr;
+
+	hash = tdb_hash(&key);
+	if (!tdb_keylocked(tdb, hash))
+		return 0;
+	if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
+		return 0;
+	if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
+		tdb_unlock(tdb, BUCKET(hash), locktype);
+	return rec_ptr;
+}
+
+enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb)
+{
+	return tdb->ecode;
+}
+
+static struct tdb_errname {
+	enum TDB_ERROR ecode; const char *estring;
+} emap[] = { {TDB_SUCCESS, "Success"},
+	     {TDB_ERR_CORRUPT, "Corrupt database"},
+	     {TDB_ERR_IO, "IO Error"},
+	     {TDB_ERR_LOCK, "Locking error"},
+	     {TDB_ERR_OOM, "Out of memory"},
+	     {TDB_ERR_EXISTS, "Record exists"},
+	     {TDB_ERR_NOLOCK, "Lock exists on other keys"},
+	     {TDB_ERR_NOEXIST, "Record does not exist"} };
+
+/* Error string for the last tdb error */
+const char *tdb_errorstr(TDB_CONTEXT *tdb)
+{
+	u32 i;
+	for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
+		if (tdb->ecode == emap[i].ecode)
+			return emap[i].estring;
+	return "Invalid error code";
+}
+
+/* update an entry in place - this only works if the new data size
+   is <= the old data size and the key exists.
+   on failure return -1.
+*/
+
+static int tdb_update(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf)
+{
+	struct list_struct rec;
+	tdb_off rec_ptr;
+
+	/* find entry */
+	if (!(rec_ptr = tdb_find(tdb, key, tdb_hash(&key), &rec)))
+		return -1;
+
+	/* must be long enough key, data and tailer */
+	if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) {
+		tdb->ecode = TDB_SUCCESS; /* Not really an error */
+		return -1;
+	}
+
+	if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
+		      dbuf.dptr, dbuf.dsize) == -1)
+		return -1;
+
+	if (dbuf.dsize != rec.data_len) {
+		/* update size */
+		rec.data_len = dbuf.dsize;
+		return rec_write(tdb, rec_ptr, &rec);
+	}
+ 
+	return 0;
+}
+
+/* find an entry in the database given a key */
+/* If an entry doesn't exist tdb_err will be set to
+ * TDB_ERR_NOEXIST. If a key has no data attached
+ * tdb_err will not be set. Both will return a
+ * zero pptr and zero dsize.
+ */
+
+TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	tdb_off rec_ptr;
+	struct list_struct rec;
+	TDB_DATA ret;
+
+	/* find which hash bucket it is in */
+	if (!(rec_ptr = tdb_find_lock(tdb,key,F_RDLCK,&rec)))
+		return tdb_null;
+
+	if (rec.data_len)
+		ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
+					  rec.data_len);
+	else
+		ret.dptr = NULL;
+	ret.dsize = rec.data_len;
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+	return ret;
+}
+
+/* check if an entry in the database exists 
+
+   note that 1 is returned if the key is found and 0 is returned if not found
+   this doesn't match the conventions in the rest of this module, but is
+   compatible with gdbm
+*/
+int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	struct list_struct rec;
+	
+	if (tdb_find_lock(tdb, key, F_RDLCK, &rec) == 0)
+		return 0;
+	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
+	return 1;
+}
+
+/* record lock stops delete underneath */
+static int lock_record(TDB_CONTEXT *tdb, tdb_off off)
+{
+	return off ? tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0) : 0;
+}
+/*
+  Write locks override our own fcntl readlocks, so check it here.
+  Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
+  an error to fail to get the lock here.
+*/
+ 
+static int write_lock_record(TDB_CONTEXT *tdb, tdb_off off)
+{
+	struct tdb_traverse_lock *i;
+	for (i = &tdb->travlocks; i; i = i->next)
+		if (i->off == off)
+			return -1;
+	return tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1);
+}
+
+/*
+  Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
+  an error to fail to get the lock here.
+*/
+
+static int write_unlock_record(TDB_CONTEXT *tdb, tdb_off off)
+{
+	return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
+}
+/* fcntl locks don't stack: avoid unlocking someone else's */
+static int unlock_record(TDB_CONTEXT *tdb, tdb_off off)
+{
+	struct tdb_traverse_lock *i;
+	u32 count = 0;
+
+	if (off == 0)
+		return 0;
+	for (i = &tdb->travlocks; i; i = i->next)
+		if (i->off == off)
+			count++;
+	return (count == 1 ? tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0) : 0);
+}
+
+/* actually delete an entry in the database given the offset */
+static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
+{
+	tdb_off last_ptr, i;
+	struct list_struct lastrec;
+
+	if (tdb->read_only) return -1;
+
+	if (write_lock_record(tdb, rec_ptr) == -1) {
+		/* Someone traversing here: mark it as dead */
+		rec->magic = TDB_DEAD_MAGIC;
+		return rec_write(tdb, rec_ptr, rec);
+	}
+	if (write_unlock_record(tdb, rec_ptr) != 0)
+		return -1;
+
+	/* find previous record in hash chain */
+	if (ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
+		return -1;
+	for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
+		if (rec_read(tdb, i, &lastrec) == -1)
+			return -1;
+
+	/* unlink it: next ptr is at start of record. */
+	if (last_ptr == 0)
+		last_ptr = TDB_HASH_TOP(rec->full_hash);
+	if (ofs_write(tdb, last_ptr, &rec->next) == -1)
+		return -1;
+
+	/* recover the space */
+	if (tdb_free(tdb, rec_ptr, rec) == -1)
+		return -1;
+	return 0;
+}
+
+/* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
+static int tdb_next_lock(TDB_CONTEXT *tdb, struct tdb_traverse_lock *tlock,
+			 struct list_struct *rec)
+{
+	int want_next = (tlock->off != 0);
+
+	/* No traversal allows if you've called tdb_lockkeys() */
+	if (tdb->lockedkeys)
+		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
+
+	/* Lock each chain from the start one. */
+	for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
+		if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1)
+			return -1;
+
+		/* No previous record?  Start at top of chain. */
+		if (!tlock->off) {
+			if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
+				     &tlock->off) == -1)
+				goto fail;
+		} else {
+			/* Otherwise unlock the previous record. */
+			if (unlock_record(tdb, tlock->off) != 0)
+				goto fail;
+		}
+
+		if (want_next) {
+			/* We have offset of old record: grab next */
+			if (rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+			tlock->off = rec->next;
+		}
+
+		/* Iterate through chain */
+		while( tlock->off) {
+			tdb_off current;
+			if (rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+			if (!TDB_DEAD(rec)) {
+				/* Woohoo: we found one! */
+				if (lock_record(tdb, tlock->off) != 0)
+					goto fail;
+				return tlock->off;
+			}
+			/* Try to clean dead ones from old traverses */
+			current = tlock->off;
+			tlock->off = rec->next;
+			if (do_delete(tdb, current, rec) != 0)
+				goto fail;
+		}
+		tdb_unlock(tdb, tlock->hash, F_WRLCK);
+		want_next = 0;
+	}
+	/* We finished iteration without finding anything */
+	return TDB_ERRCODE(TDB_SUCCESS, 0);
+
+ fail:
+	tlock->off = 0;
+	if (tdb_unlock(tdb, tlock->hash, F_WRLCK) != 0)
+		TDB_LOG((tdb, 0, "tdb_next_lock: On error unlock failed!\n"));
+	return -1;
+}
+
+/* traverse the entire database - calling fn(tdb, key, data) on each element.
+   return -1 on error or the record count traversed
+   if fn is NULL then it is not called
+   a non-zero return value from fn() indicates that the traversal should stop
+  */
+int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *state)
+{
+	TDB_DATA key, dbuf;
+	struct list_struct rec;
+	struct tdb_traverse_lock tl = { NULL, 0, 0 };
+	int ret, count = 0;
+
+	/* This was in the initializaton, above, but the IRIX compiler
+	 * did not like it.  crh
+	 */
+	tl.next = tdb->travlocks.next;
+
+	/* fcntl locks don't stack: beware traverse inside traverse */
+	tdb->travlocks.next = &tl;
+
+	/* tdb_next_lock places locks on the record returned, and its chain */
+	while ((ret = tdb_next_lock(tdb, &tl, &rec)) > 0) {
+		count++;
+		/* now read the full record */
+		key.dptr = tdb_alloc_read(tdb, tl.off + sizeof(rec), 
+					  rec.key_len + rec.data_len);
+		if (!key.dptr) {
+			ret = -1;
+			if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0)
+				goto out;
+			if (unlock_record(tdb, tl.off) != 0)
+				TDB_LOG((tdb, 0, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
+			goto out;
+		}
+		key.dsize = rec.key_len;
+		dbuf.dptr = key.dptr + rec.key_len;
+		dbuf.dsize = rec.data_len;
+
+		/* Drop chain lock, call out */
+		if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0) {
+			ret = -1;
+			goto out;
+		}
+		if (fn && fn(tdb, key, dbuf, state)) {
+			/* They want us to terminate traversal */
+			ret = count;
+			if (unlock_record(tdb, tl.off) != 0) {
+				TDB_LOG((tdb, 0, "tdb_traverse: unlock_record failed!\n"));;
+				ret = -1;
+			}
+			tdb->travlocks.next = tl.next;
+			SAFE_FREE(key.dptr);
+			return count;
+		}
+		SAFE_FREE(key.dptr);
+	}
+out:
+	tdb->travlocks.next = tl.next;
+	if (ret < 0)
+		return -1;
+	else
+		return count;
+}
+
+/* find the first entry in the database and return its key */
+TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb)
+{
+	TDB_DATA key;
+	struct list_struct rec;
+
+	/* release any old lock */
+	if (unlock_record(tdb, tdb->travlocks.off) != 0)
+		return tdb_null;
+	tdb->travlocks.off = tdb->travlocks.hash = 0;
+
+	if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0)
+		return tdb_null;
+	/* now read the key */
+	key.dsize = rec.key_len;
+	key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
+	if (tdb_unlock(tdb, BUCKET(tdb->travlocks.hash), F_WRLCK) != 0)
+		TDB_LOG((tdb, 0, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
+	return key;
+}
+
+/* find the next entry in the database, returning its key */
+TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA oldkey)
+{
+	u32 oldhash;
+	TDB_DATA key = tdb_null;
+	struct list_struct rec;
+	char *k = NULL;
+
+	/* Is locked key the old key?  If so, traverse will be reliable. */
+	if (tdb->travlocks.off) {
+		if (tdb_lock(tdb,tdb->travlocks.hash,F_WRLCK))
+			return tdb_null;
+		if (rec_read(tdb, tdb->travlocks.off, &rec) == -1
+		    || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
+					    rec.key_len))
+		    || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
+			/* No, it wasn't: unlock it and start from scratch */
+			if (unlock_record(tdb, tdb->travlocks.off) != 0)
+				return tdb_null;
+			if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
+				return tdb_null;
+			tdb->travlocks.off = 0;
+		}
+
+		SAFE_FREE(k);
+	}
+
+	if (!tdb->travlocks.off) {
+		/* No previous element: do normal find, and lock record */
+		tdb->travlocks.off = tdb_find_lock(tdb, oldkey, F_WRLCK, &rec);
+		if (!tdb->travlocks.off)
+			return tdb_null;
+		tdb->travlocks.hash = BUCKET(rec.full_hash);
+		if (lock_record(tdb, tdb->travlocks.off) != 0) {
+			TDB_LOG((tdb, 0, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
+			return tdb_null;
+		}
+	}
+	oldhash = tdb->travlocks.hash;
+
+	/* Grab next record: locks chain and returned record,
+	   unlocks old record */
+	if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
+		key.dsize = rec.key_len;
+		key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
+					  key.dsize);
+		/* Unlock the chain of this new record */
+		if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
+			TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
+	}
+	/* Unlock the chain of old record */
+	if (tdb_unlock(tdb, BUCKET(oldhash), F_WRLCK) != 0)
+		TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
+	return key;
+}
+
+/* delete an entry in the database given a key */
+int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	tdb_off rec_ptr;
+	struct list_struct rec;
+	int ret;
+
+	if (!(rec_ptr = tdb_find_lock(tdb, key, F_WRLCK, &rec)))
+		return -1;
+	ret = do_delete(tdb, rec_ptr, &rec);
+	if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
+		TDB_LOG((tdb, 0, "tdb_delete: WARNING tdb_unlock failed!\n"));
+	return ret;
+}
+
+/* store an element in the database, replacing any existing element
+   with the same key 
+
+   return 0 on success, -1 on failure
+*/
+int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
+{
+	struct list_struct rec;
+	u32 hash;
+	tdb_off rec_ptr;
+	char *p = NULL;
+	int ret = 0;
+
+	/* find which hash bucket it is in */
+	hash = tdb_hash(&key);
+	if (!tdb_keylocked(tdb, hash))
+		return -1;
+	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
+		return -1;
+
+	/* check for it existing, on insert. */
+	if (flag == TDB_INSERT) {
+		if (tdb_exists(tdb, key)) {
+			tdb->ecode = TDB_ERR_EXISTS;
+			goto fail;
+		}
+	} else {
+		/* first try in-place update, on modify or replace. */
+		if (tdb_update(tdb, key, dbuf) == 0)
+			goto out;
+		if (flag == TDB_MODIFY && tdb->ecode == TDB_ERR_NOEXIST)
+			goto fail;
+	}
+	/* reset the error code potentially set by the tdb_update() */
+	tdb->ecode = TDB_SUCCESS;
+
+	/* delete any existing record - if it doesn't exist we don't
+           care.  Doing this first reduces fragmentation, and avoids
+           coalescing with `allocated' block before it's updated. */
+	if (flag != TDB_INSERT)
+		tdb_delete(tdb, key);
+
+	/* Copy key+value *before* allocating free space in case malloc
+	   fails and we are left with a dead spot in the tdb. */
+
+	if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
+		tdb->ecode = TDB_ERR_OOM;
+		goto fail;
+	}
+
+	memcpy(p, key.dptr, key.dsize);
+	if (dbuf.dsize)
+		memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
+
+	/* now we're into insert / modify / replace of a record which
+	 * we know could not be optimised by an in-place store (for
+	 * various reasons).  */
+	if (!(rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec)))
+		goto fail;
+
+	/* Read hash top into next ptr */
+	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
+		goto fail;
+
+	rec.key_len = key.dsize;
+	rec.data_len = dbuf.dsize;
+	rec.full_hash = hash;
+	rec.magic = TDB_MAGIC;
+
+	/* write out and point the top of the hash chain at it */
+	if (rec_write(tdb, rec_ptr, &rec) == -1
+	    || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
+	    || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
+		/* Need to tdb_unallocate() here */
+		goto fail;
+	}
+ out:
+	SAFE_FREE(p); 
+	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+	return ret;
+fail:
+	ret = -1;
+	goto out;
+}
+
+/* Attempt to append data to an entry in place - this only works if the new data size
+   is <= the old data size and the key exists.
+   on failure return -1. Record must be locked before calling.
+*/
+static int tdb_append_inplace(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA new_dbuf)
+{
+	struct list_struct rec;
+	tdb_off rec_ptr;
+
+	/* find entry */
+	if (!(rec_ptr = tdb_find(tdb, key, tdb_hash(&key), &rec)))
+		return -1;
+
+	/* Append of 0 is always ok. */
+	if (new_dbuf.dsize == 0)
+		return 0;
+
+	/* must be long enough for key, old data + new data and tailer */
+	if (rec.rec_len < key.dsize + rec.data_len + new_dbuf.dsize + sizeof(tdb_off)) {
+		/* No room. */
+		tdb->ecode = TDB_SUCCESS; /* Not really an error */
+		return -1;
+	}
+
+	if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len + rec.data_len,
+		      new_dbuf.dptr, new_dbuf.dsize) == -1)
+		return -1;
+
+	/* update size */
+	rec.data_len += new_dbuf.dsize;
+	return rec_write(tdb, rec_ptr, &rec);
+}
+
+/* Append to an entry. Create if not exist. */
+
+int tdb_append(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA new_dbuf)
+{
+	struct list_struct rec;
+	u32 hash;
+	tdb_off rec_ptr;
+	char *p = NULL;
+	int ret = 0;
+	size_t new_data_size = 0;
+
+	/* find which hash bucket it is in */
+	hash = tdb_hash(&key);
+	if (!tdb_keylocked(tdb, hash))
+		return -1;
+	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
+		return -1;
+
+	/* first try in-place. */
+	if (tdb_append_inplace(tdb, key, new_dbuf) == 0)
+		goto out;
+
+	/* reset the error code potentially set by the tdb_append_inplace() */
+	tdb->ecode = TDB_SUCCESS;
+
+	/* find entry */
+	if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
+		if (tdb->ecode != TDB_ERR_NOEXIST)
+			goto fail;
+
+		/* Not found - create. */
+
+		ret = tdb_store(tdb, key, new_dbuf, TDB_INSERT);
+		goto out;
+	}
+
+	new_data_size = rec.data_len + new_dbuf.dsize;
+
+	/* Copy key+old_value+value *before* allocating free space in case malloc
+	   fails and we are left with a dead spot in the tdb. */
+
+	if (!(p = (char *)malloc(key.dsize + new_data_size))) {
+		tdb->ecode = TDB_ERR_OOM;
+		goto fail;
+	}
+
+	/* Copy the key in place. */
+	memcpy(p, key.dptr, key.dsize);
+
+	/* Now read the old data into place. */
+	if (rec.data_len &&
+		tdb_read(tdb, rec_ptr + sizeof(rec) + rec.key_len, p + key.dsize, rec.data_len, 0) == -1)
+			goto fail;
+
+	/* Finally append the new data. */
+	if (new_dbuf.dsize)
+		memcpy(p+key.dsize+rec.data_len, new_dbuf.dptr, new_dbuf.dsize);
+
+	/* delete any existing record - if it doesn't exist we don't
+           care.  Doing this first reduces fragmentation, and avoids
+           coalescing with `allocated' block before it's updated. */
+
+	tdb_delete(tdb, key);
+
+	if (!(rec_ptr = tdb_allocate(tdb, key.dsize + new_data_size, &rec)))
+		goto fail;
+
+	/* Read hash top into next ptr */
+	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
+		goto fail;
+
+	rec.key_len = key.dsize;
+	rec.data_len = new_data_size;
+	rec.full_hash = hash;
+	rec.magic = TDB_MAGIC;
+
+	/* write out and point the top of the hash chain at it */
+	if (rec_write(tdb, rec_ptr, &rec) == -1
+	    || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+new_data_size)==-1
+	    || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
+		/* Need to tdb_unallocate() here */
+		goto fail;
+	}
+
+ out:
+	SAFE_FREE(p); 
+	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+	return ret;
+
+fail:
+	ret = -1;
+	goto out;
+}
+
+static int tdb_already_open(dev_t device,
+			    ino_t ino)
+{
+	TDB_CONTEXT *i;
+	
+	for (i = tdbs; i; i = i->next) {
+		if (i->device == device && i->inode == ino) {
+			return 1;
+		}
+	}
+
+	return 0;
+}
+
+/* open the database, creating it if necessary 
+
+   The open_flags and mode are passed straight to the open call on the
+   database file. A flags value of O_WRONLY is invalid. The hash size
+   is advisory, use zero for a default value.
+
+   Return is NULL on error, in which case errno is also set.  Don't 
+   try to call tdb_error or tdb_errname, just do strerror(errno).
+
+   @param name may be NULL for internal databases. */
+TDB_CONTEXT *tdb_open(const char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode)
+{
+	return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL);
+}
+
+
+TDB_CONTEXT *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
+			 int open_flags, mode_t mode,
+			 tdb_log_func log_fn)
+{
+	TDB_CONTEXT *tdb;
+	struct stat st;
+	int rev = 0, locked;
+	unsigned char *vp;
+	u32 vertest;
+
+	if (!(tdb = calloc(1, sizeof *tdb))) {
+		/* Can't log this */
+		errno = ENOMEM;
+		goto fail;
+	}
+	tdb->fd = -1;
+	tdb->name = NULL;
+	tdb->map_ptr = NULL;
+	tdb->lockedkeys = NULL;
+	tdb->flags = tdb_flags;
+	tdb->open_flags = open_flags;
+	tdb->log_fn = log_fn;
+	
+	if ((open_flags & O_ACCMODE) == O_WRONLY) {
+		TDB_LOG((tdb, 0, "tdb_open_ex: can't open tdb %s write-only\n",
+			 name));
+		errno = EINVAL;
+		goto fail;
+	}
+	
+	if (hash_size == 0)
+		hash_size = DEFAULT_HASH_SIZE;
+	if ((open_flags & O_ACCMODE) == O_RDONLY) {
+		tdb->read_only = 1;
+		/* read only databases don't do locking or clear if first */
+		tdb->flags |= TDB_NOLOCK;
+		tdb->flags &= ~TDB_CLEAR_IF_FIRST;
+	}
+
+	/* internal databases don't mmap or lock, and start off cleared */
+	if (tdb->flags & TDB_INTERNAL) {
+		tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
+		tdb->flags &= ~TDB_CLEAR_IF_FIRST;
+		if (tdb_new_database(tdb, hash_size) != 0) {
+			TDB_LOG((tdb, 0, "tdb_open_ex: tdb_new_database failed!"));
+			goto fail;
+		}
+		goto internal;
+	}
+
+	if ((tdb->fd = open(name, open_flags, mode)) == -1) {
+		TDB_LOG((tdb, 5, "tdb_open_ex: could not open file %s: %s\n",
+			 name, strerror(errno)));
+		goto fail;	/* errno set by open(2) */
+	}
+
+	/* ensure there is only one process initialising at once */
+	if (tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0) == -1) {
+		TDB_LOG((tdb, 0, "tdb_open_ex: failed to get global lock on %s: %s\n",
+			 name, strerror(errno)));
+		goto fail;	/* errno set by tdb_brlock */
+	}
+
+	/* we need to zero database if we are the only one with it open */
+	if ((locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))
+	    && (tdb_flags & TDB_CLEAR_IF_FIRST)) {
+		open_flags |= O_CREAT;
+		if (ftruncate(tdb->fd, 0) == -1) {
+			TDB_LOG((tdb, 0, "tdb_open_ex: "
+				 "failed to truncate %s: %s\n",
+				 name, strerror(errno)));
+			goto fail; /* errno set by ftruncate */
+		}
+	}
+
+	if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
+	    || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0
+	    || tdb->header.version != TDB_VERSION
+	    || (tdb->header.hash_size != hash_size
+		&& !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION))))) {
+		/* its not a valid database - possibly initialise it */
+		if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
+			errno = EIO; /* ie bad format or something */
+			goto fail;
+		}
+		rev = (tdb->flags & TDB_CONVERT);
+	}
+	vp = (unsigned char *)&tdb->header.version;
+	vertest = (((u32)vp[0]) << 24) | (((u32)vp[1]) << 16) |
+		  (((u32)vp[2]) << 8) | (u32)vp[3];
+	tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
+	if (!rev)
+		tdb->flags &= ~TDB_CONVERT;
+	else {
+		tdb->flags |= TDB_CONVERT;
+		convert(&tdb->header, sizeof(tdb->header));
+	}
+	if (fstat(tdb->fd, &st) == -1)
+		goto fail;
+
+	/* Is it already in the open list?  If so, fail. */
+	if (tdb_already_open(st.st_dev, st.st_ino)) {
+		TDB_LOG((tdb, 2, "tdb_open_ex: "
+			 "%s (%d,%d) is already open in this process\n",
+			 name, st.st_dev, st.st_ino));
+		errno = EBUSY;
+		goto fail;
+	}
+
+	if (!(tdb->name = (char *)strdup(name))) {
+		errno = ENOMEM;
+		goto fail;
+	}
+
+	tdb->map_size = st.st_size;
+	tdb->device = st.st_dev;
+	tdb->inode = st.st_ino;
+	tdb->locked = calloc(tdb->header.hash_size+1, sizeof(tdb->locked[0]));
+	if (!tdb->locked) {
+		TDB_LOG((tdb, 2, "tdb_open_ex: "
+			 "failed to allocate lock structure for %s\n",
+			 name));
+		errno = ENOMEM;
+		goto fail;
+	}
+	tdb_mmap(tdb);
+	if (locked) {
+		if (!tdb->read_only)
+			if (tdb_clear_spinlocks(tdb) != 0) {
+				TDB_LOG((tdb, 0, "tdb_open_ex: "
+				"failed to clear spinlock\n"));
+				goto fail;
+			}
+		if (tdb_brlock(tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0) == -1) {
+			TDB_LOG((tdb, 0, "tdb_open_ex: "
+				 "failed to take ACTIVE_LOCK on %s: %s\n",
+				 name, strerror(errno)));
+			goto fail;
+		}
+	}
+	/* leave this lock in place to indicate it's in use */
+	if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1)
+		goto fail;
+
+ internal:
+	/* Internal (memory-only) databases skip all the code above to
+	 * do with disk files, and resume here by releasing their
+	 * global lock and hooking into the active list. */
+	if (tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0) == -1)
+		goto fail;
+	tdb->next = tdbs;
+	tdbs = tdb;
+	return tdb;
+
+ fail:
+	{ int save_errno = errno;
+
+	if (!tdb)
+		return NULL;
+	
+	if (tdb->map_ptr) {
+		if (tdb->flags & TDB_INTERNAL)
+			SAFE_FREE(tdb->map_ptr);
+		else
+			tdb_munmap(tdb);
+	}
+	SAFE_FREE(tdb->name);
+	if (tdb->fd != -1)
+		if (close(tdb->fd) != 0)
+			TDB_LOG((tdb, 5, "tdb_open_ex: failed to close tdb->fd on error!\n"));
+	SAFE_FREE(tdb->locked);
+	SAFE_FREE(tdb);
+	errno = save_errno;
+	return NULL;
+	}
+}
+
+/**
+ * Close a database.
+ *
+ * @returns -1 for error; 0 for success.
+ **/
+int tdb_close(TDB_CONTEXT *tdb)
+{
+	TDB_CONTEXT **i;
+	int ret = 0;
+
+	if (tdb->map_ptr) {
+		if (tdb->flags & TDB_INTERNAL)
+			SAFE_FREE(tdb->map_ptr);
+		else
+			tdb_munmap(tdb);
+	}
+	SAFE_FREE(tdb->name);
+	if (tdb->fd != -1)
+		ret = close(tdb->fd);
+	SAFE_FREE(tdb->locked);
+	SAFE_FREE(tdb->lockedkeys);
+
+	/* Remove from contexts list */
+	for (i = &tdbs; *i; i = &(*i)->next) {
+		if (*i == tdb) {
+			*i = tdb->next;
+			break;
+		}
+	}
+
+	memset(tdb, 0, sizeof(*tdb));
+	SAFE_FREE(tdb);
+
+	return ret;
+}
+
+/* lock/unlock entire database */
+int tdb_lockall(TDB_CONTEXT *tdb)
+{
+	u32 i;
+
+	/* There are no locks on read-only dbs */
+	if (tdb->read_only)
+		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
+	if (tdb->lockedkeys)
+		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
+	for (i = 0; i < tdb->header.hash_size; i++) 
+		if (tdb_lock(tdb, i, F_WRLCK))
+			break;
+
+	/* If error, release locks we have... */
+	if (i < tdb->header.hash_size) {
+		u32 j;
+
+		for ( j = 0; j < i; j++)
+			tdb_unlock(tdb, j, F_WRLCK);
+		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
+	}
+
+	return 0;
+}
+void tdb_unlockall(TDB_CONTEXT *tdb)
+{
+	u32 i;
+	for (i=0; i < tdb->header.hash_size; i++)
+		tdb_unlock(tdb, i, F_WRLCK);
+}
+
+int tdb_lockkeys(TDB_CONTEXT *tdb, u32 number, TDB_DATA keys[])
+{
+	u32 i, j, hash;
+
+	/* Can't lock more keys if already locked */
+	if (tdb->lockedkeys)
+		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
+	if (!(tdb->lockedkeys = malloc(sizeof(u32) * (number+1))))
+		return TDB_ERRCODE(TDB_ERR_OOM, -1);
+	/* First number in array is # keys */
+	tdb->lockedkeys[0] = number;
+
+	/* Insertion sort by bucket */
+	for (i = 0; i < number; i++) {
+		hash = tdb_hash(&keys[i]);
+		for (j = 0; j < i && BUCKET(tdb->lockedkeys[j+1]) < BUCKET(hash); j++);
+			memmove(&tdb->lockedkeys[j+2], &tdb->lockedkeys[j+1], sizeof(u32) * (i-j));
+		tdb->lockedkeys[j+1] = hash;
+	}
+	/* Finally, lock in order */
+	for (i = 0; i < number; i++)
+		if (tdb_lock(tdb, i, F_WRLCK))
+			break;
+
+	/* If error, release locks we have... */
+	if (i < number) {
+		for ( j = 0; j < i; j++)
+			tdb_unlock(tdb, j, F_WRLCK);
+		SAFE_FREE(tdb->lockedkeys);
+		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
+	}
+	return 0;
+}
+
+/* Unlock the keys previously locked by tdb_lockkeys() */
+void tdb_unlockkeys(TDB_CONTEXT *tdb)
+{
+	u32 i;
+	for (i = 0; i < tdb->lockedkeys[0]; i++)
+		tdb_unlock(tdb, tdb->lockedkeys[i+1], F_WRLCK);
+	SAFE_FREE(tdb->lockedkeys);
+}
+
+/* lock/unlock one hash chain. This is meant to be used to reduce
+   contention - it cannot guarantee how many records will be locked */
+int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
+}
+
+int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	return tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
+}
+
+int tdb_chainlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_RDLCK);
+}
+
+int tdb_chainunlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
+{
+	return tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_RDLCK);
+}
+
+
+/* register a loging function */
+void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
+{
+	tdb->log_fn = fn;
+}
+
+
+/* reopen a tdb - this is used after a fork to ensure that we have an independent
+   seek pointer from our parent and to re-establish locks */
+int tdb_reopen(TDB_CONTEXT *tdb)
+{
+	struct stat st;
+
+	if (tdb_munmap(tdb) != 0) {
+		TDB_LOG((tdb, 0, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (close(tdb->fd) != 0)
+		TDB_LOG((tdb, 0, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
+	tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
+	if (tdb->fd == -1) {
+		TDB_LOG((tdb, 0, "tdb_reopen: open failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (fstat(tdb->fd, &st) != 0) {
+		TDB_LOG((tdb, 0, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
+		goto fail;
+	}
+	if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
+		TDB_LOG((tdb, 0, "tdb_reopen: file dev/inode has changed!\n"));
+		goto fail;
+	}
+	tdb_mmap(tdb);
+	if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1) {
+		TDB_LOG((tdb, 0, "tdb_reopen: failed to obtain active lock\n"));
+		goto fail;
+	}
+
+	return 0;
+
+fail:
+	tdb_close(tdb);
+	return -1;
+}
+
+/* reopen all tdb's */
+int tdb_reopen_all(void)
+{
+	TDB_CONTEXT *tdb;
+
+	for (tdb=tdbs; tdb; tdb = tdb->next) {
+		if (tdb_reopen(tdb) != 0) return -1;
+	}
+
+	return 0;
+}
diff --git a/source4/tdb/tdb.h b/source4/tdb/tdb.h
new file mode 100644
index 0000000000..6f3b1ff756
--- /dev/null
+++ b/source4/tdb/tdb.h
@@ -0,0 +1,144 @@
+#ifndef __TDB_H__
+#define __TDB_H__
+
+/* 
+   Unix SMB/CIFS implementation.
+   Samba database functions
+   Copyright (C) Andrew Tridgell 1999
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+
+/* flags to tdb_store() */
+#define TDB_REPLACE 1
+#define TDB_INSERT 2
+#define TDB_MODIFY 3
+
+/* flags for tdb_open() */
+#define TDB_DEFAULT 0 /* just a readability place holder */
+#define TDB_CLEAR_IF_FIRST 1
+#define TDB_INTERNAL 2 /* don't store on disk */
+#define TDB_NOLOCK   4 /* don't do any locking */
+#define TDB_NOMMAP   8 /* don't use mmap */
+#define TDB_CONVERT 16 /* convert endian (internal use) */
+#define TDB_BIGENDIAN 32 /* header is big-endian (internal use) */
+
+#define TDB_ERRCODE(code, ret) ((tdb->ecode = (code)), ret)
+
+/* error codes */
+enum TDB_ERROR {TDB_SUCCESS=0, TDB_ERR_CORRUPT, TDB_ERR_IO, TDB_ERR_LOCK, 
+		TDB_ERR_OOM, TDB_ERR_EXISTS, TDB_ERR_NOEXIST, TDB_ERR_NOLOCK, TDB_ERR_LOCK_TIMEOUT };
+
+#ifndef u32
+#define u32 unsigned
+#endif
+
+typedef struct {
+	char *dptr;
+	size_t dsize;
+} TDB_DATA;
+
+typedef u32 tdb_len;
+typedef u32 tdb_off;
+
+/* this is stored at the front of every database */
+struct tdb_header {
+	char magic_food[32]; /* for /etc/magic */
+	u32 version; /* version of the code */
+	u32 hash_size; /* number of hash entries */
+	tdb_off rwlocks;
+	tdb_off reserved[31];
+};
+
+struct tdb_lock_type {
+	u32 count;
+	u32 ltype;
+};
+
+struct tdb_traverse_lock {
+	struct tdb_traverse_lock *next;
+	u32 off;
+	u32 hash;
+};
+
+/* this is the context structure that is returned from a db open */
+typedef struct tdb_context {
+	char *name; /* the name of the database */
+	void *map_ptr; /* where it is currently mapped */
+	int fd; /* open file descriptor for the database */
+	tdb_len map_size; /* how much space has been mapped */
+	int read_only; /* opened read-only */
+	struct tdb_lock_type *locked; /* array of chain locks */
+	enum TDB_ERROR ecode; /* error code for last tdb error */
+	struct tdb_header header; /* a cached copy of the header */
+	u32 flags; /* the flags passed to tdb_open */
+	u32 *lockedkeys; /* array of locked keys: first is #keys */
+	struct tdb_traverse_lock travlocks; /* current traversal locks */
+	struct tdb_context *next; /* all tdbs to avoid multiple opens */
+	dev_t device;	/* uniquely identifies this tdb */
+	ino_t inode;	/* uniquely identifies this tdb */
+	void (*log_fn)(struct tdb_context *tdb, int level, const char *, ...); /* logging function */
+	int open_flags; /* flags used in the open - needed by reopen */
+} TDB_CONTEXT;
+
+typedef int (*tdb_traverse_func)(TDB_CONTEXT *, TDB_DATA, TDB_DATA, void *);
+typedef void (*tdb_log_func)(TDB_CONTEXT *, int , const char *, ...);
+
+TDB_CONTEXT *tdb_open(const char *name, int hash_size, int tdb_flags,
+		      int open_flags, mode_t mode);
+TDB_CONTEXT *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
+			 int open_flags, mode_t mode,
+			 tdb_log_func log_fn);
+
+int tdb_reopen(TDB_CONTEXT *tdb);
+int tdb_reopen_all(void);
+void tdb_logging_function(TDB_CONTEXT *tdb, tdb_log_func);
+enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb);
+const char *tdb_errorstr(TDB_CONTEXT *tdb);
+TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key);
+int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key);
+int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag);
+int tdb_append(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA new_dbuf);
+int tdb_close(TDB_CONTEXT *tdb);
+TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb);
+TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA key);
+int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *state);
+int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key);
+int tdb_lockkeys(TDB_CONTEXT *tdb, u32 number, TDB_DATA keys[]);
+void tdb_unlockkeys(TDB_CONTEXT *tdb);
+int tdb_lockall(TDB_CONTEXT *tdb);
+void tdb_unlockall(TDB_CONTEXT *tdb);
+
+/* Low level locking functions: use with care */
+void tdb_set_lock_alarm(sig_atomic_t *palarm);
+int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key);
+int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key);
+
+/* Debug functions. Not used in production. */
+void tdb_dump_all(TDB_CONTEXT *tdb);
+int tdb_printfreelist(TDB_CONTEXT *tdb);
+
+extern TDB_DATA tdb_null;
+
+#ifdef  __cplusplus
+}
+#endif
+
+#endif /* tdb.h */
diff --git a/source4/tdb/tdb.magic b/source4/tdb/tdb.magic
new file mode 100644
index 0000000000..f5619e7327
--- /dev/null
+++ b/source4/tdb/tdb.magic
@@ -0,0 +1,10 @@
+# Magic file(1) information about tdb files.
+#
+# Install this into /etc/magic or the corresponding location for your
+# system, or pass as a -m argument to file(1).
+
+# You may use and redistribute this file without restriction.
+
+0	string	TDB\ file		TDB database
+>32	lelong	=0x2601196D		version 6, little-endian
+>>36	lelong	x			hash size %d bytes
diff --git a/source4/tdb/tdbbackup.c b/source4/tdb/tdbbackup.c
new file mode 100644
index 0000000000..7b344de6c4
--- /dev/null
+++ b/source4/tdb/tdbbackup.c
@@ -0,0 +1,315 @@
+/* 
+   Unix SMB/CIFS implementation.
+   low level tdb backup and restore utility
+   Copyright (C) Andrew Tridgell              2002
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+/*
+
+  This program is meant for backup/restore of tdb databases. Typical usage would be:
+     tdbbackup *.tdb
+  when Samba shuts down cleanly, which will make a backup of all the local databases
+  to *.bak files. Then on Samba startup you would use:
+     tdbbackup -v *.tdb
+  and this will check the databases for corruption and if corruption is detected then
+  the backup will be restored.
+
+  You may also like to do a backup on a regular basis while Samba is
+  running, perhaps using cron.
+
+  The reason this program is needed is to cope with power failures
+  while Samba is running. A power failure could lead to database
+  corruption and Samba will then not start correctly.
+
+  Note that many of the databases in Samba are transient and thus
+  don't need to be backed up, so you can optimise the above a little
+  by only running the backup on the critical databases.
+
+ */
+
+#include <errno.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <ctype.h>
+#include <signal.h>
+#include "tdb.h"
+
+static int failed;
+
+static char *add_suffix(const char *name, const char *suffix)
+{
+	char *ret;
+	int len = strlen(name) + strlen(suffix) + 1;
+	ret = malloc(len);
+	if (!ret) {
+		fprintf(stderr,"Out of memory!\n");
+		exit(1);
+	}
+	strncpy(ret, name, len);
+	strncat(ret, suffix, len);
+	return ret;
+}
+
+static int copy_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	TDB_CONTEXT *tdb_new = (TDB_CONTEXT *)state;
+
+	if (tdb_store(tdb_new, key, dbuf, TDB_INSERT) != 0) {
+		fprintf(stderr,"Failed to insert into %s\n", tdb_new->name);
+		failed = 1;
+		return 1;
+	}
+	return 0;
+}
+
+
+static int test_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	return 0;
+}
+
+/*
+  carefully backup a tdb, validating the contents and
+  only doing the backup if its OK
+  this function is also used for restore
+*/
+static int backup_tdb(const char *old_name, const char *new_name)
+{
+	TDB_CONTEXT *tdb;
+	TDB_CONTEXT *tdb_new;
+	char *tmp_name;
+	struct stat st;
+	int count1, count2;
+
+	tmp_name = add_suffix(new_name, ".tmp");
+
+	/* stat the old tdb to find its permissions */
+	if (stat(old_name, &st) != 0) {
+		perror(old_name);
+		return 1;
+	}
+
+	/* open the old tdb */
+	tdb = tdb_open(old_name, 0, 0, O_RDWR, 0);
+	if (!tdb) {
+		printf("Failed to open %s\n", old_name);
+		return 1;
+	}
+
+	/* create the new tdb */
+	unlink(tmp_name);
+	tdb_new = tdb_open(tmp_name, tdb->header.hash_size, 
+			   TDB_DEFAULT, O_RDWR|O_CREAT|O_EXCL, 
+			   st.st_mode & 0777);
+	if (!tdb_new) {
+		perror(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* lock the old tdb */
+	if (tdb_lockall(tdb) != 0) {
+		fprintf(stderr,"Failed to lock %s\n", old_name);
+		tdb_close(tdb);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	failed = 0;
+
+	/* traverse and copy */
+	count1 = tdb_traverse(tdb, copy_fn, (void *)tdb_new);
+	if (count1 < 0 || failed) {
+		fprintf(stderr,"failed to copy %s\n", old_name);
+		tdb_close(tdb);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* close the old tdb */
+	tdb_close(tdb);
+
+	/* close the new tdb and re-open read-only */
+	tdb_close(tdb_new);
+	tdb_new = tdb_open(tmp_name, 0, TDB_DEFAULT, O_RDONLY, 0);
+	if (!tdb_new) {
+		fprintf(stderr,"failed to reopen %s\n", tmp_name);
+		unlink(tmp_name);
+		perror(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+	
+	/* traverse the new tdb to confirm */
+	count2 = tdb_traverse(tdb_new, test_fn, 0);
+	if (count2 != count1) {
+		fprintf(stderr,"failed to copy %s\n", old_name);
+		tdb_close(tdb_new);
+		unlink(tmp_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	/* make sure the new tdb has reached stable storage */
+	fsync(tdb_new->fd);
+
+	/* close the new tdb and rename it to .bak */
+	tdb_close(tdb_new);
+	unlink(new_name);
+	if (rename(tmp_name, new_name) != 0) {
+		perror(new_name);
+		free(tmp_name);
+		return 1;
+	}
+
+	printf("%s : %d records\n", old_name, count1);
+	free(tmp_name);
+
+	return 0;
+}
+
+
+
+/*
+  verify a tdb and if it is corrupt then restore from *.bak
+*/
+static int verify_tdb(const char *fname, const char *bak_name)
+{
+	TDB_CONTEXT *tdb;
+	int count = -1;
+
+	/* open the tdb */
+	tdb = tdb_open(fname, 0, 0, O_RDONLY, 0);
+
+	/* traverse the tdb, then close it */
+	if (tdb) {
+		count = tdb_traverse(tdb, test_fn, NULL);
+		tdb_close(tdb);
+	}
+
+	/* count is < 0 means an error */
+	if (count < 0) {
+		printf("restoring %s\n", fname);
+		return backup_tdb(bak_name, fname);
+	}
+
+	printf("%s : %d records\n", fname, count);
+
+	return 0;
+}
+
+
+/*
+  see if one file is newer than another
+*/
+static int file_newer(const char *fname1, const char *fname2)
+{
+	struct stat st1, st2;
+	if (stat(fname1, &st1) != 0) {
+		return 0;
+	}
+	if (stat(fname2, &st2) != 0) {
+		return 1;
+	}
+	return (st1.st_mtime > st2.st_mtime);
+}
+
+static void usage(void)
+{
+	printf("Usage: tdbbackup [options] <fname...>\n\n");
+	printf("   -h            this help message\n");
+	printf("   -s suffix     set the backup suffix\n");
+	printf("   -v            veryify mode (restore if corrupt)\n");
+}
+		
+
+ int main(int argc, char *argv[])
+{
+	int i;
+	int ret = 0;
+	int c;
+	int verify = 0;
+	char *suffix = ".bak";
+	extern int optind;
+	extern char *optarg;
+
+	while ((c = getopt(argc, argv, "vhs:")) != -1) {
+		switch (c) {
+		case 'h':
+			usage();
+			exit(0);
+		case 'v':
+			verify = 1;
+			break;
+		case 's':
+			suffix = optarg;
+			break;
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		usage();
+		exit(1);
+	}
+
+	for (i=0; i<argc; i++) {
+		const char *fname = argv[i];
+		char *bak_name;
+
+		bak_name = add_suffix(fname, suffix);
+
+		if (verify) {
+			if (verify_tdb(fname, bak_name) != 0) {
+				ret = 1;
+			}
+		} else {
+			if (file_newer(fname, bak_name) &&
+			    backup_tdb(fname, bak_name) != 0) {
+				ret = 1;
+			}
+		}
+
+		free(bak_name);
+	}
+
+	return ret;
+}
+
+#ifdef VALGRIND
+size_t valgrind_strlen(const char *s)
+{
+	size_t count;
+	for(count = 0; *s++; count++)
+		;
+	return count;
+}
+#endif
diff --git a/source4/tdb/tdbdump.c b/source4/tdb/tdbdump.c
new file mode 100644
index 0000000000..9c1dc2761b
--- /dev/null
+++ b/source4/tdb/tdbdump.c
@@ -0,0 +1,89 @@
+/* 
+   Unix SMB/CIFS implementation.
+   simple tdb dump util
+   Copyright (C) Andrew Tridgell              2001
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include <errno.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <ctype.h>
+#include <signal.h>
+#include "tdb.h"
+
+static void print_data(TDB_DATA d)
+{
+	unsigned char *p = d.dptr;
+	int len = d.dsize;
+	while (len--) {
+		if (isprint(*p) && !strchr("\"\\", *p)) {
+			fputc(*p, stdout);
+		} else {
+			printf("\\%02X", *p);
+		}
+		p++;
+	}
+}
+
+static int traverse_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	printf("{\n");
+	printf("key = \"");
+	print_data(key);
+	printf("\"\n");
+	printf("data = \"");
+	print_data(dbuf);
+	printf("\"\n");
+	printf("}\n");
+	return 0;
+}
+
+static int dump_tdb(const char *fname)
+{
+	TDB_CONTEXT *tdb;
+	
+	tdb = tdb_open(fname, 0, 0, O_RDONLY, 0);
+	if (!tdb) {
+		printf("Failed to open %s\n", fname);
+		return 1;
+	}
+
+	tdb_traverse(tdb, traverse_fn, NULL);
+	return 0;
+}
+
+ int main(int argc, char *argv[])
+{
+	char *fname;
+
+	if (argc < 2) {
+		printf("Usage: tdbdump <fname>\n");
+		exit(1);
+	}
+
+	fname = argv[1];
+
+	return dump_tdb(fname);
+}
diff --git a/source4/tdb/tdbtest.c b/source4/tdb/tdbtest.c
new file mode 100644
index 0000000000..89295a3291
--- /dev/null
+++ b/source4/tdb/tdbtest.c
@@ -0,0 +1,263 @@
+#include <stdlib.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <signal.h>
+#include "tdb.h"
+#include <gdbm.h>
+
+/* a test program for tdb - the trivial database */
+
+
+
+#define DELETE_PROB 7
+#define STORE_PROB 5
+
+static TDB_CONTEXT *db;
+static GDBM_FILE gdbm;
+
+struct timeval tp1,tp2;
+
+static void start_timer(void)
+{
+	gettimeofday(&tp1,NULL);
+}
+
+static double end_timer(void)
+{
+	gettimeofday(&tp2,NULL);
+	return((tp2.tv_sec - tp1.tv_sec) + 
+	       (tp2.tv_usec - tp1.tv_usec)*1.0e-6);
+}
+
+static void fatal(char *why)
+{
+	perror(why);
+	exit(1);
+}
+
+static void tdb_log(TDB_CONTEXT *tdb, int level, const char *format, ...)
+{
+	va_list ap;
+    
+	va_start(ap, format);
+	vfprintf(stdout, format, ap);
+	va_end(ap);
+	fflush(stdout);
+}
+
+static void compare_db(void)
+{
+	TDB_DATA d, key, nextkey;
+	datum gd, gkey, gnextkey;
+
+	key = tdb_firstkey(db);
+	while (key.dptr) {
+		d = tdb_fetch(db, key);
+		gkey.dptr = key.dptr;
+		gkey.dsize = key.dsize;
+
+		gd = gdbm_fetch(gdbm, gkey);
+
+		if (!gd.dptr) fatal("key not in gdbm");
+		if (gd.dsize != d.dsize) fatal("data sizes differ");
+		if (memcmp(gd.dptr, d.dptr, d.dsize)) {
+			fatal("data differs");
+		}
+
+		nextkey = tdb_nextkey(db, key);
+		free(key.dptr);
+		free(d.dptr);
+		free(gd.dptr);
+		key = nextkey;
+	}
+
+	gkey = gdbm_firstkey(gdbm);
+	while (gkey.dptr) {
+		gd = gdbm_fetch(gdbm, gkey);
+		key.dptr = gkey.dptr;
+		key.dsize = gkey.dsize;
+
+		d = tdb_fetch(db, key);
+
+		if (!d.dptr) fatal("key not in db");
+		if (d.dsize != gd.dsize) fatal("data sizes differ");
+		if (memcmp(d.dptr, gd.dptr, gd.dsize)) {
+			fatal("data differs");
+		}
+
+		gnextkey = gdbm_nextkey(gdbm, gkey);
+		free(gkey.dptr);
+		free(gd.dptr);
+		free(d.dptr);
+		gkey = gnextkey;
+	}
+}
+
+static char *randbuf(int len)
+{
+	char *buf;
+	int i;
+	buf = (char *)malloc(len+1);
+
+	for (i=0;i<len;i++) {
+		buf[i] = 'a' + (rand() % 26);
+	}
+	buf[i] = 0;
+	return buf;
+}
+
+static void addrec_db(void)
+{
+	int klen, dlen;
+	char *k, *d;
+	TDB_DATA key, data;
+
+	klen = 1 + (rand() % 4);
+	dlen = 1 + (rand() % 100);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+
+	key.dptr = k;
+	key.dsize = klen+1;
+
+	data.dptr = d;
+	data.dsize = dlen+1;
+
+	if (rand() % DELETE_PROB == 0) {
+		tdb_delete(db, key);
+	} else if (rand() % STORE_PROB == 0) {
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+	} else {
+		data = tdb_fetch(db, key);
+		if (data.dptr) free(data.dptr);
+	}
+
+	free(k);
+	free(d);
+}
+
+static void addrec_gdbm(void)
+{
+	int klen, dlen;
+	char *k, *d;
+	datum key, data;
+
+	klen = 1 + (rand() % 4);
+	dlen = 1 + (rand() % 100);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+
+	key.dptr = k;
+	key.dsize = klen+1;
+
+	data.dptr = d;
+	data.dsize = dlen+1;
+
+	if (rand() % DELETE_PROB == 0) {
+		gdbm_delete(gdbm, key);
+	} else if (rand() % STORE_PROB == 0) {
+		if (gdbm_store(gdbm, key, data, GDBM_REPLACE) != 0) {
+			fatal("gdbm_store failed");
+		}
+	} else {
+		data = gdbm_fetch(gdbm, key);
+		if (data.dptr) free(data.dptr);
+	}
+
+	free(k);
+	free(d);
+}
+
+static int traverse_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+#if 0
+	printf("[%s] [%s]\n", key.dptr, dbuf.dptr);
+#endif
+	tdb_delete(tdb, key);
+	return 0;
+}
+
+static void merge_test(void)
+{
+	int i;
+	char keys[5][2];
+	TDB_DATA key, data;
+	
+	for (i = 0; i < 5; i++) {
+		sprintf(keys[i], "%d", i);
+		key.dptr = keys[i];
+		key.dsize = 2;
+		
+		data.dptr = "test";
+		data.dsize = 4;
+		
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+	}
+
+	key.dptr = keys[0];
+	tdb_delete(db, key);
+	key.dptr = keys[4];
+	tdb_delete(db, key);
+	key.dptr = keys[2];
+	tdb_delete(db, key);
+	key.dptr = keys[1];
+	tdb_delete(db, key);
+	key.dptr = keys[3];
+	tdb_delete(db, key);
+}
+	
+int main(int argc, char *argv[])
+{
+	int i, seed=0;
+	int loops = 10000;
+
+	unlink("test.gdbm");
+
+	db = tdb_open("test.tdb", 0, TDB_CLEAR_IF_FIRST, 
+		      O_RDWR | O_CREAT | O_TRUNC, 0600);
+	gdbm = gdbm_open("test.gdbm", 512, GDBM_WRITER|GDBM_NEWDB|GDBM_FAST, 
+			 0600, NULL);
+
+	if (!db || !gdbm) {
+		fatal("db open failed");
+	}
+
+	tdb_logging_function(db, tdb_log);
+	
+#if 1
+	srand(seed);
+	start_timer();
+	for (i=0;i<loops;i++) addrec_gdbm();
+	printf("gdbm got %.2f ops/sec\n", i/end_timer());
+#endif
+
+	merge_test();
+
+	srand(seed);
+	start_timer();
+	for (i=0;i<loops;i++) addrec_db();
+	printf("tdb got %.2f ops/sec\n", i/end_timer());
+
+	compare_db();
+
+	printf("traversed %d records\n", tdb_traverse(db, traverse_fn, NULL));
+	printf("traversed %d records\n", tdb_traverse(db, traverse_fn, NULL));
+
+	tdb_close(db);
+	gdbm_close(gdbm);
+
+	return 0;
+}
diff --git a/source4/tdb/tdbtool.c b/source4/tdb/tdbtool.c
new file mode 100644
index 0000000000..f5e486be14
--- /dev/null
+++ b/source4/tdb/tdbtool.c
@@ -0,0 +1,482 @@
+/* 
+   Unix SMB/CIFS implementation.
+   Samba database functions
+   Copyright (C) Andrew Tridgell              1999-2000
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000
+   Copyright (C) Andrew Esh                        2001
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include <errno.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <ctype.h>
+#include <signal.h>
+#include "tdb.h"
+
+/* a tdb tool for manipulating a tdb database */
+
+#define FSTRING_LEN 256
+typedef char fstring[FSTRING_LEN];
+
+typedef struct connections_key {
+	pid_t pid;
+	int cnum;
+	fstring name;
+} connections_key;
+
+typedef struct connections_data {
+	int magic;
+	pid_t pid;
+	int cnum;
+	uid_t uid;
+	gid_t gid;
+	char name[24];
+	char addr[24];
+	char machine[128];
+	time_t start;
+} connections_data;
+
+static TDB_CONTEXT *tdb;
+
+static int print_rec(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state);
+
+static void print_asc(unsigned char *buf,int len)
+{
+	int i;
+
+	/* We're probably printing ASCII strings so don't try to display
+	   the trailing NULL character. */
+
+	if (buf[len - 1] == 0)
+	        len--;
+
+	for (i=0;i<len;i++)
+		printf("%c",isprint(buf[i])?buf[i]:'.');
+}
+
+static void print_data(unsigned char *buf,int len)
+{
+	int i=0;
+	if (len<=0) return;
+	printf("[%03X] ",i);
+	for (i=0;i<len;) {
+		printf("%02X ",(int)buf[i]);
+		i++;
+		if (i%8 == 0) printf(" ");
+		if (i%16 == 0) {      
+			print_asc(&buf[i-16],8); printf(" ");
+			print_asc(&buf[i-8],8); printf("\n");
+			if (i<len) printf("[%03X] ",i);
+		}
+	}
+	if (i%16) {
+		int n;
+		
+		n = 16 - (i%16);
+		printf(" ");
+		if (n>8) printf(" ");
+		while (n--) printf("   ");
+		
+		n = i%16;
+		if (n > 8) n = 8;
+		print_asc(&buf[i-(i%16)],n); printf(" ");
+		n = (i%16) - n;
+		if (n>0) print_asc(&buf[i-n],n); 
+		printf("\n");    
+	}
+}
+
+static void help(void)
+{
+	printf("
+tdbtool: 
+  create    dbname     : create a database
+  open      dbname     : open an existing database
+  erase                : erase the database
+  dump                 : dump the database as strings
+  insert    key  data  : insert a record
+  store     key  data  : store a record (replace)
+  show      key        : show a record by key
+  delete    key        : delete a record by key
+  list                 : print the database hash table and freelist
+  free                 : print the database freelist
+  1 | first            : print the first record
+  n | next             : print the next record
+  q | quit             : terminate
+  \\n                   : repeat 'next' command
+");
+}
+
+static void terror(char *why)
+{
+	printf("%s\n", why);
+}
+
+static char *get_token(int startover)
+{
+	static char tmp[1024];
+	static char *cont = NULL;
+	char *insert, *start;
+	char *k = strtok(NULL, " ");
+
+	if (!k)
+	  return NULL;
+
+	if (startover)
+	  start = tmp;
+	else
+	  start = cont;
+
+	strcpy(start, k);
+	insert = start + strlen(start) - 1;
+	while (*insert == '\\') {
+	  *insert++ = ' ';
+	  k = strtok(NULL, " ");
+	  if (!k)
+	    break;
+	  strcpy(insert, k);
+	  insert = start + strlen(start) - 1;
+	}
+
+	/* Get ready for next call */
+	cont = start + strlen(start) + 1;
+	return start;
+}
+
+static void create_tdb(void)
+{
+	char *tok = get_token(1);
+	if (!tok) {
+		help();
+		return;
+	}
+	if (tdb) tdb_close(tdb);
+	tdb = tdb_open(tok, 0, TDB_CLEAR_IF_FIRST,
+		       O_RDWR | O_CREAT | O_TRUNC, 0600);
+	if (!tdb) {
+		printf("Could not create %s: %s\n", tok, strerror(errno));
+	}
+}
+
+static void open_tdb(void)
+{
+	char *tok = get_token(1);
+	if (!tok) {
+		help();
+		return;
+	}
+	if (tdb) tdb_close(tdb);
+	tdb = tdb_open(tok, 0, 0, O_RDWR, 0600);
+	if (!tdb) {
+		printf("Could not open %s: %s\n", tok, strerror(errno));
+	}
+}
+
+static void insert_tdb(void)
+{
+	char *k = get_token(1);
+	char *d = get_token(0);
+	TDB_DATA key, dbuf;
+
+	if (!k || !d) {
+		help();
+		return;
+	}
+
+	key.dptr = k;
+	key.dsize = strlen(k)+1;
+	dbuf.dptr = d;
+	dbuf.dsize = strlen(d)+1;
+
+	if (tdb_store(tdb, key, dbuf, TDB_INSERT) == -1) {
+		terror("insert failed");
+	}
+}
+
+static void store_tdb(void)
+{
+	char *k = get_token(1);
+	char *d = get_token(0);
+	TDB_DATA key, dbuf;
+
+	if (!k || !d) {
+		help();
+		return;
+	}
+
+	key.dptr = k;
+	key.dsize = strlen(k)+1;
+	dbuf.dptr = d;
+	dbuf.dsize = strlen(d)+1;
+
+	printf("Storing key:\n");
+	print_rec(tdb, key, dbuf, NULL);
+
+	if (tdb_store(tdb, key, dbuf, TDB_REPLACE) == -1) {
+		terror("store failed");
+	}
+}
+
+static void show_tdb(void)
+{
+	char *k = get_token(1);
+	TDB_DATA key, dbuf;
+
+	if (!k) {
+		help();
+		return;
+	}
+
+	key.dptr = k;
+/*	key.dsize = strlen(k)+1;*/
+	key.dsize = strlen(k);
+
+	dbuf = tdb_fetch(tdb, key);
+	if (!dbuf.dptr) {
+		terror("fetch failed");
+		return;
+	}
+	/* printf("%s : %*.*s\n", k, (int)dbuf.dsize, (int)dbuf.dsize, dbuf.dptr); */
+	print_rec(tdb, key, dbuf, NULL);
+}
+
+static void delete_tdb(void)
+{
+	char *k = get_token(1);
+	TDB_DATA key;
+
+	if (!k) {
+		help();
+		return;
+	}
+
+	key.dptr = k;
+	key.dsize = strlen(k)+1;
+
+	if (tdb_delete(tdb, key) != 0) {
+		terror("delete failed");
+	}
+}
+
+#if 0
+static int print_conn_key(TDB_DATA key)
+{
+	printf( "pid    =%5d   ", ((connections_key*)key.dptr)->pid);
+	printf( "cnum   =%10d  ", ((connections_key*)key.dptr)->cnum);
+	printf( "name   =[%s]\n", ((connections_key*)key.dptr)->name);
+	return 0;
+}
+
+static int print_conn_data(TDB_DATA dbuf)
+{
+	printf( "pid    =%5d   ", ((connections_data*)dbuf.dptr)->pid);
+	printf( "cnum   =%10d  ", ((connections_data*)dbuf.dptr)->cnum);
+	printf( "name   =[%s]\n", ((connections_data*)dbuf.dptr)->name);
+	
+	printf( "uid    =%5d   ",  ((connections_data*)dbuf.dptr)->uid);
+	printf( "addr   =[%s]\n", ((connections_data*)dbuf.dptr)->addr);
+	printf( "gid    =%5d   ",  ((connections_data*)dbuf.dptr)->gid);
+	printf( "machine=[%s]\n", ((connections_data*)dbuf.dptr)->machine);
+	printf( "start  = %s\n",   ctime(&((connections_data*)dbuf.dptr)->start));
+	return 0;
+}
+#endif
+
+static int print_rec(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+#if 0
+	print_conn_key(key);
+	print_conn_data(dbuf);
+	return 0;
+#else
+	printf("\nkey %d bytes\n", key.dsize);
+	print_asc(key.dptr, key.dsize);
+	printf("\ndata %d bytes\n", dbuf.dsize);
+	print_data(dbuf.dptr, dbuf.dsize);
+	return 0;
+#endif
+}
+
+static int print_key(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	print_asc(key.dptr, key.dsize);
+	printf("\n");
+	return 0;
+}
+
+static int total_bytes;
+
+static int traverse_fn(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf, void *state)
+{
+	total_bytes += dbuf.dsize;
+	return 0;
+}
+
+static void info_tdb(void)
+{
+	int count;
+	total_bytes = 0;
+	if ((count = tdb_traverse(tdb, traverse_fn, NULL) == -1))
+		printf("Error = %s\n", tdb_errorstr(tdb));
+	else
+		printf("%d records totalling %d bytes\n", count, total_bytes);
+}
+
+static char *tdb_getline(char *prompt)
+{
+	static char line[1024];
+	char *p;
+	fputs(prompt, stdout);
+	line[0] = 0;
+	p = fgets(line, sizeof(line)-1, stdin);
+	if (p) p = strchr(p, '\n');
+	if (p) *p = 0;
+	return p?line:NULL;
+}
+
+static int do_delete_fn(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf,
+                     void *state)
+{
+    return tdb_delete(the_tdb, key);
+}
+
+static void first_record(TDB_CONTEXT *the_tdb, TDB_DATA *pkey)
+{
+	TDB_DATA dbuf;
+	*pkey = tdb_firstkey(the_tdb);
+	
+	dbuf = tdb_fetch(the_tdb, *pkey);
+	if (!dbuf.dptr) terror("fetch failed");
+	else {
+		/* printf("%s : %*.*s\n", k, (int)dbuf.dsize, (int)dbuf.dsize, dbuf.dptr); */
+		print_rec(the_tdb, *pkey, dbuf, NULL);
+	}
+}
+
+static void next_record(TDB_CONTEXT *the_tdb, TDB_DATA *pkey)
+{
+	TDB_DATA dbuf;
+	*pkey = tdb_nextkey(the_tdb, *pkey);
+	
+	dbuf = tdb_fetch(the_tdb, *pkey);
+	if (!dbuf.dptr) 
+		terror("fetch failed");
+	else
+		/* printf("%s : %*.*s\n", k, (int)dbuf.dsize, (int)dbuf.dsize, dbuf.dptr); */
+		print_rec(the_tdb, *pkey, dbuf, NULL);
+}
+
+int main(int argc, char *argv[])
+{
+    int bIterate = 0;
+    char *line;
+    char *tok;
+	TDB_DATA iterate_kbuf;
+
+    if (argv[1]) {
+	static char tmp[1024];
+        sprintf(tmp, "open %s", argv[1]);
+        tok=strtok(tmp," ");
+        open_tdb();
+    }
+
+    while ((line = tdb_getline("tdb> "))) {
+
+        /* Shell command */
+        
+        if (line[0] == '!') {
+            system(line + 1);
+            continue;
+        }
+        
+        if ((tok = strtok(line," "))==NULL) {
+           if (bIterate)
+              next_record(tdb, &iterate_kbuf);
+           continue;
+        }
+        if (strcmp(tok,"create") == 0) {
+            bIterate = 0;
+            create_tdb();
+            continue;
+        } else if (strcmp(tok,"open") == 0) {
+            open_tdb();
+            continue;
+        } else if ((strcmp(tok, "q") == 0) ||
+                   (strcmp(tok, "quit") == 0)) {
+            break;
+	}
+            
+        /* all the rest require a open database */
+        if (!tdb) {
+            bIterate = 0;
+            terror("database not open");
+            help();
+            continue;
+        }
+            
+        if (strcmp(tok,"insert") == 0) {
+            bIterate = 0;
+            insert_tdb();
+        } else if (strcmp(tok,"store") == 0) {
+            bIterate = 0;
+            store_tdb();
+        } else if (strcmp(tok,"show") == 0) {
+            bIterate = 0;
+            show_tdb();
+        } else if (strcmp(tok,"erase") == 0) {
+            bIterate = 0;
+            tdb_traverse(tdb, do_delete_fn, NULL);
+        } else if (strcmp(tok,"delete") == 0) {
+            bIterate = 0;
+            delete_tdb();
+        } else if (strcmp(tok,"dump") == 0) {
+            bIterate = 0;
+            tdb_traverse(tdb, print_rec, NULL);
+        } else if (strcmp(tok,"list") == 0) {
+            tdb_dump_all(tdb);
+        } else if (strcmp(tok, "free") == 0) {
+            tdb_printfreelist(tdb);
+        } else if (strcmp(tok,"info") == 0) {
+            info_tdb();
+        } else if ( (strcmp(tok, "1") == 0) ||
+                    (strcmp(tok, "first") == 0)) {
+            bIterate = 1;
+            first_record(tdb, &iterate_kbuf);
+        } else if ((strcmp(tok, "n") == 0) ||
+                   (strcmp(tok, "next") == 0)) {
+            next_record(tdb, &iterate_kbuf);
+        } else if ((strcmp(tok, "keys") == 0)) {
+                bIterate = 0;
+                tdb_traverse(tdb, print_key, NULL);
+        } else {
+            help();
+        }
+    }
+
+    if (tdb) tdb_close(tdb);
+
+    return 0;
+}
diff --git a/source4/tdb/tdbtorture.c b/source4/tdb/tdbtorture.c
new file mode 100644
index 0000000000..e27bbff990
--- /dev/null
+++ b/source4/tdb/tdbtorture.c
@@ -0,0 +1,226 @@
+#include <stdlib.h>
+#include <time.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <sys/wait.h>
+#include "tdb.h"
+
+/* this tests tdb by doing lots of ops from several simultaneous
+   writers - that stresses the locking code. Build with TDB_DEBUG=1
+   for best effect */
+
+
+
+#define REOPEN_PROB 30
+#define DELETE_PROB 8
+#define STORE_PROB 4
+#define APPEND_PROB 6
+#define LOCKSTORE_PROB 0
+#define TRAVERSE_PROB 20
+#define CULL_PROB 100
+#define KEYLEN 3
+#define DATALEN 100
+#define LOCKLEN 20
+
+static TDB_CONTEXT *db;
+
+static void tdb_log(TDB_CONTEXT *tdb, int level, const char *format, ...)
+{
+	va_list ap;
+    
+	va_start(ap, format);
+	vfprintf(stdout, format, ap);
+	va_end(ap);
+	fflush(stdout);
+#if 0
+	{
+		char *ptr;
+		asprintf(&ptr,"xterm -e gdb /proc/%d/exe %d", getpid(), getpid());
+		system(ptr);
+		free(ptr);
+	}
+#endif	
+}
+
+static void fatal(char *why)
+{
+	perror(why);
+	exit(1);
+}
+
+static char *randbuf(int len)
+{
+	char *buf;
+	int i;
+	buf = (char *)malloc(len+1);
+
+	for (i=0;i<len;i++) {
+		buf[i] = 'a' + (rand() % 26);
+	}
+	buf[i] = 0;
+	return buf;
+}
+
+static int cull_traverse(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf,
+			 void *state)
+{
+	if (random() % CULL_PROB == 0) {
+		tdb_delete(tdb, key);
+	}
+	return 0;
+}
+
+static void addrec_db(void)
+{
+	int klen, dlen, slen;
+	char *k, *d, *s;
+	TDB_DATA key, data, lockkey;
+
+	klen = 1 + (rand() % KEYLEN);
+	dlen = 1 + (rand() % DATALEN);
+	slen = 1 + (rand() % LOCKLEN);
+
+	k = randbuf(klen);
+	d = randbuf(dlen);
+	s = randbuf(slen);
+
+	key.dptr = k;
+	key.dsize = klen+1;
+
+	data.dptr = d;
+	data.dsize = dlen+1;
+
+	lockkey.dptr = s;
+	lockkey.dsize = slen+1;
+
+#if REOPEN_PROB
+	if (random() % REOPEN_PROB == 0) {
+		tdb_reopen_all();
+		goto next;
+	} 
+#endif
+
+#if DELETE_PROB
+	if (random() % DELETE_PROB == 0) {
+		tdb_delete(db, key);
+		goto next;
+	}
+#endif
+
+#if STORE_PROB
+	if (random() % STORE_PROB == 0) {
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+		goto next;
+	}
+#endif
+
+#if APPEND_PROB
+	if (random() % APPEND_PROB == 0) {
+		if (tdb_append(db, key, data) != 0) {
+			fatal("tdb_append failed");
+		}
+		goto next;
+	}
+#endif
+
+#if LOCKSTORE_PROB
+	if (random() % LOCKSTORE_PROB == 0) {
+		tdb_chainlock(db, lockkey);
+		data = tdb_fetch(db, key);
+		if (tdb_store(db, key, data, TDB_REPLACE) != 0) {
+			fatal("tdb_store failed");
+		}
+		if (data.dptr) free(data.dptr);
+		tdb_chainunlock(db, lockkey);
+		goto next;
+	} 
+#endif
+
+#if TRAVERSE_PROB
+	if (random() % TRAVERSE_PROB == 0) {
+		tdb_traverse(db, cull_traverse, NULL);
+		goto next;
+	}
+#endif
+
+	data = tdb_fetch(db, key);
+	if (data.dptr) free(data.dptr);
+
+next:
+	free(k);
+	free(d);
+	free(s);
+}
+
+static int traverse_fn(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf,
+                       void *state)
+{
+	tdb_delete(tdb, key);
+	return 0;
+}
+
+#ifndef NPROC
+#define NPROC 6
+#endif
+
+#ifndef NLOOPS
+#define NLOOPS 200000
+#endif
+
+int main(int argc, char *argv[])
+{
+	int i, seed=0;
+	int loops = NLOOPS;
+	pid_t pids[NPROC];
+
+	pids[0] = getpid();
+
+	for (i=0;i<NPROC-1;i++) {
+		if ((pids[i+1]=fork()) == 0) break;
+	}
+
+	db = tdb_open("torture.tdb", 2, TDB_CLEAR_IF_FIRST, 
+		      O_RDWR | O_CREAT, 0600);
+	if (!db) {
+		fatal("db open failed");
+	}
+	tdb_logging_function(db, tdb_log);
+
+	srand(seed + getpid());
+	srandom(seed + getpid() + time(NULL));
+	for (i=0;i<loops;i++) addrec_db();
+
+	tdb_traverse(db, NULL, NULL);
+	tdb_traverse(db, traverse_fn, NULL);
+	tdb_traverse(db, traverse_fn, NULL);
+
+	tdb_close(db);
+
+	if (getpid() == pids[0]) {
+		for (i=0;i<NPROC-1;i++) {
+			int status;
+			if (waitpid(pids[i+1], &status, 0) != pids[i+1]) {
+				printf("failed to wait for %d\n",
+				       (int)pids[i+1]);
+				exit(1);
+			}
+			if (WEXITSTATUS(status) != 0) {
+				printf("child %d exited with status %d\n",
+				       (int)pids[i+1], WEXITSTATUS(status));
+				exit(1);
+			}
+		}
+		printf("OK\n");
+	}
+
+	return 0;
+}
diff --git a/source4/tdb/tdbutil.c b/source4/tdb/tdbutil.c
new file mode 100644
index 0000000000..0d8f6128cc
--- /dev/null
+++ b/source4/tdb/tdbutil.c
@@ -0,0 +1,687 @@
+/* 
+   Unix SMB/CIFS implementation.
+   tdb utility functions
+   Copyright (C) Andrew Tridgell 1992-1998
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include "includes.h"
+#include <fnmatch.h>
+
+/* these are little tdb utility functions that are meant to make
+   dealing with a tdb database a little less cumbersome in Samba */
+
+static SIG_ATOMIC_T gotalarm;
+
+/***************************************************************
+ Signal function to tell us we timed out.
+****************************************************************/
+
+static void gotalarm_sig(void)
+{
+	gotalarm = 1;
+}
+
+/***************************************************************
+ Make a TDB_DATA and keep the const warning in one place
+****************************************************************/
+
+static TDB_DATA make_tdb_data(const char *dptr, size_t dsize)
+{
+	TDB_DATA ret;
+	ret.dptr = dptr;
+	ret.dsize = dsize;
+	return ret;
+}
+
+/****************************************************************************
+ Lock a chain with timeout (in seconds).
+****************************************************************************/
+
+static int tdb_chainlock_with_timeout_internal( TDB_CONTEXT *tdb, TDB_DATA key, unsigned int timeout, int rw_type)
+{
+	/* Allow tdb_chainlock to be interrupted by an alarm. */
+	int ret;
+	gotalarm = 0;
+	tdb_set_lock_alarm(&gotalarm);
+
+	if (timeout) {
+		CatchSignal(SIGALRM, SIGNAL_CAST gotalarm_sig);
+		alarm(timeout);
+	}
+
+	if (rw_type == F_RDLCK)
+		ret = tdb_chainlock_read(tdb, key);
+	else
+		ret = tdb_chainlock(tdb, key);
+
+	if (timeout) {
+		alarm(0);
+		CatchSignal(SIGALRM, SIGNAL_CAST SIG_IGN);
+		if (gotalarm) {
+			DEBUG(0,("tdb_chainlock_with_timeout_internal: alarm (%u) timed out for key %s in tdb %s\n",
+				timeout, key.dptr, tdb->name ));
+			/* TODO: If we time out waiting for a lock, it might
+			 * be nice to use F_GETLK to get the pid of the
+			 * process currently holding the lock and print that
+			 * as part of the debugging message. -- mbp */
+			return -1;
+		}
+	}
+
+	return ret;
+}
+
+/****************************************************************************
+ Write lock a chain. Return -1 if timeout or lock failed.
+****************************************************************************/
+
+int tdb_chainlock_with_timeout( TDB_CONTEXT *tdb, TDB_DATA key, unsigned int timeout)
+{
+	return tdb_chainlock_with_timeout_internal(tdb, key, timeout, F_WRLCK);
+}
+
+/****************************************************************************
+ Lock a chain by string. Return -1 if timeout or lock failed.
+****************************************************************************/
+
+int tdb_lock_bystring(TDB_CONTEXT *tdb, const char *keyval, unsigned int timeout)
+{
+	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
+	
+	return tdb_chainlock_with_timeout_internal(tdb, key, timeout, F_WRLCK);
+}
+
+/****************************************************************************
+ Unlock a chain by string.
+****************************************************************************/
+
+void tdb_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+{
+	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
+
+	tdb_chainunlock(tdb, key);
+}
+
+/****************************************************************************
+ Read lock a chain by string. Return -1 if timeout or lock failed.
+****************************************************************************/
+
+int tdb_read_lock_bystring(TDB_CONTEXT *tdb, const char *keyval, unsigned int timeout)
+{
+	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
+	
+	return tdb_chainlock_with_timeout_internal(tdb, key, timeout, F_RDLCK);
+}
+
+/****************************************************************************
+ Read unlock a chain by string.
+****************************************************************************/
+
+void tdb_read_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+{
+	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
+	
+	tdb_chainunlock_read(tdb, key);
+}
+
+
+/****************************************************************************
+ Fetch a int32 value by a arbitrary blob key, return -1 if not found.
+ Output is int32 in native byte order.
+****************************************************************************/
+
+int32 tdb_fetch_int32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len)
+{
+	TDB_DATA key = make_tdb_data(keyval, len);
+	TDB_DATA data;
+	int32 ret;
+
+	data = tdb_fetch(tdb, key);
+	if (!data.dptr || data.dsize != sizeof(int32)) {
+		SAFE_FREE(data.dptr);
+		return -1;
+	}
+
+	ret = IVAL(data.dptr,0);
+	SAFE_FREE(data.dptr);
+	return ret;
+}
+
+/****************************************************************************
+ Fetch a int32 value by string key, return -1 if not found.
+ Output is int32 in native byte order.
+****************************************************************************/
+
+int32 tdb_fetch_int32(TDB_CONTEXT *tdb, const char *keystr)
+{
+	return tdb_fetch_int32_byblob(tdb, keystr, strlen(keystr) + 1);
+}
+
+/****************************************************************************
+ Store a int32 value by an arbitary blob key, return 0 on success, -1 on failure.
+ Input is int32 in native byte order. Output in tdb is in little-endian.
+****************************************************************************/
+
+int tdb_store_int32_byblob(TDB_CONTEXT *tdb, const char *keystr, size_t len, int32 v)
+{
+	TDB_DATA key = make_tdb_data(keystr, len);
+	TDB_DATA data;
+	int32 v_store;
+
+	SIVAL(&v_store,0,v);
+	data.dptr = (void *)&v_store;
+	data.dsize = sizeof(int32);
+
+	return tdb_store(tdb, key, data, TDB_REPLACE);
+}
+
+/****************************************************************************
+ Store a int32 value by string key, return 0 on success, -1 on failure.
+ Input is int32 in native byte order. Output in tdb is in little-endian.
+****************************************************************************/
+
+int tdb_store_int32(TDB_CONTEXT *tdb, const char *keystr, int32 v)
+{
+	return tdb_store_int32_byblob(tdb, keystr, strlen(keystr) + 1, v);
+}
+
+/****************************************************************************
+ Fetch a uint32 value by a arbitrary blob key, return -1 if not found.
+ Output is uint32 in native byte order.
+****************************************************************************/
+
+BOOL tdb_fetch_uint32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len, uint32 *value)
+{
+	TDB_DATA key = make_tdb_data(keyval, len);
+	TDB_DATA data;
+
+	data = tdb_fetch(tdb, key);
+	if (!data.dptr || data.dsize != sizeof(uint32)) {
+		SAFE_FREE(data.dptr);
+		return False;
+	}
+
+	*value = IVAL(data.dptr,0);
+	SAFE_FREE(data.dptr);
+	return True;
+}
+
+/****************************************************************************
+ Fetch a uint32 value by string key, return -1 if not found.
+ Output is uint32 in native byte order.
+****************************************************************************/
+
+BOOL tdb_fetch_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32 *value)
+{
+	return tdb_fetch_uint32_byblob(tdb, keystr, strlen(keystr) + 1, value);
+}
+
+/****************************************************************************
+ Store a uint32 value by an arbitary blob key, return 0 on success, -1 on failure.
+ Input is uint32 in native byte order. Output in tdb is in little-endian.
+****************************************************************************/
+
+BOOL tdb_store_uint32_byblob(TDB_CONTEXT *tdb, const char *keystr, size_t len, uint32 value)
+{
+	TDB_DATA key = make_tdb_data(keystr, len);
+	TDB_DATA data;
+	uint32 v_store;
+	BOOL ret = True;
+
+	SIVAL(&v_store, 0, value);
+	data.dptr = (void *)&v_store;
+	data.dsize = sizeof(uint32);
+
+	if (tdb_store(tdb, key, data, TDB_REPLACE) == -1)
+		ret = False;
+
+	return ret;
+}
+
+/****************************************************************************
+ Store a uint32 value by string key, return 0 on success, -1 on failure.
+ Input is uint32 in native byte order. Output in tdb is in little-endian.
+****************************************************************************/
+
+BOOL tdb_store_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32 value)
+{
+	return tdb_store_uint32_byblob(tdb, keystr, strlen(keystr) + 1, value);
+}
+/****************************************************************************
+ Store a buffer by a null terminated string key.  Return 0 on success, -1
+ on failure.
+****************************************************************************/
+
+int tdb_store_by_string(TDB_CONTEXT *tdb, const char *keystr, TDB_DATA data, int flags)
+{
+	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
+	
+	return tdb_store(tdb, key, data, flags);
+}
+
+/****************************************************************************
+ Fetch a buffer using a null terminated string key.  Don't forget to call
+ free() on the result dptr.
+****************************************************************************/
+
+TDB_DATA tdb_fetch_by_string(TDB_CONTEXT *tdb, const char *keystr)
+{
+	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
+
+	return tdb_fetch(tdb, key);
+}
+
+/****************************************************************************
+ Delete an entry using a null terminated string key. 
+****************************************************************************/
+
+int tdb_delete_by_string(TDB_CONTEXT *tdb, const char *keystr)
+{
+	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
+
+	return tdb_delete(tdb, key);
+}
+
+/****************************************************************************
+ Atomic integer change. Returns old value. To create, set initial value in *oldval. 
+****************************************************************************/
+
+int32 tdb_change_int32_atomic(TDB_CONTEXT *tdb, const char *keystr, int32 *oldval, int32 change_val)
+{
+	int32 val;
+	int32 ret = -1;
+
+	if (tdb_lock_bystring(tdb, keystr,0) == -1)
+		return -1;
+
+	if ((val = tdb_fetch_int32(tdb, keystr)) == -1) {
+		/* The lookup failed */
+		if (tdb_error(tdb) != TDB_ERR_NOEXIST) {
+			/* but not becouse it didn't exist */
+			goto err_out;
+		}
+		
+		/* Start with 'old' value */
+		val = *oldval;
+
+	} else {
+		/* It worked, set return value (oldval) to tdb data */
+		*oldval = val;
+	}
+
+	/* Increment value for storage and return next time */
+	val += change_val;
+		
+	if (tdb_store_int32(tdb, keystr, val) == -1)
+		goto err_out;
+
+	ret = 0;
+
+  err_out:
+
+	tdb_unlock_bystring(tdb, keystr);
+	return ret;
+}
+
+/****************************************************************************
+ Atomic unsigned integer change. Returns old value. To create, set initial value in *oldval. 
+****************************************************************************/
+
+BOOL tdb_change_uint32_atomic(TDB_CONTEXT *tdb, const char *keystr, uint32 *oldval, uint32 change_val)
+{
+	uint32 val;
+	BOOL ret = False;
+
+	if (tdb_lock_bystring(tdb, keystr,0) == -1)
+		return False;
+
+	if (!tdb_fetch_uint32(tdb, keystr, &val)) {
+		/* It failed */
+		if (tdb_error(tdb) != TDB_ERR_NOEXIST) { 
+			/* and not becouse it didn't exist */
+			goto err_out;
+		}
+
+		/* Start with 'old' value */
+		val = *oldval;
+
+	} else {
+		/* it worked, set return value (oldval) to tdb data */
+		*oldval = val;
+
+	}
+
+	/* get a new value to store */
+	val += change_val;
+		
+	if (!tdb_store_uint32(tdb, keystr, val))
+		goto err_out;
+
+	ret = True;
+
+  err_out:
+
+	tdb_unlock_bystring(tdb, keystr);
+	return ret;
+}
+
+/****************************************************************************
+ Useful pair of routines for packing/unpacking data consisting of
+ integers and strings.
+****************************************************************************/
+
+size_t tdb_pack(char *buf, int bufsize, const char *fmt, ...)
+{
+	va_list ap;
+	uint16 w;
+	uint32 d;
+	int i;
+	void *p;
+	int len;
+	char *s;
+	char c;
+	char *buf0 = buf;
+	const char *fmt0 = fmt;
+	int bufsize0 = bufsize;
+
+	va_start(ap, fmt);
+
+	while (*fmt) {
+		switch ((c = *fmt++)) {
+		case 'w':
+			len = 2;
+			w = (uint16)va_arg(ap, int);
+			if (bufsize >= len)
+				SSVAL(buf, 0, w);
+			break;
+		case 'd':
+			len = 4;
+			d = va_arg(ap, uint32);
+			if (bufsize >= len)
+				SIVAL(buf, 0, d);
+			break;
+		case 'p':
+			len = 4;
+			p = va_arg(ap, void *);
+			d = p?1:0;
+			if (bufsize >= len)
+				SIVAL(buf, 0, d);
+			break;
+		case 'P':
+			s = va_arg(ap,char *);
+			w = strlen(s);
+			len = w + 1;
+			if (bufsize >= len)
+				memcpy(buf, s, len);
+			break;
+		case 'f':
+			s = va_arg(ap,char *);
+			w = strlen(s);
+			len = w + 1;
+			if (bufsize >= len)
+				memcpy(buf, s, len);
+			break;
+		case 'B':
+			i = va_arg(ap, int);
+			s = va_arg(ap, char *);
+			len = 4+i;
+			if (bufsize >= len) {
+				SIVAL(buf, 0, i);
+				memcpy(buf+4, s, i);
+			}
+			break;
+		default:
+			DEBUG(0,("Unknown tdb_pack format %c in %s\n", 
+				 c, fmt));
+			len = 0;
+			break;
+		}
+
+		buf += len;
+		bufsize -= len;
+	}
+
+	va_end(ap);
+
+	DEBUG(18,("tdb_pack(%s, %d) -> %d\n", 
+		 fmt0, bufsize0, (int)PTR_DIFF(buf, buf0)));
+	
+	return PTR_DIFF(buf, buf0);
+}
+
+/****************************************************************************
+ Useful pair of routines for packing/unpacking data consisting of
+ integers and strings.
+****************************************************************************/
+
+int tdb_unpack(char *buf, int bufsize, const char *fmt, ...)
+{
+	va_list ap;
+	uint16 *w;
+	uint32 *d;
+	int len;
+	int *i;
+	void **p;
+	char *s, **b;
+	char c;
+	char *buf0 = buf;
+	const char *fmt0 = fmt;
+	int bufsize0 = bufsize;
+
+	va_start(ap, fmt);
+	
+	while (*fmt) {
+		switch ((c=*fmt++)) {
+		case 'w':
+			len = 2;
+			w = va_arg(ap, uint16 *);
+			if (bufsize < len)
+				goto no_space;
+			*w = SVAL(buf, 0);
+			break;
+		case 'd':
+			len = 4;
+			d = va_arg(ap, uint32 *);
+			if (bufsize < len)
+				goto no_space;
+			*d = IVAL(buf, 0);
+			break;
+		case 'p':
+			len = 4;
+			p = va_arg(ap, void **);
+			if (bufsize < len)
+				goto no_space;
+			*p = (void *)IVAL(buf, 0);
+			break;
+		case 'P':
+			s = va_arg(ap,char *);
+			len = strlen(buf) + 1;
+			if (bufsize < len || len > sizeof(pstring))
+				goto no_space;
+			memcpy(s, buf, len);
+			break;
+		case 'f':
+			s = va_arg(ap,char *);
+			len = strlen(buf) + 1;
+			if (bufsize < len || len > sizeof(fstring))
+				goto no_space;
+			memcpy(s, buf, len);
+			break;
+		case 'B':
+			i = va_arg(ap, int *);
+			b = va_arg(ap, char **);
+			len = 4;
+			if (bufsize < len)
+				goto no_space;
+			*i = IVAL(buf, 0);
+			if (! *i) {
+				*b = NULL;
+				break;
+			}
+			len += *i;
+			if (bufsize < len)
+				goto no_space;
+			*b = (char *)malloc(*i);
+			if (! *b)
+				goto no_space;
+			memcpy(*b, buf+4, *i);
+			break;
+		default:
+			DEBUG(0,("Unknown tdb_unpack format %c in %s\n", 
+				 c, fmt));
+
+			len = 0;
+			break;
+		}
+
+		buf += len;
+		bufsize -= len;
+	}
+
+	va_end(ap);
+
+	DEBUG(18,("tdb_unpack(%s, %d) -> %d\n", 
+		 fmt0, bufsize0, (int)PTR_DIFF(buf, buf0)));
+
+	return PTR_DIFF(buf, buf0);
+
+ no_space:
+	return -1;
+}
+
+/****************************************************************************
+ Log tdb messages via DEBUG().
+****************************************************************************/
+
+static void tdb_log(TDB_CONTEXT *tdb, int level, const char *format, ...)
+{
+	va_list ap;
+	char *ptr = NULL;
+
+	va_start(ap, format);
+	vasprintf(&ptr, format, ap);
+	va_end(ap);
+	
+	if (!ptr || !*ptr)
+		return;
+
+	DEBUG(level, ("tdb(%s): %s", tdb->name ? tdb->name : "unnamed", ptr));
+	SAFE_FREE(ptr);
+}
+
+/****************************************************************************
+ Like tdb_open() but also setup a logging function that redirects to
+ the samba DEBUG() system.
+****************************************************************************/
+
+TDB_CONTEXT *tdb_open_log(const char *name, int hash_size, int tdb_flags,
+			  int open_flags, mode_t mode)
+{
+	TDB_CONTEXT *tdb;
+
+	if (!lp_use_mmap())
+		tdb_flags |= TDB_NOMMAP;
+
+	tdb = tdb_open_ex(name, hash_size, tdb_flags, 
+				    open_flags, mode, tdb_log);
+	if (!tdb)
+		return NULL;
+
+	return tdb;
+}
+
+
+/****************************************************************************
+ Allow tdb_delete to be used as a tdb_traversal_fn.
+****************************************************************************/
+
+int tdb_traverse_delete_fn(TDB_CONTEXT *the_tdb, TDB_DATA key, TDB_DATA dbuf,
+                     void *state)
+{
+    return tdb_delete(the_tdb, key);
+}
+
+
+
+/**
+ * Search across the whole tdb for keys that match the given pattern
+ * return the result as a list of keys
+ *
+ * @param tdb pointer to opened tdb file context
+ * @param pattern searching pattern used by fnmatch(3) functions
+ *
+ * @return list of keys found by looking up with given pattern
+ **/
+TDB_LIST_NODE *tdb_search_keys(TDB_CONTEXT *tdb, const char* pattern)
+{
+	TDB_DATA key, next;
+	TDB_LIST_NODE *list = NULL;
+	TDB_LIST_NODE *rec = NULL;
+	TDB_LIST_NODE *tmp = NULL;
+	
+	for (key = tdb_firstkey(tdb); key.dptr; key = next) {
+		/* duplicate key string to ensure null-termination */
+		char *key_str = (char*) strndup(key.dptr, key.dsize);
+		if (!key_str) {
+			DEBUG(0, ("tdb_search_keys: strndup() failed!\n"));
+			smb_panic("strndup failed!\n");
+		}
+		
+		DEBUG(18, ("checking %s for match to pattern %s\n", key_str, pattern));
+		
+		next = tdb_nextkey(tdb, key);
+
+		/* do the pattern checking */
+		if (fnmatch(pattern, key_str, 0) == 0) {
+			rec = (TDB_LIST_NODE*) malloc(sizeof(*rec));
+			ZERO_STRUCTP(rec);
+
+			rec->node_key = key;
+	
+			DLIST_ADD_END(list, rec, tmp);
+		
+			DEBUG(18, ("checking %s matched pattern %s\n", key_str, pattern));
+		} else {
+			free(key.dptr);
+		}
+		
+		/* free duplicated key string */
+		free(key_str);
+	}
+	
+	return list;
+
+};
+
+
+/**
+ * Free the list returned by tdb_search_keys
+ *
+ * @param node list of results found by tdb_search_keys
+ **/
+void tdb_search_list_free(TDB_LIST_NODE* node)
+{
+	TDB_LIST_NODE *next_node;
+	
+	while (node) {
+		next_node = node->next;
+		SAFE_FREE(node);
+		node = next_node;
+	};
+};
+
+
diff --git a/source4/tdb/tdbutil.h b/source4/tdb/tdbutil.h
new file mode 100644
index 0000000000..01473446a1
--- /dev/null
+++ b/source4/tdb/tdbutil.h
@@ -0,0 +1,37 @@
+/* 
+   Unix SMB/CIFS implementation.
+   tdb utility functions
+   Copyright (C) Andrew Tridgell 1999
+   
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#ifndef __TDBUTIL_H__
+#define __TDBUTIL_H__
+
+
+/* single node of a list returned by tdb_search_keys */
+typedef struct keys_node 
+{
+	struct keys_node *prev, *next;
+	TDB_DATA node_key;
+} TDB_LIST_NODE;
+
+
+TDB_LIST_NODE *tdb_search_keys(TDB_CONTEXT*, const char*);
+void tdb_search_list_free(TDB_LIST_NODE*);
+
+
+#endif /* __TDBUTIL_H__ */