r10253: a fairly large tdb cleanup and re-organise. Nearly all of this change

just involves splitting up the core tdb.c code into separate files on
logical boundaries, but there are some minor functional changes as well:

 - move the 'struct tdb_context' into tdb_private.h, hiding it from
   users. This was done to allow the structure to change without
   breaking code that uses tdb.

 - added accessor functions tdb_fd(), tdb_name(), and tdb_log_fn() to
   access the elements of struct tdb_context that were used by
   external code but are no longer visible

 - simplied tdb_append() to use tdb_fetch()/tdb_store(), which is just
   as good due to the way tdb locks work

 - changed some of the types (such as tdb_off to tdb_off_t) to make
   syntax highlighting work better

 - removed the old optional spinlock code. It was a bad idea.

 - fixed a bug in tdb_reopen_all() that caused tdbtorture to sometimes
   fail or report nasty looking errors. This is the only real bug
   fixed in this commit. Jeremy/Jerry, you might like to pickup this
   change for Samba3, as that could definately affect smbd in
   Samba3.

The aim of all of these changes is to make the tdb
transactions/journaling code I am working on easier to write. I
started to write it on top of the existing tdb.c code and it got very
messy. Splitting up the code makes it much easier to follow.

There are more cleanups we could do in tdb, such as using uint32_t
instead of u32 (suggested by metze). I'll leave those for another day.
(This used to be commit 4673cdd0d261614e707b72a7a348bb0e7dbb2482)

11 files changed, 2128 insertions, 2395 deletions

diff --git a/source4/lib/tdb/common/dump.c b/source4/lib/tdb/common/dump.c
new file mode 100644
index 0000000000..0e203cc0d8
--- /dev/null
+++ b/source4/lib/tdb/common/dump.c
@@ -0,0 +1,133 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+static tdb_off_t tdb_dump_record(struct tdb_context *tdb, tdb_off_t offset)
+{
+	struct list_struct rec;
+	tdb_off_t 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=0x%08x next=0x%08x 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_t);
+
+	if (tdb_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 int)tailer, (unsigned int)(rec.rec_len + sizeof(rec)));
+	}
+	return rec.next;
+}
+
+static int tdb_dump_chain(struct tdb_context *tdb, int i)
+{
+	tdb_off_t rec_ptr, top;
+
+	top = TDB_HASH_TOP(i);
+
+	if (tdb_lock(tdb, i, F_WRLCK) != 0)
+		return -1;
+
+	if (tdb_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(struct 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(struct tdb_context *tdb)
+{
+	int ret;
+	long total_free = 0;
+	tdb_off_t 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 (tdb_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)] (end = 0x%08x)\n", 
+		       rec_ptr, rec.rec_len, rec.rec_len, rec_ptr + 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);
+}
+
diff --git a/source4/lib/tdb/common/error.c b/source4/lib/tdb/common/error.c
new file mode 100644
index 0000000000..5b7696877c
--- /dev/null
+++ b/source4/lib/tdb/common/error.c
@@ -0,0 +1,56 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+enum TDB_ERROR tdb_error(struct 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(struct 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";
+}
+
diff --git a/source4/lib/tdb/common/freelist.c b/source4/lib/tdb/common/freelist.c
new file mode 100644
index 0000000000..018b7bc298
--- /dev/null
+++ b/source4/lib/tdb/common/freelist.c
@@ -0,0 +1,323 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+/* read a freelist record and check for simple errors */
+static int rec_free_read(struct tdb_context *tdb, tdb_off_t 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 0x%x 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;
+}
+
+
+
+/* Remove an element from the freelist.  Must have alloc lock. */
+static int remove_from_freelist(struct tdb_context *tdb, tdb_off_t off, tdb_off_t next)
+{
+	tdb_off_t last_ptr, i;
+
+	/* read in the freelist top */
+	last_ptr = FREELIST_TOP;
+	while (tdb_ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
+		if (i == off) {
+			/* We've found it! */
+			return tdb_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);
+}
+
+
+/* update a record tailer (must hold allocation lock) */
+static int update_tailer(struct tdb_context *tdb, tdb_off_t offset,
+			 const struct list_struct *rec)
+{
+	tdb_off_t totalsize;
+
+	/* Offset of tailer from record header */
+	totalsize = sizeof(*rec) + rec->rec_len;
+	return tdb_ofs_write(tdb, offset + totalsize - sizeof(tdb_off_t),
+			 &totalsize);
+}
+
+/* Add an element into the freelist. Merge adjacent records if
+   neccessary. */
+int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
+{
+	tdb_off_t 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_t);
+	if (left > TDB_DATA_START(tdb->header.hash_size)) {
+		struct list_struct l;
+		tdb_off_t leftsize;
+		
+		/* Read in tailer and jump back to header */
+		if (tdb_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 (tdb_ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
+	    tdb_rec_write(tdb, offset, rec) == -1 ||
+	    tdb_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;
+}
+
+
+/* 
+   the core of tdb_allocate - called when we have decided which
+   free list entry to use
+ */
+static tdb_off_t tdb_allocate_ofs(struct tdb_context *tdb, tdb_len_t length, tdb_off_t rec_ptr,
+				struct list_struct *rec, tdb_off_t last_ptr)
+{
+	struct list_struct newrec;
+	tdb_off_t newrec_ptr;
+
+	memset(&newrec, '\0', sizeof(newrec));
+
+	/* 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 (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1) {
+		return 0;
+	}
+	
+	/* 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 (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
+		return 0;
+	}
+	
+	/* Did we create new block? */
+	if (newrec_ptr) {
+		/* Update allocated record tailer (we
+		   shortened it). */
+		if (update_tailer(tdb, rec_ptr, rec) == -1) {
+			return 0;
+		}
+		
+		/* Free new record */
+		if (tdb_free(tdb, newrec_ptr, &newrec) == -1) {
+			return 0;
+		}
+	}
+	
+	/* all done - return the new record offset */
+	return rec_ptr;
+}
+
+/* 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
+ */
+tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct list_struct *rec)
+{
+	tdb_off_t rec_ptr, last_ptr, newrec_ptr;
+	struct {
+		tdb_off_t rec_ptr, last_ptr;
+		tdb_len_t rec_len;
+	} bestfit;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1)
+		return 0;
+
+	/* Extra bytes required for tailer */
+	length += sizeof(tdb_off_t);
+
+ again:
+	last_ptr = FREELIST_TOP;
+
+	/* read in the freelist top */
+	if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
+		goto fail;
+
+	bestfit.rec_ptr = 0;
+
+	/* 
+	   this is a best fit allocation strategy. Originally we used
+	   a first fit strategy, but it suffered from massive fragmentation
+	   issues when faced with a slowly increasing record size.
+	 */
+	while (rec_ptr) {
+		if (rec_free_read(tdb, rec_ptr, rec) == -1) {
+			goto fail;
+		}
+
+		if (rec->rec_len >= length) {
+			if (bestfit.rec_ptr == 0 ||
+			    rec->rec_len < bestfit.rec_len) {
+				bestfit.rec_len = rec->rec_len;
+				bestfit.rec_ptr = rec_ptr;
+				bestfit.last_ptr = last_ptr;
+				/* consider a fit to be good enough if
+				   we aren't wasting more than half
+				   the space */
+				if (bestfit.rec_len < 2*length) {
+					break;
+				}
+			}
+		}
+
+		/* move to the next record */
+		last_ptr = rec_ptr;
+		rec_ptr = rec->next;
+	}
+
+	if (bestfit.rec_ptr != 0) {
+		if (rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
+			goto fail;
+		}
+
+		newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, rec, bestfit.last_ptr);
+		tdb_unlock(tdb, -1, F_WRLCK);
+		return newrec_ptr;
+	}
+
+	/* 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;
+}
+
diff --git a/source4/lib/tdb/common/io.c b/source4/lib/tdb/common/io.c
new file mode 100644
index 0000000000..dbcc83119c
--- /dev/null
+++ b/source4/lib/tdb/common/io.c
@@ -0,0 +1,374 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+
+#include "tdb_private.h"
+
+#ifndef HAVE_PREAD
+ static ssize_t pread(int fd, void *buf, size_t count, off_t offset)
+{
+	if (lseek(fd, offset, SEEK_SET) != offset) {
+		errno = EIO;
+		return -1;
+	}
+	return read(fd, buf, count);
+}
+#endif
+
+#ifndef HAVE_PWRITE
+ static ssize_t pwrite(int fd, const void *buf, size_t count, off_t offset)
+{
+	if (lseek(fd, offset, SEEK_SET) != offset) {
+		errno = EIO;
+		return -1;
+	}
+	return write(fd, buf, count);
+}
+#endif
+
+/* 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
+*/
+int tdb_oob(struct tdb_context *tdb, tdb_off_t 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 */
+int tdb_write(struct tdb_context *tdb, tdb_off_t off, void *buf, tdb_len_t len)
+{
+	if (tdb_oob(tdb, off + len, 0) != 0)
+		return -1;
+
+	if (tdb->map_ptr) {
+		memcpy(off + (char *)tdb->map_ptr, buf, len);
+	} else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
+		/* 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;
+}
+
+/* Endian conversion: we only ever deal with 4 byte quantities */
+void *tdb_convert(void *buf, u32 size)
+{
+	u32 i, *p = buf;
+	for (i = 0; i < size / 4; i++)
+		p[i] = TDB_BYTEREV(p[i]);
+	return buf;
+}
+
+
+/* read a lump of data at a specified offset, maybe convert */
+int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf, tdb_len_t len, int cv)
+{
+	ssize_t ret;
+
+	if (tdb_oob(tdb, off + len, 0) != 0) {
+		return -1;
+	}
+
+	if (tdb->map_ptr) {
+		memcpy(buf, off + (char *)tdb->map_ptr, len);
+	} else {
+		ssize_t ret = pread(tdb->fd, buf, len, off);
+		if (ret != (ssize_t)len) {
+			/* Ensure ecode is set for log fn. */
+			tdb->ecode = TDB_ERR_IO;
+			TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d ret=%d (%s) map_size=%d\n",
+				 off, len, ret, strerror(errno), tdb->map_size));
+			return TDB_ERRCODE(TDB_ERR_IO, -1);
+		}
+	}
+	if (cv)
+		tdb_convert(buf, len);
+	return 0;
+}
+
+
+
+/*
+  do an unlocked scan of the hash table heads to find the next non-zero head. The value
+  will then be confirmed with the lock held
+*/		
+void tdb_next_hash_chain(struct tdb_context *tdb, u32 *chain)
+{
+	u32 h = *chain;
+	if (tdb->map_ptr) {
+		for (;h < tdb->header.hash_size;h++) {
+			if (0 != *(u32 *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
+				break;
+			}
+		}
+	} else {
+		u32 off;
+		for (;h < tdb->header.hash_size;h++) {
+			if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
+				break;
+			}
+		}
+	}
+	(*chain) = h;
+}
+
+
+int tdb_munmap(struct 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;
+}
+
+void tdb_mmap(struct 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
+}
+
+/* expand a file.  we prefer to use ftruncate, as that is what posix
+  says to use for mmap expansion */
+static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
+{
+	char buf[1024];
+	if (ftruncate(tdb->fd, size+addition) == -1) {
+		char b = 0;
+		if (pwrite(tdb->fd,  &b, 1, (size+addition) - 1) != 1) {
+			TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n", 
+				 size+addition, strerror(errno)));
+			return -1;
+		}
+	}
+
+	/* 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;
+		int ret = pwrite(tdb->fd, buf, n, size);
+		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 */
+int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
+{
+	struct list_struct rec;
+	tdb_off_t 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 (tdb_expand_file(tdb, tdb->map_size, size) != 0)
+			goto fail;
+	}
+
+	tdb->map_size += size;
+
+	if (tdb->flags & TDB_INTERNAL) {
+		char *new_map_ptr = realloc(tdb->map_ptr, tdb->map_size);
+		if (!new_map_ptr) {
+			tdb->map_size -= size;
+			goto fail;
+		}
+		tdb->map_ptr = new_map_ptr;
+	} 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;
+}
+
+/* read/write a tdb_off_t */
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
+{
+	return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
+}
+
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
+{
+	tdb_off_t off = *d;
+	return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
+}
+
+
+/* read a lump of data, allocating the space for it */
+unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
+{
+	unsigned 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 record */
+int tdb_rec_read(struct tdb_context *tdb, tdb_off_t 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,"tdb_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);
+}
+
+int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
+{
+	struct list_struct r = *rec;
+	return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
+}
+
diff --git a/source4/lib/tdb/common/lock.c b/source4/lib/tdb/common/lock.c
new file mode 100644
index 0000000000..596f7bfc95
--- /dev/null
+++ b/source4/lib/tdb/common/lock.c
@@ -0,0 +1,235 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+/* 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(). */
+int tdb_brlock(struct tdb_context *tdb, tdb_off_t 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);
+	} while (ret == -1 && errno == EINTR);
+
+	if (ret == -1) {
+		if (!probe && lck_type != F_SETLK) {
+			/* Ensure error code is set for log fun to examine. */
+			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));
+		}
+		/* Generic lock error. errno set by fcntl.
+		 * EAGAIN is an expected return from non-blocking
+		 * locks. */
+		if (errno != EAGAIN) {
+		TDB_LOG((tdb, 5, "tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d: %s\n", 
+				 tdb->fd, offset, rw_type, lck_type, 
+				 strerror(errno)));
+		}
+		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
+	}
+	return 0;
+}
+
+/* lock a list in the database. list -1 is the alloc list */
+int tdb_lock(struct 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_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 */
+int tdb_unlock(struct 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 */
+		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;
+}
+
+
+
+/* lock/unlock entire database */
+int tdb_lockall(struct tdb_context *tdb)
+{
+	u32 i;
+
+	/* There are no locks on read-only dbs */
+	if (tdb->read_only)
+		return TDB_ERRCODE(TDB_ERR_LOCK, -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(struct tdb_context *tdb)
+{
+	u32 i;
+	for (i=0; i < tdb->header.hash_size; i++)
+		tdb_unlock(tdb, i, F_WRLCK);
+}
+
+/* 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(struct tdb_context *tdb, TDB_DATA key)
+{
+	return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
+}
+
+int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key)
+{
+	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
+}
+
+int tdb_chainlock_read(struct tdb_context *tdb, TDB_DATA key)
+{
+	return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
+}
+
+int tdb_chainunlock_read(struct tdb_context *tdb, TDB_DATA key)
+{
+	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
+}
+
+
+
+/* record lock stops delete underneath */
+int tdb_lock_record(struct tdb_context *tdb, tdb_off_t 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.
+*/
+int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t 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.
+*/
+int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off)
+{
+	return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
+}
+
+/* fcntl locks don't stack: avoid unlocking someone else's */
+int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t 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);
+}
diff --git a/source4/lib/tdb/common/open.c b/source4/lib/tdb/common/open.c
new file mode 100644
index 0000000000..d86ffff646
--- /dev/null
+++ b/source4/lib/tdb/common/open.c
@@ -0,0 +1,409 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+/* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
+static struct tdb_context *tdbs = NULL;
+
+
+/* This is based on the hash algorithm from gdbm */
+static unsigned int default_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);  
+}
+
+
+/* initialise a new database with a specified hash size */
+static int tdb_new_database(struct 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_t);
+	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;
+	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 = 0;
+	}
+
+  fail:
+	SAFE_FREE(newdb);
+	return ret;
+}
+
+
+
+static int tdb_already_open(dev_t device,
+			    ino_t ino)
+{
+	struct 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. */
+struct 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, NULL);
+}
+
+/* a default logging function */
+static void null_log_fn(struct tdb_context *tdb, int level, const char *fmt, ...)
+{
+}
+
+
+struct 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_hash_func hash_fn)
+{
+	struct tdb_context *tdb;
+	struct stat st;
+	int rev = 0, locked = 0;
+	uint8_t *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->flags = tdb_flags;
+	tdb->open_flags = open_flags;
+	tdb->log_fn = log_fn?log_fn:null_log_fn;
+	tdb->hash_fn = hash_fn ? hash_fn : default_tdb_hash;
+
+	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 ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
+	    (locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))) {
+		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
+		&& !(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 = (uint8_t *)&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;
+		tdb_convert(&tdb->header, sizeof(tdb->header));
+	}
+	if (fstat(tdb->fd, &st) == -1)
+		goto fail;
+
+	if (tdb->header.rwlocks != 0) {
+		TDB_LOG((tdb, 5, "tdb_open_ex: spinlocks no longer supported\n"));
+		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, (int)st.st_dev, (int)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_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;
+		}
+
+	}
+
+	/* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
+	   we didn't get the initial exclusive lock as we need to let all other
+	   users know we're using it. */
+
+	if (tdb_flags & TDB_CLEAR_IF_FIRST) {
+		/* 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(struct tdb_context *tdb)
+{
+	struct 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);
+
+	/* 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;
+}
+
+/* register a loging function */
+void tdb_logging_function(struct tdb_context *tdb, void (*fn)(struct tdb_context *, int , const char *, ...))
+{
+	tdb->log_fn = fn?fn:null_log_fn;
+}
+
+
+/* reopen a tdb - this can be used after a fork to ensure that we have an independent
+   seek pointer from our parent and to re-establish locks */
+int tdb_reopen(struct tdb_context *tdb)
+{
+	struct stat st;
+
+	if (tdb->flags & TDB_INTERNAL)
+		return 0; /* Nothing to do. */
+	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 ((tdb->flags & TDB_CLEAR_IF_FIRST) && 
+	    (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;
+	}
+	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);
+
+	return 0;
+
+fail:
+	tdb_close(tdb);
+	return -1;
+}
+
+/* reopen all tdb's */
+int tdb_reopen_all(void)
+{
+	struct tdb_context *tdb;
+
+	for (tdb=tdbs; tdb; tdb = tdb->next) {
+		if (tdb_reopen(tdb) != 0)
+			return -1;
+	}
+
+	return 0;
+}
diff --git a/source4/lib/tdb/common/spinlock.c b/source4/lib/tdb/common/spinlock.c
deleted file mode 100644
index 232cd8dfc6..0000000000
--- a/source4/lib/tdb/common/spinlock.c
+++ /dev/null
@@ -1,482 +0,0 @@
-/* 
-   Unix SMB/CIFS implementation.
-
-   trivial database library 
-
-   Copyright (C) Anton Blanchard                   2001
-   
-     ** NOTE! The following LGPL license applies to the tdb
-     ** library. This does NOT imply that all of Samba is released
-     ** under the LGPL
-   
-   This library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2 of the License, or (at your option) any later version.
-
-   This library 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
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with this library; if not, write to the Free Software
-   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
-*/
-#if HAVE_CONFIG_H
-#include <config.h>
-#endif
-
-#ifndef _SAMBA_BUILD_
-#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"
-
-#define DEBUG
-#else
-#include "includes.h"
-#include "lib/tdb/include/tdb.h"
-#include "system/filesys.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) && defined(sgi) && (_COMPILER_VERSION >= 730)
-
-/* Implement spinlocks on IRIX using the MIPSPro atomic fetch operations. See
- * sync(3) for the details of the intrinsic operations.
- *
- * "sgi" and "_COMPILER_VERSION" are always defined by MIPSPro.
- */
-
-#ifndef _SAMBA_BUILD_
-
-/* MIPSPro 7.3 has "__inline" as an extension, but not "inline. */
-#define inline __inline
-
-#endif /* _SAMBA_BUILD_ */
-
-/* Returns 0 if the lock is acquired, EBUSY otherwise. */
-static inline int __spin_trylock(spinlock_t *lock)
-{
-        unsigned int val;
-        val = __lock_test_and_set(lock, 1);
-        return val == 0 ? 0 : EBUSY;
-}
-
-static inline void __spin_unlock(spinlock_t *lock)
-{
-        __lock_release(lock);
-}
-
-static inline void __spin_lock_init(spinlock_t *lock)
-{
-        __lock_release(lock);
-}
-
-/* Returns 1 if the lock is held, 0 otherwise. */
-static inline int __spin_is_locked(spinlock_t *lock)
-{
-        unsigned int val;
-        val = __add_and_fetch(lock, 0);
-	return val;
-}
-
-#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)
-{
-#if defined(HAVE_SYSCONF) && defined(SYSCONF_SC_NPROC_ONLN)
-        return (sysconf(_SC_NPROC_ONLN) > 1) ? 1 : 0;
-#else
-	return 0;
-#endif
-}
-
-/*
- * 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 int size, i;
-	tdb_rwlock_t *rwlocks;
-
-	size = TDB_SPINLOCK_SIZE(hash_size);
-	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 int 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/lib/tdb/common/tdb.c b/source4/lib/tdb/common/tdb.c
index 3477d7e4be..f099c2d1aa 100644
--- a/source4/lib/tdb/common/tdb.c
+++ b/source4/lib/tdb/common/tdb.c
@@ -3,7 +3,7 @@
 
    trivial database library
 
-   Copyright (C) Andrew Tridgell              1999-2004
+   Copyright (C) Andrew Tridgell              1999-2005
    Copyright (C) Paul `Rusty' Russell		   2000
    Copyright (C) Jeremy Allison			   2000-2003
    
@@ -26,1007 +26,28 @@
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
 
+#include "tdb_private.h"
 
-#ifndef _SAMBA_BUILD_
-#if HAVE_CONFIG_H
-#include <config.h>
-#endif
-
-#include <stdlib.h>
-#include <stdio.h>
-#include <stdint.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 "tdb.h"
-#else
-#include "includes.h"
-#include "lib/tdb/include/tdb.h"
-#include "system/time.h"
-#include "system/shmem.h"
-#include "system/filesys.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))
-#define TDB_DATA_START(hash_size) (TDB_HASH_TOP(hash_size-1) + TDB_SPINLOCK_SIZE(hash_size))
-
-
-/* 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 x
-
-/* 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
-
-#ifndef discard_const_p
-# if defined(__intptr_t_defined) || defined(HAVE_INTPTR_T)
-#  define discard_const(ptr) ((void *)((intptr_t)(ptr)))
-# else
-#  define discard_const(ptr) ((void *)(ptr))
-# endif
-# define discard_const_p(type, ptr) ((type *)discard_const(ptr))
-#endif
-
-/* free memory if the pointer is valid and zero the pointer */
-#ifndef SAFE_FREE
-#define SAFE_FREE(x) do { if ((x) != NULL) {free(discard_const_p(void *, (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)
-		}
-	*/
-};
-
-/* 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);
-	} while (ret == -1 && errno == EINTR);
-
-	if (ret == -1) {
-		if (!probe && lck_type != F_SETLK) {
-			/* Ensure error code is set for log fun to examine. */
-			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));
-		}
-		/* Generic lock error. errno set by fcntl.
-		 * EAGAIN is an expected return from non-blocking
-		 * locks. */
-		if (errno != EAGAIN) {
-		TDB_LOG((tdb, 5, "tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d: %s\n", 
-				 tdb->fd, offset, rw_type, lck_type, 
-				 strerror(errno)));
-		}
-		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 %d 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 default_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 0x%x 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=0x%08x next=0x%08x 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 int)tailer, (unsigned int)(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)] (end = 0x%08x)\n", 
-		       rec_ptr, rec.rec_len, rec.rec_len, rec_ptr + 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_DATA_START(tdb->header.hash_size)) {
-		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) {
-		char *new_map_ptr = realloc(tdb->map_ptr, tdb->map_size);
-		if (!new_map_ptr) {
-			tdb->map_size -= size;
-			goto fail;
-		}
-		tdb->map_ptr = new_map_ptr;
-	} 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;
-}
-
-
-/* 
-   the core of tdb_allocate - called when we have decided which
-   free list entry to use
- */
-static tdb_off tdb_allocate_ofs(TDB_CONTEXT *tdb, tdb_len length, tdb_off rec_ptr,
-				struct list_struct *rec, tdb_off last_ptr)
-{
-	struct list_struct newrec;
-	tdb_off newrec_ptr;
-
-	memset(&newrec, '\0', sizeof(newrec));
-
-	/* 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) {
-		return 0;
-	}
-	
-	/* 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) {
-		return 0;
-	}
-	
-	/* Did we create new block? */
-	if (newrec_ptr) {
-		/* Update allocated record tailer (we
-		   shortened it). */
-		if (update_tailer(tdb, rec_ptr, rec) == -1) {
-			return 0;
-		}
-		
-		/* Free new record */
-		if (tdb_free(tdb, newrec_ptr, &newrec) == -1) {
-			return 0;
-		}
-	}
-	
-	/* all done - return the new record offset */
-	return rec_ptr;
-}
-
-/* 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 {
-		tdb_off rec_ptr, last_ptr;
-		tdb_len rec_len;
-	} bestfit;
-
-	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;
-
-	bestfit.rec_ptr = 0;
-
-	/* 
-	   this is a best fit allocation strategy. Originally we used
-	   a first fit strategy, but it suffered from massive fragmentation
-	   issues when faced with a slowly increasing record size.
-	 */
-	while (rec_ptr) {
-		if (rec_free_read(tdb, rec_ptr, rec) == -1) {
-			goto fail;
-		}
-
-		if (rec->rec_len >= length) {
-			if (bestfit.rec_ptr == 0 ||
-			    rec->rec_len < bestfit.rec_len) {
-				bestfit.rec_len = rec->rec_len;
-				bestfit.rec_ptr = rec_ptr;
-				bestfit.last_ptr = last_ptr;
-				/* consider a fit to be good enough if we aren't wasting more than half the space */
-				if (bestfit.rec_len < 2*length) {
-					break;
-				}
-			}
-		}
-
-		/* move to the next record */
-		last_ptr = rec_ptr;
-		rec_ptr = rec->next;
-	}
-
-	if (bestfit.rec_ptr != 0) {
-		if (rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
-			goto fail;
-		}
-
-		newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, rec, bestfit.last_ptr);
-		tdb_unlock(tdb, -1, F_WRLCK);
-		return newrec_ptr;
-	}
-
-	/* 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,
+static tdb_off_t tdb_find(struct tdb_context *tdb, TDB_DATA key, u32 hash,
 			struct list_struct *r)
 {
-	tdb_off rec_ptr;
+	tdb_off_t rec_ptr;
 	
 	/* read in the hash top */
-	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
+	if (tdb_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)
+		if (tdb_rec_read(tdb, rec_ptr, r) == -1)
 			return 0;
 
 		if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
-			char *k;
+			unsigned char *k;
 			/* a very likely hit - read the key */
 			k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r), 
 					   r->key_len);
@@ -1045,8 +66,8 @@ static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
 }
 
 /* As tdb_find, but if you succeed, keep the lock */
-static tdb_off tdb_find_lock_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, int locktype,
-			     struct list_struct *rec)
+tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, int locktype,
+			   struct list_struct *rec)
 {
 	u32 rec_ptr;
 
@@ -1057,48 +78,22 @@ static tdb_off tdb_find_lock_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, int
 	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_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA dbuf)
+static int tdb_update_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, TDB_DATA dbuf)
 {
 	struct list_struct rec;
-	tdb_off rec_ptr;
+	tdb_off_t rec_ptr;
 
 	/* find entry */
 	if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
 		return -1;
 
 	/* must be long enough key, data and tailer */
-	if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) {
+	if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off_t)) {
 		tdb->ecode = TDB_SUCCESS; /* Not really an error */
 		return -1;
 	}
@@ -1110,7 +105,7 @@ static int tdb_update_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA db
 	if (dbuf.dsize != rec.data_len) {
 		/* update size */
 		rec.data_len = dbuf.dsize;
-		return rec_write(tdb, rec_ptr, &rec);
+		return tdb_rec_write(tdb, rec_ptr, &rec);
 	}
  
 	return 0;
@@ -1122,10 +117,9 @@ static int tdb_update_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA db
  * then the TDB_DATA will have zero length but
  * a non-zero pointer
  */
-
-TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
+TDB_DATA tdb_fetch(struct tdb_context *tdb, TDB_DATA key)
 {
-	tdb_off rec_ptr;
+	tdb_off_t rec_ptr;
 	struct list_struct rec;
 	TDB_DATA ret;
 	u32 hash;
@@ -1148,7 +142,7 @@ TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
    this doesn't match the conventions in the rest of this module, but is
    compatible with gdbm
 */
-static int tdb_exists_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
+static int tdb_exists_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash)
 {
 	struct list_struct rec;
 	
@@ -1158,82 +152,39 @@ static int tdb_exists_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
 	return 1;
 }
 
-int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
+int tdb_exists(struct tdb_context *tdb, TDB_DATA key)
 {
 	u32 hash = tdb->hash_fn(&key);
 	return tdb_exists_hash(tdb, key, hash);
 }
 
-/* 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)
+int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct list_struct*rec)
 {
-	tdb_off last_ptr, i;
+	tdb_off_t last_ptr, i;
 	struct list_struct lastrec;
 
 	if (tdb->read_only) return -1;
 
-	if (write_lock_record(tdb, rec_ptr) == -1) {
+	if (tdb_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);
+		return tdb_rec_write(tdb, rec_ptr, rec);
 	}
-	if (write_unlock_record(tdb, rec_ptr) != 0)
+	if (tdb_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)
+	if (tdb_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)
+		if (tdb_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)
+	if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1)
 		return -1;
 
 	/* recover the space */
@@ -1242,279 +193,22 @@ static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
 	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);
-
-	/* Lock each chain from the start one. */
-	for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
-
-		/* this is an optimisation for the common case where
-		   the hash chain is empty, which is particularly
-		   common for the use of tdb with ldb, where large
-		   hashes are used. In that case we spend most of our
-		   time in tdb_brlock(), locking empty hash chains.
-
-		   To avoid this, we do an unlocked pre-check to see
-		   if the hash chain is empty before starting to look
-		   inside it. If it is empty then we can avoid that
-		   hash chain. If it isn't empty then we can't believe
-		   the value we get back, as we read it without a
-		   lock, so instead we get the lock and re-fetch the
-		   value below.
-
-		   Notice that not doing this optimisation on the
-		   first hash chain is critical. We must guarantee
-		   that we have done at least one fcntl lock at the
-		   start of a search to guarantee that memory is
-		   coherent on SMP systems. If records are added by
-		   others during the search then thats OK, and we
-		   could possibly miss those with this trick, but we
-		   could miss them anyway without this trick, so the
-		   semantics don't change.
-
-		   With a non-indexed ldb search this trick gains us a
-		   factor of around 80 in speed on a linux 2.6.x
-		   system (testing using ldbtest).
-		 */
-		if (!tlock->off && tlock->hash != 0) {
-			u32 off;
-			if (tdb->map_ptr) {
-				for (;tlock->hash < tdb->header.hash_size;tlock->hash++) {
-					if (0 != *(u32 *)(TDB_HASH_TOP(tlock->hash) + (unsigned char *)tdb->map_ptr)) {
-						break;
-					}
-				}
-				if (tlock->hash == tdb->header.hash_size) {
-					continue;
-				}
-			} else {
-				if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash), &off) == 0 &&
-				    off == 0) {
-					continue;
-				}
-			}
-		}
-
-		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;
-
-			/* Detect infinite loops. From "Shlomi Yaakobovich" <Shlomi@exanet.com>. */
-			if (tlock->off == rec->next) {
-				TDB_LOG((tdb, 0, "tdb_next_lock: loop detected.\n"));
-				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 (!tdb->read_only && 
-			    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 *private)
-{
-	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, private)) {
-			/* 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_hash(tdb, oldkey, tdb->hash_fn(&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 */
-static int tdb_delete_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
+static int tdb_delete_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash)
 {
-	tdb_off rec_ptr;
+	tdb_off_t rec_ptr;
 	struct list_struct rec;
 	int ret;
 
 	if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK, &rec)))
 		return -1;
-	ret = do_delete(tdb, rec_ptr, &rec);
+	ret = tdb_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;
 }
 
-int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
+int tdb_delete(struct tdb_context *tdb, TDB_DATA key)
 {
 	u32 hash = tdb->hash_fn(&key);
 	return tdb_delete_hash(tdb, key, hash);
@@ -1525,11 +219,11 @@ int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
 
    return 0 on success, -1 on failure
 */
-int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
+int tdb_store(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
 {
 	struct list_struct rec;
 	u32 hash;
-	tdb_off rec_ptr;
+	tdb_off_t rec_ptr;
 	char *p = NULL;
 	int ret = 0;
 
@@ -1581,7 +275,7 @@ int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
 		goto fail;
 
 	/* Read hash top into next ptr */
-	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
+	if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
 		goto fail;
 
 	rec.key_len = key.dsize;
@@ -1590,9 +284,9 @@ int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
 	rec.magic = TDB_MAGIC;
 
 	/* write out and point the top of the hash chain at it */
-	if (rec_write(tdb, rec_ptr, &rec) == -1
+	if (tdb_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) {
+	    || tdb_ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
 		/* Need to tdb_unallocate() here */
 		goto fail;
 	}
@@ -1605,499 +299,68 @@ fail:
 	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, u32 hash, TDB_DATA new_dbuf)
-{
-	struct list_struct rec;
-	tdb_off rec_ptr;
-
-	/* find entry */
-	if (!(rec_ptr = tdb_find(tdb, key, hash, &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)
+int tdb_append(struct 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;
+	TDB_DATA dbuf;
+	int ret = -1;
 
 	/* find which hash bucket it is in */
 	hash = tdb->hash_fn(&key);
 	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
 		return -1;
 
-	/* first try in-place. */
-	if (tdb_append_inplace(tdb, key, hash, new_dbuf) == 0)
-		goto out;
+	dbuf = tdb_fetch(tdb, key);
 
-	/* 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;
+	if (dbuf.dptr == NULL) {
+		dbuf.dptr = malloc(new_dbuf.dsize);
+	} else {
+		dbuf.dptr = realloc(dbuf.dptr, dbuf.dsize + new_dbuf.dsize);
 	}
 
-	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))) {
+	if (dbuf.dptr == NULL) {
 		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_hash(tdb, key, hash);
-
-	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, NULL);
-}
-
-/* a default logging function */
-static void null_log_fn(TDB_CONTEXT *tdb, int level, const char *fmt, ...)
-{
-}
-
-
-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_hash_func hash_fn)
-{
-	TDB_CONTEXT *tdb;
-	struct stat st;
-	int rev = 0, locked = 0;
-	uint8_t *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->flags = tdb_flags;
-	tdb->open_flags = open_flags;
-	tdb->log_fn = log_fn?log_fn:null_log_fn;
-	tdb->hash_fn = hash_fn ? hash_fn : default_tdb_hash;
-
-	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 ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
-		(locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))) {
-		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
-		&& !(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 = (uint8_t *)&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, (int)st.st_dev, (int)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;
-		}
-
-	}
-
-	/* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
-	   we didn't get the initial exclusive lock as we need to let all other
-	   users know we're using it. */
-
-	if (tdb_flags & TDB_CLEAR_IF_FIRST) {
-	/* 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;
+		goto failed;
 	}
 
+	memcpy(dbuf.dptr + dbuf.dsize, new_dbuf.dptr, new_dbuf.dsize);
+	dbuf.dsize += new_dbuf.dsize;
 
- 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;
+	ret = tdb_store(tdb, key, dbuf, 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)
-		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);
-
-	/* 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);
-
+failed:
+	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
+	SAFE_FREE(dbuf.dptr);
 	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);
-	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);
-}
-
-/* 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_fn(&key)), F_WRLCK);
-}
-
-int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
-{
-	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
-}
 
-int tdb_chainlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
-{
-	return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
-}
-
-int tdb_chainunlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
+/*
+  return the name of the current tdb file
+  useful for external logging functions
+*/
+const char *tdb_name(struct tdb_context *tdb)
 {
-	return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
+	return tdb->name;
 }
 
-
-/* register a loging function */
-void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
+/*
+  return the underlying file descriptor being used by tdb, or -1
+  useful for external routines that want to check the device/inode
+  of the fd
+*/
+int tdb_fd(struct tdb_context *tdb)
 {
-	tdb->log_fn = fn?fn:null_log_fn;
+	return tdb->fd;
 }
 
-
-/* reopen a tdb - this can be 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->flags & TDB_INTERNAL)
-		return 0; /* Nothing to do. */
-	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->flags & TDB_CLEAR_IF_FIRST) && (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)
+/*
+  return the current logging function
+  useful for external tdb routines that wish to log tdb errors
+*/
+tdb_log_func tdb_log_fn(struct tdb_context *tdb)
 {
-	TDB_CONTEXT *tdb;
-
-	for (tdb=tdbs; tdb; tdb = tdb->next) {
-		/* Ensure no clear-if-first. */
-		tdb->flags &= ~TDB_CLEAR_IF_FIRST;
-		if (tdb_reopen(tdb) != 0)
-			return -1;
-	}
-
-	return 0;
+	return tdb->log_fn;
 }
diff --git a/source4/lib/tdb/common/tdb_private.h b/source4/lib/tdb/common/tdb_private.h
new file mode 100644
index 0000000000..8e246e3283
--- /dev/null
+++ b/source4/lib/tdb/common/tdb_private.h
@@ -0,0 +1,208 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library - private includes
+
+   Copyright (C) Andrew Tridgell              2005
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#ifndef _SAMBA_BUILD_
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdint.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 "tdbconfig.h"
+#include "tdb.h"
+#else
+#include "includes.h"
+#include "lib/tdb/include/tdb.h"
+#include "system/time.h"
+#include "system/shmem.h"
+#include "system/filesys.h"
+#endif
+
+#ifndef u32
+#define u32 unsigned
+#endif
+
+typedef u32 tdb_len_t;
+typedef u32 tdb_off_t;
+
+#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_t))
+#define TDB_DATA_START(hash_size) TDB_HASH_TOP(hash_size-1)
+
+
+/* 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 x
+
+/* 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
+
+#ifndef discard_const_p
+# if defined(__intptr_t_defined) || defined(HAVE_INTPTR_T)
+#  define discard_const(ptr) ((void *)((intptr_t)(ptr)))
+# else
+#  define discard_const(ptr) ((void *)(ptr))
+# endif
+# define discard_const_p(type, ptr) ((type *)discard_const(ptr))
+#endif
+
+/* free memory if the pointer is valid and zero the pointer */
+#ifndef SAFE_FREE
+#define SAFE_FREE(x) do { if ((x) != NULL) {free(discard_const_p(void *, (x))); (x)=NULL;} } while(0)
+#endif
+
+#define BUCKET(hash) ((hash) % tdb->header.hash_size)
+
+#define DOCONV() (tdb->flags & TDB_CONVERT)
+#define CONVERT(x) (DOCONV() ? tdb_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_t next; /* offset of the next record in the list */
+	tdb_len_t rec_len; /* total byte length of record */
+	tdb_len_t key_len; /* byte length of key */
+	tdb_len_t 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)
+		}
+	*/
+};
+
+
+/* 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_t rwlocks;
+	tdb_off_t reserved[31];
+};
+
+struct tdb_lock_type {
+	u32 count;
+	u32 ltype;
+};
+
+struct tdb_traverse_lock {
+	struct tdb_traverse_lock *next;
+	u32 off;
+	u32 hash;
+};
+
+
+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_t 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 */
+	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 *, ...) PRINTF_ATTRIBUTE(3,4); /* logging function */
+	unsigned int (*hash_fn)(TDB_DATA *key);
+	int open_flags; /* flags used in the open - needed by reopen */
+	struct tdb_transaction *transaction;
+};
+
+
+/*
+  internal prototypes
+*/
+int tdb_munmap(struct tdb_context *tdb);
+void tdb_mmap(struct tdb_context *tdb);
+int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf, tdb_len_t len, int cv);
+int tdb_write(struct tdb_context *tdb, tdb_off_t off, void *buf, tdb_len_t len);
+int tdb_oob(struct tdb_context *tdb, tdb_off_t len, int probe);
+int tdb_lock(struct tdb_context *tdb, int list, int ltype);
+int tdb_unlock(struct tdb_context *tdb, int list, int ltype);
+int tdb_expand(struct tdb_context *tdb, tdb_off_t size);
+int tdb_brlock(struct tdb_context *tdb, tdb_off_t offset, int rw_type, int lck_type, int probe);
+int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+void *tdb_convert(void *buf, u32 size);
+int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
+tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct list_struct *rec);
+int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
+int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off);
+int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
+int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
+int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct list_struct *rec);
+unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
+tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, int locktype,
+			   struct list_struct *rec);
+void tdb_next_hash_chain(struct tdb_context *tdb, u32 *chain);
+
diff --git a/source4/lib/tdb/common/tdbutil.c b/source4/lib/tdb/common/tdbutil.c
index df9ce4e238..327555db10 100644
--- a/source4/lib/tdb/common/tdbutil.c
+++ b/source4/lib/tdb/common/tdbutil.c
@@ -1,7 +1,9 @@
 /* 
    Unix SMB/CIFS implementation.
+
    tdb utility functions
-   Copyright (C) Andrew Tridgell 1992-1998
+
+   Copyright (C) Andrew Tridgell 1992-2005
    
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -17,6 +19,10 @@
    along with this program; if not, write to the Free Software
    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
+/*
+  NOTE: these utility functions are specific to Samba, and are not part
+  of the core tdb code
+*/
 
 #include "includes.h"
 #include "lib/tdb/include/tdbutil.h"
@@ -36,7 +42,7 @@
 static TDB_DATA make_tdb_data(const char *dptr, size_t dsize)
 {
 	TDB_DATA ret;
-	ret.dptr = discard_const_p(char, dptr);
+	ret.dptr = discard_const_p(unsigned char, dptr);
 	ret.dsize = dsize;
 	return ret;
 }
@@ -45,7 +51,7 @@ static TDB_DATA make_tdb_data(const char *dptr, size_t dsize)
  Lock a chain by string. Return -1 if lock failed.
 ****************************************************************************/
 
-int tdb_lock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+int tdb_lock_bystring(struct tdb_context *tdb, const char *keyval)
 {
 	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
 	
@@ -56,7 +62,7 @@ int tdb_lock_bystring(TDB_CONTEXT *tdb, const char *keyval)
  Unlock a chain by string.
 ****************************************************************************/
 
-void tdb_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+void tdb_unlock_bystring(struct tdb_context *tdb, const char *keyval)
 {
 	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
 
@@ -67,7 +73,7 @@ void tdb_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
  Read lock a chain by string. Return -1 if lock failed.
 ****************************************************************************/
 
-int tdb_read_lock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+int tdb_read_lock_bystring(struct tdb_context *tdb, const char *keyval)
 {
 	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
 	
@@ -78,7 +84,7 @@ int tdb_read_lock_bystring(TDB_CONTEXT *tdb, const char *keyval)
  Read unlock a chain by string.
 ****************************************************************************/
 
-void tdb_read_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
+void tdb_read_unlock_bystring(struct tdb_context *tdb, const char *keyval)
 {
 	TDB_DATA key = make_tdb_data(keyval, strlen(keyval)+1);
 	
@@ -91,7 +97,7 @@ void tdb_read_unlock_bystring(TDB_CONTEXT *tdb, const char *keyval)
  Output is int32_t in native byte order.
 ****************************************************************************/
 
-int32_t tdb_fetch_int32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len)
+int32_t tdb_fetch_int32_byblob(struct tdb_context *tdb, const char *keyval, size_t len)
 {
 	TDB_DATA key = make_tdb_data(keyval, len);
 	TDB_DATA data;
@@ -113,7 +119,7 @@ int32_t tdb_fetch_int32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len)
  Output is int32_t in native byte order.
 ****************************************************************************/
 
-int32_t tdb_fetch_int32(TDB_CONTEXT *tdb, const char *keystr)
+int32_t tdb_fetch_int32(struct tdb_context *tdb, const char *keystr)
 {
 	return tdb_fetch_int32_byblob(tdb, keystr, strlen(keystr) + 1);
 }
@@ -123,7 +129,7 @@ int32_t tdb_fetch_int32(TDB_CONTEXT *tdb, const char *keystr)
  Input is int32_t 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_t v)
+int tdb_store_int32_byblob(struct tdb_context *tdb, const char *keystr, size_t len, int32_t v)
 {
 	TDB_DATA key = make_tdb_data(keystr, len);
 	TDB_DATA data;
@@ -141,7 +147,7 @@ int tdb_store_int32_byblob(TDB_CONTEXT *tdb, const char *keystr, size_t len, int
  Input is int32_t in native byte order. Output in tdb is in little-endian.
 ****************************************************************************/
 
-int tdb_store_int32(TDB_CONTEXT *tdb, const char *keystr, int32_t v)
+int tdb_store_int32(struct tdb_context *tdb, const char *keystr, int32_t v)
 {
 	return tdb_store_int32_byblob(tdb, keystr, strlen(keystr) + 1, v);
 }
@@ -151,7 +157,7 @@ int tdb_store_int32(TDB_CONTEXT *tdb, const char *keystr, int32_t v)
  Output is uint32_t in native byte order.
 ****************************************************************************/
 
-BOOL tdb_fetch_uint32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len, uint32_t *value)
+BOOL tdb_fetch_uint32_byblob(struct tdb_context *tdb, const char *keyval, size_t len, uint32_t *value)
 {
 	TDB_DATA key = make_tdb_data(keyval, len);
 	TDB_DATA data;
@@ -172,7 +178,7 @@ BOOL tdb_fetch_uint32_byblob(TDB_CONTEXT *tdb, const char *keyval, size_t len, u
  Output is uint32_t in native byte order.
 ****************************************************************************/
 
-BOOL tdb_fetch_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32_t *value)
+BOOL tdb_fetch_uint32(struct tdb_context *tdb, const char *keystr, uint32_t *value)
 {
 	return tdb_fetch_uint32_byblob(tdb, keystr, strlen(keystr) + 1, value);
 }
@@ -182,7 +188,7 @@ BOOL tdb_fetch_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32_t *value)
  Input is uint32_t 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_t value)
+BOOL tdb_store_uint32_byblob(struct tdb_context *tdb, const char *keystr, size_t len, uint32_t value)
 {
 	TDB_DATA key = make_tdb_data(keystr, len);
 	TDB_DATA data;
@@ -204,7 +210,7 @@ BOOL tdb_store_uint32_byblob(TDB_CONTEXT *tdb, const char *keystr, size_t len, u
  Input is uint32_t in native byte order. Output in tdb is in little-endian.
 ****************************************************************************/
 
-BOOL tdb_store_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32_t value)
+BOOL tdb_store_uint32(struct tdb_context *tdb, const char *keystr, uint32_t value)
 {
 	return tdb_store_uint32_byblob(tdb, keystr, strlen(keystr) + 1, value);
 }
@@ -213,7 +219,7 @@ BOOL tdb_store_uint32(TDB_CONTEXT *tdb, const char *keystr, uint32_t value)
  on failure.
 ****************************************************************************/
 
-int tdb_store_bystring(TDB_CONTEXT *tdb, const char *keystr, TDB_DATA data, int flags)
+int tdb_store_bystring(struct tdb_context *tdb, const char *keystr, TDB_DATA data, int flags)
 {
 	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
 	
@@ -225,7 +231,7 @@ int tdb_store_bystring(TDB_CONTEXT *tdb, const char *keystr, TDB_DATA data, int
  free() on the result dptr.
 ****************************************************************************/
 
-TDB_DATA tdb_fetch_bystring(TDB_CONTEXT *tdb, const char *keystr)
+TDB_DATA tdb_fetch_bystring(struct tdb_context *tdb, const char *keystr)
 {
 	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
 
@@ -236,7 +242,7 @@ TDB_DATA tdb_fetch_bystring(TDB_CONTEXT *tdb, const char *keystr)
  Delete an entry using a null terminated string key. 
 ****************************************************************************/
 
-int tdb_delete_bystring(TDB_CONTEXT *tdb, const char *keystr)
+int tdb_delete_bystring(struct tdb_context *tdb, const char *keystr)
 {
 	TDB_DATA key = make_tdb_data(keystr, strlen(keystr)+1);
 
@@ -247,7 +253,7 @@ int tdb_delete_bystring(TDB_CONTEXT *tdb, const char *keystr)
  Atomic integer change. Returns old value. To create, set initial value in *oldval. 
 ****************************************************************************/
 
-int32_t tdb_change_int32_atomic(TDB_CONTEXT *tdb, const char *keystr, int32_t *oldval, int32_t change_val)
+int32_t tdb_change_int32_atomic(struct tdb_context *tdb, const char *keystr, int32_t *oldval, int32_t change_val)
 {
 	int32_t val;
 	int32_t ret = -1;
@@ -288,7 +294,7 @@ int32_t tdb_change_int32_atomic(TDB_CONTEXT *tdb, const char *keystr, int32_t *o
  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_t *oldval, uint32_t change_val)
+BOOL tdb_change_uint32_atomic(struct tdb_context *tdb, const char *keystr, uint32_t *oldval, uint32_t change_val)
 {
 	uint32_t val;
 	BOOL ret = False;
@@ -327,6 +333,18 @@ BOOL tdb_change_uint32_atomic(TDB_CONTEXT *tdb, const char *keystr, uint32_t *ol
 }
 
 /****************************************************************************
+ Allow tdb_delete to be used as a tdb_traversal_fn.
+****************************************************************************/
+
+int tdb_traverse_delete_fn(struct tdb_context *the_tdb, TDB_DATA key, TDB_DATA dbuf,
+                     void *state)
+{
+    return tdb_delete(the_tdb, key);
+}
+
+
+
+/****************************************************************************
  Useful pair of routines for packing/unpacking data consisting of
  integers and strings.
 ****************************************************************************/
@@ -345,6 +363,7 @@ size_t tdb_pack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 	char *buf0 = buf;
 	const char *fmt0 = fmt;
 	int bufsize0 = bufsize;
+	tdb_log_func log_fn = tdb_log_fn(tdb);
 
 	va_start(ap, fmt);
 
@@ -399,8 +418,8 @@ size_t tdb_pack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 			}
 			break;
 		default:
-			tdb->log_fn(tdb, 0,"Unknown tdb_pack format %c in %s\n", 
-				    c, fmt);
+			log_fn(tdb, 0,"Unknown tdb_pack format %c in %s\n", 
+			       c, fmt);
 			len = 0;
 			break;
 		}
@@ -414,8 +433,8 @@ size_t tdb_pack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 
 	va_end(ap);
 
-	tdb->log_fn(tdb, 18,"tdb_pack(%s, %d) -> %d\n", 
-		    fmt0, bufsize0, (int)PTR_DIFF(buf, buf0));
+	log_fn(tdb, 18,"tdb_pack(%s, %d) -> %d\n", 
+	       fmt0, bufsize0, (int)PTR_DIFF(buf, buf0));
 	
 	return PTR_DIFF(buf, buf0);
 }
@@ -439,6 +458,7 @@ int tdb_unpack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 	char *buf0 = buf;
 	const char *fmt0 = fmt;
 	int bufsize0 = bufsize;
+	tdb_log_func log_fn = tdb_log_fn(tdb);
 
 	va_start(ap, fmt);
 	
@@ -506,8 +526,8 @@ int tdb_unpack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 			memcpy(*b, buf+4, *i);
 			break;
 		default:
-			tdb->log_fn(tdb, 0, "Unknown tdb_unpack format %c in %s\n", 
-				 c, fmt);
+			log_fn(tdb, 0, "Unknown tdb_unpack format %c in %s\n", 
+			       c, fmt);
 
 			len = 0;
 			break;
@@ -519,91 +539,11 @@ int tdb_unpack(TDB_CONTEXT *tdb, char *buf, int bufsize, const char *fmt, ...)
 
 	va_end(ap);
 
-	tdb->log_fn(tdb, 18, "tdb_unpack(%s, %d) -> %d\n", 
-		    fmt0, bufsize0, (int)PTR_DIFF(buf, buf0));
+	log_fn(tdb, 18, "tdb_unpack(%s, %d) -> %d\n", 
+	       fmt0, bufsize0, (int)PTR_DIFF(buf, buf0));
 
 	return PTR_DIFF(buf, buf0);
 
  no_space:
 	return -1;
 }
-
-/****************************************************************************
- 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;
-	
-	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 0
-		if (!key_str) {
-			tdb->log_fn(tdb, 0, "tdb_search_keys: strndup() failed!\n");
-			smb_panic("strndup failed!\n");
-		}
-#endif
-		tdb->log_fn(tdb, 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, TDB_LIST_NODE *);
-		
-			tdb->log_fn(tdb, 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_key.dptr);
-		SAFE_FREE(node);
-		node = next_node;
-	}
-}
diff --git a/source4/lib/tdb/common/traverse.c b/source4/lib/tdb/common/traverse.c
new file mode 100644
index 0000000000..d14be355a2
--- /dev/null
+++ b/source4/lib/tdb/common/traverse.c
@@ -0,0 +1,274 @@
+ /* 
+   Unix SMB/CIFS implementation.
+
+   trivial database library
+
+   Copyright (C) Andrew Tridgell              1999-2005
+   Copyright (C) Paul `Rusty' Russell		   2000
+   Copyright (C) Jeremy Allison			   2000-2003
+   
+     ** NOTE! The following LGPL license applies to the tdb
+     ** library. This does NOT imply that all of Samba is released
+     ** under the LGPL
+   
+   This library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2 of the License, or (at your option) any later version.
+
+   This library 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this library; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+*/
+
+#include "tdb_private.h"
+
+/* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
+static int tdb_next_lock(struct tdb_context *tdb, struct tdb_traverse_lock *tlock,
+			 struct list_struct *rec)
+{
+	int want_next = (tlock->off != 0);
+
+	/* Lock each chain from the start one. */
+	for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
+		if (!tlock->off && tlock->hash != 0) {
+			/* this is an optimisation for the common case where
+			   the hash chain is empty, which is particularly
+			   common for the use of tdb with ldb, where large
+			   hashes are used. In that case we spend most of our
+			   time in tdb_brlock(), locking empty hash chains.
+			   
+			   To avoid this, we do an unlocked pre-check to see
+			   if the hash chain is empty before starting to look
+			   inside it. If it is empty then we can avoid that
+			   hash chain. If it isn't empty then we can't believe
+			   the value we get back, as we read it without a
+			   lock, so instead we get the lock and re-fetch the
+			   value below.
+			   
+			   Notice that not doing this optimisation on the
+			   first hash chain is critical. We must guarantee
+			   that we have done at least one fcntl lock at the
+			   start of a search to guarantee that memory is
+			   coherent on SMP systems. If records are added by
+			   others during the search then thats OK, and we
+			   could possibly miss those with this trick, but we
+			   could miss them anyway without this trick, so the
+			   semantics don't change.
+			   
+			   With a non-indexed ldb search this trick gains us a
+			   factor of around 80 in speed on a linux 2.6.x
+			   system (testing using ldbtest).
+			*/
+			tdb_next_hash_chain(tdb, &tlock->hash);
+			if (tlock->hash == tdb->header.hash_size) {
+				continue;
+			}
+		}
+
+		if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1)
+			return -1;
+
+		/* No previous record?  Start at top of chain. */
+		if (!tlock->off) {
+			if (tdb_ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
+				     &tlock->off) == -1)
+				goto fail;
+		} else {
+			/* Otherwise unlock the previous record. */
+			if (tdb_unlock_record(tdb, tlock->off) != 0)
+				goto fail;
+		}
+
+		if (want_next) {
+			/* We have offset of old record: grab next */
+			if (tdb_rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+			tlock->off = rec->next;
+		}
+
+		/* Iterate through chain */
+		while( tlock->off) {
+			tdb_off_t current;
+			if (tdb_rec_read(tdb, tlock->off, rec) == -1)
+				goto fail;
+
+			/* Detect infinite loops. From "Shlomi Yaakobovich" <Shlomi@exanet.com>. */
+			if (tlock->off == rec->next) {
+				TDB_LOG((tdb, 0, "tdb_next_lock: loop detected.\n"));
+				goto fail;
+			}
+
+			if (!TDB_DEAD(rec)) {
+				/* Woohoo: we found one! */
+				if (tdb_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 (!tdb->read_only && 
+			    tdb_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(struct tdb_context *tdb, tdb_traverse_func fn, void *private)
+{
+	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 (tdb_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, private)) {
+			/* They want us to terminate traversal */
+			ret = count;
+			if (tdb_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(struct tdb_context *tdb)
+{
+	TDB_DATA key;
+	struct list_struct rec;
+
+	/* release any old lock */
+	if (tdb_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(struct tdb_context *tdb, TDB_DATA oldkey)
+{
+	u32 oldhash;
+	TDB_DATA key = tdb_null;
+	struct list_struct rec;
+	unsigned 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 (tdb_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 (tdb_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_hash(tdb, oldkey, tdb->hash_fn(&oldkey), F_WRLCK, &rec);
+		if (!tdb->travlocks.off)
+			return tdb_null;
+		tdb->travlocks.hash = BUCKET(rec.full_hash);
+		if (tdb_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;
+}
+