X-Git-Url: https://git.sur5r.net/?a=blobdiff_plain;f=servers%2Fslapd%2Fback-bdb%2Fdn2id.c;h=528c57815f15089372f2b11178150b740a4b9c08;hb=82c1b0d1f0a27e374961c910473f549878e81e24;hp=72178a6fe2b0ce53b1658f60e7504cec184a721b;hpb=ee1cddf6fd730afcfd9aa60551daf2ee99ccbef0;p=openldap

diff --git a/servers/slapd/back-bdb/dn2id.c b/servers/slapd/back-bdb/dn2id.c
index 72178a6fe2..528c57815f 100644
--- a/servers/slapd/back-bdb/dn2id.c
+++ b/servers/slapd/back-bdb/dn2id.c
@@ -1,8 +1,17 @@
 /* dn2id.c - routines to deal with the dn2id index */
 /* $OpenLDAP$ */
-/*
- * Copyright 1998-2003 The OpenLDAP Foundation, All Rights Reserved.
- * COPYING RESTRICTIONS APPLY, see COPYRIGHT file
+/* This work is part of OpenLDAP Software <http://www.openldap.org/>.
+ *
+ * Copyright 2000-2004 The OpenLDAP Foundation.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted only as authorized by the OpenLDAP
+ * Public License.
+ *
+ * A copy of this license is available in the file LICENSE in the
+ * top-level directory of the distribution or, alternatively, at
+ * <http://www.OpenLDAP.org/license.html>.
  */
 
 #include "portable.h"
@@ -17,13 +26,12 @@
 #ifndef BDB_HIER
 int
 bdb_dn2id_add(
-	BackendDB	*be,
+	Operation *op,
 	DB_TXN *txn,
 	EntryInfo *eip,
-	Entry		*e,
-	void *ctx )
+	Entry		*e )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 	int		rc;
 	DBT		key, data;
@@ -43,7 +51,7 @@ bdb_dn2id_add(
 	key.size = e->e_nname.bv_len + 2;
 	key.ulen = key.size;
 	key.flags = DB_DBT_USERMEM;
-	buf = sl_malloc( key.size, ctx );
+	buf = op->o_tmpalloc( key.size, op->o_tmpmemctx );
 	key.data = buf;
 	buf[0] = DN_BASE_PREFIX;
 	ptr.bv_val = buf + 1;
@@ -69,7 +77,7 @@ bdb_dn2id_add(
 	}
 
 #ifndef BDB_MULTIPLE_SUFFIXES
-	if( !be_issuffix( be, &ptr )) {
+	if( !be_issuffix( op->o_bd, &ptr )) {
 #endif
 		buf[0] = DN_SUBTREE_PREFIX;
 		rc = db->put( db, txn, &key, &data, DB_NOOVERWRITE );
@@ -87,7 +95,7 @@ bdb_dn2id_add(
 		}
 		
 #ifdef BDB_MULTIPLE_SUFFIXES
-	if( !be_issuffix( be, &ptr )) {
+	if( !be_issuffix( op->o_bd, &ptr )) {
 #endif
 		dnParent( &ptr, &pdn );
 	
@@ -97,7 +105,7 @@ bdb_dn2id_add(
 		key.data = pdn.bv_val-1;
 		ptr = pdn;
 
-		rc = bdb_idl_insert_key( be, db, txn, &key, e->e_id );
+		rc = bdb_idl_insert_key( op->o_bd, db, txn, &key, e->e_id );
 
 		if( rc != 0 ) {
 #ifdef NEW_LOGGING
@@ -114,13 +122,14 @@ bdb_dn2id_add(
 #ifndef BDB_MULTIPLE_SUFFIXES
 	}
 
-	while( !be_issuffix( be, &ptr )) {
+	while( !be_issuffix( op->o_bd, &ptr ))
 #else
-	for (;;) {
+	for (;;)
 #endif
+	{
 		ptr.bv_val[-1] = DN_SUBTREE_PREFIX;
 
-		rc = bdb_idl_insert_key( be, db, txn, &key, e->e_id );
+		rc = bdb_idl_insert_key( op->o_bd, db, txn, &key, e->e_id );
 
 		if( rc != 0 ) {
 #ifdef NEW_LOGGING
@@ -135,7 +144,7 @@ bdb_dn2id_add(
 			break;
 		}
 #ifdef BDB_MULTIPLE_SUFFIXES
-		if( be_issuffix( be, &ptr )) break;
+		if( be_issuffix( op->o_bd, &ptr )) break;
 #endif
 		dnParent( &ptr, &pdn );
 
@@ -149,7 +158,7 @@ bdb_dn2id_add(
 #endif
 
 done:
-	sl_free( buf, ctx );
+	op->o_tmpfree( buf, op->o_tmpmemctx );
 #ifdef NEW_LOGGING
 	LDAP_LOG ( INDEX, RESULTS, "<= bdb_dn2id_add: %d\n", rc, 0, 0 );
 #else
@@ -160,13 +169,12 @@ done:
 
 int
 bdb_dn2id_delete(
-	BackendDB	*be,
+	Operation *op,
 	DB_TXN *txn,
 	EntryInfo	*eip,
-	Entry		*e,
-	void *ctx )
+	Entry		*e )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 	int		rc;
 	DBT		key;
@@ -183,7 +191,7 @@ bdb_dn2id_delete(
 
 	DBTzero( &key );
 	key.size = e->e_nname.bv_len + 2;
-	buf = sl_malloc( key.size, ctx );
+	buf = op->o_tmpalloc( key.size, op->o_tmpmemctx );
 	key.data = buf;
 	key.flags = DB_DBT_USERMEM;
 	buf[0] = DN_BASE_PREFIX;
@@ -207,7 +215,7 @@ bdb_dn2id_delete(
 	}
 
 #ifndef BDB_MULTIPLE_SUFFIXES
-	if( !be_issuffix( be, &ptr )) {
+	if( !be_issuffix( op->o_bd, &ptr )) {
 #endif
 		buf[0] = DN_SUBTREE_PREFIX;
 		rc = db->del( db, txn, &key, 0 );
@@ -225,7 +233,7 @@ bdb_dn2id_delete(
 		}
 
 #ifdef BDB_MULTIPLE_SUFFIXES
-	if( !be_issuffix( be, &ptr )) {
+	if( !be_issuffix( op->o_bd, &ptr )) {
 #endif
 		dnParent( &ptr, &pdn );
 
@@ -235,7 +243,7 @@ bdb_dn2id_delete(
 		key.data = pdn.bv_val - 1;
 		ptr = pdn;
 
-		rc = bdb_idl_delete_key( be, db, txn, &key, e->e_id );
+		rc = bdb_idl_delete_key( op->o_bd, db, txn, &key, e->e_id );
 
 		if( rc != 0 ) {
 #ifdef NEW_LOGGING
@@ -252,13 +260,13 @@ bdb_dn2id_delete(
 #ifndef BDB_MULTIPLE_SUFFIXES
 	}
 
-	while( !be_issuffix( be, &ptr )) {
+	while( !be_issuffix( op->o_bd, &ptr )) {
 #else
 	for (;;) {
 #endif
 		ptr.bv_val[-1] = DN_SUBTREE_PREFIX;
 
-		rc = bdb_idl_delete_key( be, db, txn, &key, e->e_id );
+		rc = bdb_idl_delete_key( op->o_bd, db, txn, &key, e->e_id );
 		if( rc != 0 ) {
 #ifdef NEW_LOGGING
 			LDAP_LOG ( INDEX, ERR, 
@@ -272,7 +280,7 @@ bdb_dn2id_delete(
 			goto done;
 		}
 #ifdef BDB_MULTIPLE_SUFFIXES
-		if( be_issuffix( be, &ptr )) break;
+		if( be_issuffix( op->o_bd, &ptr )) break;
 #endif
 		dnParent( &ptr, &pdn );
 
@@ -286,7 +294,7 @@ bdb_dn2id_delete(
 #endif
 
 done:
-	sl_free( buf, ctx );
+	op->o_tmpfree( buf, op->o_tmpmemctx );
 #ifdef NEW_LOGGING
 	LDAP_LOG ( INDEX, RESULTS, "<= bdb_dn2id_delete %d\n", rc, 0, 0 );
 #else
@@ -297,15 +305,14 @@ done:
 
 int
 bdb_dn2id(
-	BackendDB	*be,
+	Operation *op,
 	DB_TXN *txn,
 	struct berval	*dn,
-	EntryInfo *ei,
-	void *ctx )
+	EntryInfo *ei )
 {
 	int		rc;
 	DBT		key, data;
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 
 #ifdef NEW_LOGGING
@@ -313,10 +320,9 @@ bdb_dn2id(
 #else
 	Debug( LDAP_DEBUG_TRACE, "=> bdb_dn2id( \"%s\" )\n", dn->bv_val, 0, 0 );
 #endif
-
 	DBTzero( &key );
 	key.size = dn->bv_len + 2;
-	key.data = sl_malloc( key.size, ctx );
+	key.data = op->o_tmpalloc( key.size, op->o_tmpmemctx );
 	((char *)key.data)[0] = DN_BASE_PREFIX;
 	AC_MEMCPY( &((char *)key.data)[1], dn->bv_val, key.size - 1 );
 
@@ -347,7 +353,7 @@ bdb_dn2id(
 #endif
 	}
 
-	sl_free( key.data, ctx );
+	op->o_tmpfree( key.data, op->o_tmpmemctx );
 	return rc;
 }
 
@@ -372,7 +378,7 @@ bdb_dn2id_children(
 #endif
 	DBTzero( &key );
 	key.size = e->e_nname.bv_len + 2;
-	key.data = sl_malloc( key.size, op->o_tmpmemctx );
+	key.data = op->o_tmpalloc( key.size, op->o_tmpmemctx );
 	((char *)key.data)[0] = DN_ONE_PREFIX;
 	AC_MEMCPY( &((char *)key.data)[1], e->e_nname.bv_val, key.size - 1 );
 
@@ -380,7 +386,7 @@ bdb_dn2id_children(
 	if ( bdb->bi_idl_cache_size ) {
 		rc = bdb_idl_cache_get( bdb, db, &key, NULL );
 		if ( rc != LDAP_NO_SUCH_OBJECT ) {
-			sl_free( key.data, op->o_tmpmemctx );
+			op->o_tmpfree( key.data, op->o_tmpmemctx );
 			return rc;
 		}
 	}
@@ -394,7 +400,7 @@ bdb_dn2id_children(
 	data.dlen = sizeof(id);
 
 	rc = db->get( db, txn, &key, &data, bdb->bi_db_opflags );
-	sl_free( key.data, op->o_tmpmemctx );
+	op->o_tmpfree( key.data, op->o_tmpmemctx );
 
 #ifdef NEW_LOGGING
 	LDAP_LOG ( INDEX, DETAIL1, 
@@ -413,40 +419,42 @@ bdb_dn2id_children(
 
 int
 bdb_dn2idl(
-	BackendDB	*be,
-	struct berval	*dn,
-	int prefix,
-	ID *ids )
+	Operation *op,
+	Entry *e,
+	ID *ids,
+	ID *stack )
 {
 	int		rc;
 	DBT		key;
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
+	int prefix = ( op->ors_scope == LDAP_SCOPE_ONELEVEL )
+		? DN_ONE_PREFIX : DN_SUBTREE_PREFIX;
 
 #ifdef NEW_LOGGING
-	LDAP_LOG ( INDEX, ARGS, 
-		"=> bdb_dn2ididl( \"%s\" )\n", dn->bv_val, 0, 0 );
+	LDAP_LOG ( INDEX, ARGS, "=> bdb_dn2ididl( \"%s\" )\n",
+		e->e_nname.bv_val, 0, 0 );
 #else
-	Debug( LDAP_DEBUG_TRACE, "=> bdb_dn2idl( \"%s\" )\n", dn->bv_val, 0, 0 );
+	Debug( LDAP_DEBUG_TRACE, "=> bdb_dn2idl( \"%s\" )\n",
+		e->e_nname.bv_val, 0, 0 );
 #endif
 
 #ifndef	BDB_MULTIPLE_SUFFIXES
-	if (prefix == DN_SUBTREE_PREFIX && be_issuffix(be, dn))
-	{
+	if ( prefix == DN_SUBTREE_PREFIX && BEI(e)->bei_parent->bei_id == 0 ) {
 		BDB_IDL_ALL(bdb, ids);
 		return 0;
 	}
 #endif
 
 	DBTzero( &key );
-	key.size = dn->bv_len + 2;
+	key.size = e->e_nname.bv_len + 2;
 	key.ulen = key.size;
 	key.flags = DB_DBT_USERMEM;
-	key.data = ch_malloc( key.size );
+	key.data = op->o_tmpalloc( key.size, op->o_tmpmemctx );
 	((char *)key.data)[0] = prefix;
-	AC_MEMCPY( &((char *)key.data)[1], dn->bv_val, key.size - 1 );
+	AC_MEMCPY( &((char *)key.data)[1], e->e_nname.bv_val, key.size - 1 );
 
-	rc = bdb_idl_fetch_key( be, db, NULL, &key, ids );
+	rc = bdb_idl_fetch_key( op->o_bd, db, NULL, &key, ids );
 
 	if( rc != 0 ) {
 #ifdef NEW_LOGGING
@@ -472,7 +480,7 @@ bdb_dn2idl(
 #endif
 	}
 
-	ch_free( key.data );
+	op->o_tmpfree( key.data, op->o_tmpmemctx );
 	return rc;
 }
 #else	/* BDB_HIER */
@@ -480,14 +488,16 @@ bdb_dn2idl(
 /* Experimental management routines for a hierarchically structured database.
  *
  * Unsupported! Use at your own risk!
+ * -- Howard Chu, Symas Corp. 2003.
  *
  * Instead of a ldbm-style dn2id database, we use a hierarchical one. Each
  * entry in this database is a struct diskNode, keyed by entryID and with
  * the data containing the RDN and entryID of the node's children. We use
  * a B-Tree with sorted duplicates to store all the children of a node under
- * the same key. Also, the first item under the key contains an empty rdn
- * and the ID of the node's parent, to allow bottom-up tree traversal as
- * well as top-down.
+ * the same key. Also, the first item under the key contains the entry's own
+ * rdn and the ID of the node's parent, to allow bottom-up tree traversal as
+ * well as top-down. To keep this info first in the list, the nrdnlen is set
+ * to the negative of its value.
  *
  * The diskNode is a variable length structure. This definition is not
  * directly usable for in-memory manipulation.
@@ -500,51 +510,72 @@ typedef struct diskNode {
 } diskNode;
 
 /* Sort function for the sorted duplicate data items of a dn2id key.
- * Sorts based on normalized RDN, in lexical order.
+ * Sorts based on normalized RDN, in length order.
  */
 int
-bdb_hdb_compare(
+hdb_dup_compare(
 	DB *db, 
 	const DBT *usrkey,
 	const DBT *curkey
 )
 {
-	diskNode *usr = usrkey->data;
-	diskNode *cur = curkey->data;
-	short curlen;
-	char *ptr = (char *)&cur->nrdnlen;
-	int rc;
+	char *u = (char *)&(((diskNode *)(usrkey->data))->nrdnlen);
+	char *c = (char *)&(((diskNode *)(curkey->data))->nrdnlen);
+	int rc, i;
 
-	curlen = ptr[0] << 8 | ptr[1];
-
-	rc = strncmp( usr->nrdn, cur->nrdn, usr->nrdnlen );
-	if ( rc == 0 ) rc = usrlen - curlen;
-	return rc;
+	/* data is not aligned, cannot compare directly */
+#ifdef WORDS_BIGENDIAN
+	for( i = 0; i < (int)sizeof(short); i++)
+#else
+	for( i = sizeof(short)-1; i >= 0; i--)
+#endif
+	{
+		rc = u[i] - c[i];
+		if( rc ) return rc;
+	}
+	return strcmp( u+sizeof(short), c+sizeof(short) );
 }
 
 /* This function constructs a full DN for a given entry.
  */
-int bdb_fix_dn(
-	BackendDB *be,
-	Entry *e
+int hdb_fix_dn(
+	Entry *e,
+	int checkit
 )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
 	EntryInfo *ei;
 	int rlen = 0, nrlen = 0;
 	char *ptr, *nptr;
-	
-	for ( ei = BEI(e); ei; ei=ei->bei_parent ) {
+	int max = 0;
+
+	/* count length of all DN components */
+	for ( ei = BEI(e); ei && ei->bei_id; ei=ei->bei_parent ) {
 		rlen += ei->bei_rdn.bv_len + 1;
 		nrlen += ei->bei_nrdn.bv_len + 1;
+		if (ei->bei_modrdns > max) max = ei->bei_modrdns;
+	}
+
+	/* See if the entry DN was invalidated by a subtree rename */
+	if ( checkit ) {
+		if ( BEI(e)->bei_modrdns >= max ) {
+			return 0;
+		}
+		/* We found a mismatch, tell the caller to lock it */
+		if ( checkit == 1 ) {
+			return 1;
+		}
+		/* checkit == 2. do the fix. */
+		free( e->e_name.bv_val );
+		free( e->e_nname.bv_val );
 	}
+
 	e->e_name.bv_len = rlen - 1;
 	e->e_nname.bv_len = nrlen - 1;
-	e->e_name.bv_val = ch_malloc(rlen + nrlen);
-	e->e_nname.bv_val = e->e_name.bv_val + rlen;
+	e->e_name.bv_val = ch_malloc(rlen);
+	e->e_nname.bv_val = ch_malloc(nrlen);
 	ptr = e->e_name.bv_val;
 	nptr = e->e_nname.bv_val;
-	for ( ei = BEI(e); ei; ei=ei->bei_parent ) {
+	for ( ei = BEI(e); ei && ei->bei_id; ei=ei->bei_parent ) {
 		ptr = lutil_strcopy(ptr, ei->bei_rdn.bv_val);
 		nptr = lutil_strcopy(nptr, ei->bei_nrdn.bv_val);
 		if ( ei->bei_parent ) {
@@ -552,8 +583,9 @@ int bdb_fix_dn(
 			*nptr++ = ',';
 		}
 	}
-	*ptr = '\0';
-	*nptr = '\0';
+	BEI(e)->bei_modrdns = max;
+	ptr[-1] = '\0';
+	nptr[-1] = '\0';
 
 	return 0;
 }
@@ -563,28 +595,28 @@ int bdb_fix_dn(
  * child's entryID containing the parent's entryID.
  */
 int
-bdb_dn2id_add(
-	BackendDB	*be,
+hdb_dn2id_add(
+	Operation	*op,
 	DB_TXN *txn,
 	EntryInfo	*eip,
-	Entry		*e,
-	void *ctx )
+	Entry		*e )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 	DBT		key, data;
 	int		rc, rlen, nrlen;
 	diskNode *d;
 	char *ptr;
 
-	nrlen = dn_rdnlen( be, &e->e_nname );
+	nrlen = dn_rdnlen( op->o_bd, &e->e_nname );
 	if (nrlen) {
-		rlen = dn_rdnlen( be, &e->e_name );
+		rlen = dn_rdnlen( op->o_bd, &e->e_name );
 	} else {
-		rlen = 0;
+		nrlen = e->e_nname.bv_len;
+		rlen = e->e_name.bv_len;
 	}
 
-	d = sl_malloc(sizeof(diskNode) + rlen + nrlen, ctx);
+	d = op->o_tmpalloc(sizeof(diskNode) + rlen + nrlen, op->o_tmpmemctx);
 	d->entryID = e->e_id;
 	d->nrdnlen = nrlen;
 	ptr = lutil_strncopy( d->nrdn, e->e_nname.bv_val, nrlen );
@@ -598,42 +630,50 @@ bdb_dn2id_add(
 	key.size = sizeof(ID);
 	key.flags = DB_DBT_USERMEM;
 
+	/* Need to make dummy root node once. Subsequent attempts
+	 * will fail harmlessly.
+	 */
+	if ( eip->bei_id == 0 ) {
+		diskNode dummy = {0};
+		data.data = &dummy;
+		data.size = sizeof(diskNode);
+		data.flags = DB_DBT_USERMEM;
+
+		db->put( db, txn, &key, &data, DB_NODUPDATA );
+	}
+
 #ifdef SLAP_IDL_CACHE
 	if ( bdb->bi_idl_cache_size ) {
 		bdb_idl_cache_del( bdb, db, &key );
 	}
 #endif
 	data.data = d;
-	data.size = sizeof(diskNode) + rlen + nrlen + 2;
+	data.size = sizeof(diskNode) + rlen + nrlen;
 	data.flags = DB_DBT_USERMEM;
 
-	rc = db->put( db, txn, &key, &data, DB_NOOVERWRITE );
+	rc = db->put( db, txn, &key, &data, DB_NODUPDATA );
 
 	if (rc == 0) {
 		key.data = &e->e_id;
 		d->entryID = eip->bei_id;
-		d->nrdnlen = 0;
-		d->nrdn[0] = '\0';
-		d->rdn[0] = '\0';
-		data.size = sizeof(diskNode);
+		d->nrdnlen = 0 - nrlen;
 
-		rc = db->put( db, txn, &key, &data, DB_NOOVERWRITE );
+		rc = db->put( db, txn, &key, &data, DB_NODUPDATA );
 	}
 
-	sl_free( d, ctx );
+	op->o_tmpfree( d, op->o_tmpmemctx );
 
 	return rc;
 }
 
 int
-bdb_dn2id_delete(
-	BackendDB	*be,
+hdb_dn2id_delete(
+	Operation	*op,
 	DB_TXN *txn,
 	EntryInfo	*eip,
-	Entry	*e,
-	void	*ctx )
+	Entry	*e )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 	DBT		key, data;
 	DBC	*cursor;
@@ -647,12 +687,7 @@ bdb_dn2id_delete(
 	key.flags = DB_DBT_USERMEM;
 
 	DBTzero(&data);
-	data.size = sizeof(diskNode) + BEI(e)->nrdn.bv_len;
-	d = sl_malloc( data.size, ctx );
-	d->entryID = e->e_id;
-	d->nrdnlen = BEI(e)->nrdn.bv_len;
-	strcpy( d->nrdn, BEI(e)->nrdn.bv_val );
-	data.data = d;
+	data.size = sizeof(diskNode) + BEI(e)->bei_nrdn.bv_len;
 	data.ulen = data.size;
 	data.dlen = data.size;
 	data.flags = DB_DBT_USERMEM | DB_DBT_PARTIAL;
@@ -665,67 +700,152 @@ bdb_dn2id_delete(
 	rc = db->cursor( db, txn, &cursor, bdb->bi_db_opflags );
 	if ( rc ) return rc;
 
+	d = op->o_tmpalloc( data.size, op->o_tmpmemctx );
+	d->entryID = e->e_id;
+	d->nrdnlen = BEI(e)->bei_nrdn.bv_len;
+	strcpy( d->nrdn, BEI(e)->bei_nrdn.bv_val );
+	data.data = d;
+
+	/* Delete our ID from the parent's list */
 	rc = cursor->c_get( cursor, &key, &data, DB_GET_BOTH | DB_RMW );
 	if ( rc == 0 )
 		rc = cursor->c_del( cursor, 0 );
-	cursor->c_close( cursor );
 
-	key.data = &e->e_id;
-	rc = db->del( db, txn, &key, 0);
-	sl_free( d, ctx );
+	/* Delete our ID from the tree. With sorted duplicates, this
+	 * will leave any child nodes still hanging around. This is OK
+	 * for modrdn, which will add our info back in later.
+	 */
+	if ( rc == 0 ) {
+		key.data = &e->e_id;
+		rc = cursor->c_get( cursor, &key, &data, DB_SET | DB_RMW );
+		if ( rc == 0 )
+			rc = cursor->c_del( cursor, 0 );
+	}
+	cursor->c_close( cursor );
+	op->o_tmpfree( d, op->o_tmpmemctx );
 
 	return rc;
 }
 
+
 int
-bdb_dn2id(
-	BackendDB	*be,
+hdb_dn2id(
+	Operation	*op,
 	DB_TXN *txn,
 	struct berval	*in,
-	EntryInfo	*ei,
-	void *ctx )
+	EntryInfo	*ei )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
 	DB *db = bdb->bi_dn2id->bdi_db;
 	DBT		key, data;
 	DBC	*cursor;
 	int		rc = 0, nrlen;
+	diskNode *d;
 	char	*ptr;
+	ID idp = ei->bei_parent->bei_id;
 
-	nrlen = dn_rdnlen( be, &in );
+	nrlen = dn_rdnlen( op->o_bd, in );
+	if (!nrlen) nrlen = in->bv_len;
 
 	DBTzero(&key);
 	key.size = sizeof(ID);
-	key.data = &eip->bei_id;
+	key.data = &idp;
+	key.ulen = sizeof(ID);
 	key.flags = DB_DBT_USERMEM;
 
 	DBTzero(&data);
 	data.size = sizeof(diskNode) + nrlen;
-	d = sl_malloc( data.size * 3, ctx );
-	d->nrdnlen = nrlen;
-	ptr = lutil_strncopy( d->nrdn, BEI(e)->nrdn.bv_val, nrlen );
-	*ptr = '\0';
-	data.data = d;
 	data.ulen = data.size * 3;
 	data.flags = DB_DBT_USERMEM;
 
 	rc = db->cursor( db, txn, &cursor, bdb->bi_db_opflags );
 	if ( rc ) return rc;
 
+	d = op->o_tmpalloc( data.size * 3, op->o_tmpmemctx );
+	d->nrdnlen = nrlen;
+	ptr = lutil_strncopy( d->nrdn, in->bv_val, nrlen );
+	*ptr = '\0';
+	data.data = d;
+
 	rc = cursor->c_get( cursor, &key, &data, DB_GET_BOTH );
+	if ( rc == 0 ) {
+		ei->bei_id = d->entryID;
+		ei->bei_rdn.bv_len = data.size - sizeof(diskNode) - nrlen;
+		ptr = d->nrdn + nrlen + 1;
+		ber_str2bv( ptr, ei->bei_rdn.bv_len, 1, &ei->bei_rdn );
+		if ( !ei->bei_parent->bei_dkids ) {
+			db_recno_t dkids;
+			/* How many children does the parent have? */
+			/* FIXME: do we need to lock the parent
+			 * entryinfo? Seems safe...
+			 */
+			cursor->c_count( cursor, &dkids, 0 );
+			ei->bei_parent->bei_dkids = dkids;
+		}
+	}
 	cursor->c_close( cursor );
+	op->o_tmpfree( d, op->o_tmpmemctx );
+
+	return rc;
+}
+
+int
+hdb_dn2id_parent(
+	Operation *op,
+	DB_TXN *txn,
+	EntryInfo *ei,
+	ID *idp )
+{
+	struct bdb_info *bdb = (struct bdb_info *) op->o_bd->be_private;
+	DB *db = bdb->bi_dn2id->bdi_db;
+	DBT		key, data;
+	DBC	*cursor;
+	int		rc = 0;
+	diskNode *d;
+	char	*ptr;
+	unsigned char *pt2;
+
+	DBTzero(&key);
+	key.size = sizeof(ID);
+	key.data = &ei->bei_id;
+	key.ulen = sizeof(ID);
+	key.flags = DB_DBT_USERMEM;
+
+	DBTzero(&data);
+	data.flags = DB_DBT_USERMEM;
+
+	rc = db->cursor( db, txn, &cursor, bdb->bi_db_opflags );
 	if ( rc ) return rc;
 
-	AC_MEMCPY( &ei->bei_id, &d->entryID, sizeof(ID) );
-	ei->rdn.bv_len = data.size - sizeof(diskNode) - nrlen;
-	ptr = d->nrdn + nrlen + 1;
-	strcpy( ei->rdn.bv_val, ptr );
+	data.ulen = sizeof(diskNode) + (SLAP_LDAPDN_MAXLEN * 2);
+	d = op->o_tmpalloc( data.ulen, op->o_tmpmemctx );
+	data.data = d;
 
+	rc = cursor->c_get( cursor, &key, &data, DB_SET );
+	if ( rc == 0 ) {
+		if (d->nrdnlen >= 0) {
+			rc = LDAP_OTHER;
+		} else {
+			db_recno_t dkids;
+			*idp = d->entryID;
+			ei->bei_nrdn.bv_len = 0 - d->nrdnlen;
+			ber_str2bv( d->nrdn, ei->bei_nrdn.bv_len, 1, &ei->bei_nrdn );
+			ei->bei_rdn.bv_len = data.size - sizeof(diskNode) -
+				ei->bei_nrdn.bv_len;
+			ptr = d->nrdn + ei->bei_nrdn.bv_len + 1;
+			ber_str2bv( ptr, ei->bei_rdn.bv_len, 1, &ei->bei_rdn );
+			/* How many children does this node have? */
+			cursor->c_count( cursor, &dkids, 0 );
+			ei->bei_dkids = dkids;
+		}
+	}
+	cursor->c_close( cursor );
+	op->o_tmpfree( d, op->o_tmpmemctx );
 	return rc;
 }
 
 int
-bdb_dn2id_children(
+hdb_dn2id_children(
 	Operation *op,
 	DB_TXN *txn,
 	Entry *e )
@@ -747,12 +867,12 @@ bdb_dn2id_children(
 	if ( bdb->bi_idl_cache_size ) {
 		rc = bdb_idl_cache_get( bdb, db, &key, NULL );
 		if ( rc != LDAP_NO_SUCH_OBJECT ) {
-			sl_free( key.data, o->o_tmpmemctx );
 			return rc;
 		}
 	}
 #endif
 	DBTzero(&data);
+	data.data = &d;
 	data.ulen = sizeof(d);
 	data.flags = DB_DBT_USERMEM | DB_DBT_PARTIAL;
 	data.dlen = sizeof(d);
@@ -760,49 +880,262 @@ bdb_dn2id_children(
 	rc = db->cursor( db, txn, &cursor, bdb->bi_db_opflags );
 	if ( rc ) return rc;
 
-	rc = cursor->c_get( cursor, &key, &data, DB_FIRST );
+	rc = cursor->c_get( cursor, &key, &data, DB_SET );
 	if ( rc == 0 ) {
-		rc = cursor->c_get( cursor, &key, &data, DB_NEXT_DUP );
+		db_recno_t dkids;
+		rc = cursor->c_count( cursor, &dkids, 0 );
+		if ( rc == 0 ) {
+			BEI(e)->bei_dkids = dkids;
+			if ( dkids < 2 ) rc = DB_NOTFOUND;
+		}
 	}
 	cursor->c_close( cursor );
 	return rc;
 }
 
-int
-bdb_dn2idl(
-	BackendDB	*be,
-	struct berval	*dn,
-	int prefix,
-	ID *ids )
+/* bdb_dn2idl:
+ * We can't just use bdb_idl_fetch_key because
+ * 1 - our data items are longer than just an entry ID
+ * 2 - our data items are sorted alphabetically by nrdn, not by ID.
+ *
+ * We descend the tree recursively, so we define this cookie
+ * to hold our necessary state information. The bdb_dn2idl_internal
+ * function uses this cookie when calling itself.
+ */
+
+struct dn2id_cookie {
+	struct bdb_info *bdb;
+	DB *db;
+	int prefix;
+	int rc;
+	EntryInfo *ei;
+	ID id;
+	ID dbuf;
+	ID *ids;
+	void *ptr;
+	ID tmp[BDB_IDL_DB_SIZE];
+	ID *buf;
+	DBT key;
+	DBT data;
+	DBC *dbc;
+	Operation *op;
+};
+
+static int
+apply_func(
+	void *data,
+	void *arg )
 {
-	struct bdb_info *bdb = (struct bdb_info *) be->be_private;
-	int		rc;
-	ID		id;
-	idNode		*n;
+	EntryInfo *ei = data;
+	ID *idl = arg;
 
-	if (prefix == DN_SUBTREE_PREFIX && be_issuffix(be, dn)) {
-		BDB_IDL_ALL(bdb, ids);
-		return 0;
+	bdb_idl_insert( idl, ei->bei_id );
+	return 0;
+}
+
+static int
+hdb_dn2idl_internal(
+	struct dn2id_cookie *cx
+)
+{
+#ifdef SLAP_IDL_CACHE
+	if ( cx->bdb->bi_idl_cache_size ) {
+		cx->rc = bdb_idl_cache_get(cx->bdb, cx->db, &cx->key, cx->tmp);
+		if ( cx->rc == DB_NOTFOUND ) {
+			return cx->rc;
+		}
+		if ( cx->rc == LDAP_SUCCESS ) {
+			goto gotit;
+		}
 	}
+#endif
+	BDB_IDL_ZERO( cx->tmp );
 
-	rc = bdb_dn2id(be, NULL, dn, &id, 0);
-	if (rc) return rc;
+	if ( !cx->ei ) {
+		cx->ei = bdb_cache_find_info( cx->bdb, cx->id );
+		if ( !cx->ei ) {
+			cx->rc = DB_NOTFOUND;
+			goto saveit;
+		}
+	}
 
-	ldap_pvt_thread_rdwr_rlock(&bdb->bi_tree_rdwr);
-	n = bdb_find_id_node(id, bdb->bi_tree);
-	ldap_pvt_thread_rdwr_runlock(&bdb->bi_tree_rdwr);
+	bdb_cache_entryinfo_lock( cx->ei );
+
+	/* If number of kids in the cache differs from on-disk, load
+	 * up all the kids from the database
+	 */
+	if ( cx->ei->bei_ckids+1 != cx->ei->bei_dkids ) {
+		EntryInfo ei;
+		db_recno_t dkids = cx->ei->bei_dkids;
+		ei.bei_parent = cx->ei;
+
+		bdb_cache_entryinfo_unlock( cx->ei );
+
+		cx->rc = cx->db->cursor( cx->db, NULL, &cx->dbc,
+			cx->bdb->bi_db_opflags );
+		if ( cx->rc ) return cx->rc;
+
+		cx->data.data = &cx->dbuf;
+		cx->data.ulen = sizeof(ID);
+		cx->data.dlen = sizeof(ID);
+		cx->data.flags = DB_DBT_USERMEM | DB_DBT_PARTIAL;
+
+		/* The first item holds the parent ID. Ignore it. */
+		cx->rc = cx->dbc->c_get( cx->dbc, &cx->key, &cx->data, DB_SET );
+		if ( cx->rc ) {
+			cx->dbc->c_close( cx->dbc );
+			if ( cx->rc == DB_NOTFOUND ) goto saveit;
+			return cx->rc;
+		}
+
+		/* If the on-disk count is zero we've never checked it.
+		 * Count it now.
+		 */
+		if ( !dkids ) {
+			cx->dbc->c_count( cx->dbc, &dkids, 0 );
+			cx->ei->bei_dkids = dkids;
+		}
 
-	ids[0] = 0;
-	ldap_pvt_thread_rdwr_rlock(&n->i_kids_rdwr);
-	if (prefix == DN_ONE_PREFIX) {
-		rc = avl_apply(n->i_kids, insert_one, ids, -1, AVL_INORDER);
+		cx->data.data = cx->buf;
+		cx->data.ulen = BDB_IDL_UM_SIZE * sizeof(ID);
+		cx->data.flags = DB_DBT_USERMEM;
+
+		/* Fetch the rest of the IDs in a loop... */
+		while ( (cx->rc = cx->dbc->c_get( cx->dbc, &cx->key, &cx->data,
+			DB_MULTIPLE | DB_NEXT_DUP )) == 0 ) {
+			u_int8_t *j;
+			size_t len;
+			DB_MULTIPLE_INIT( cx->ptr, &cx->data );
+			while (cx->ptr) {
+				DB_MULTIPLE_NEXT( cx->ptr, &cx->data, j, len );
+				if (j) {
+					EntryInfo *ei2;
+					diskNode *d = (diskNode *)j;
+					short nrlen;
+
+					AC_MEMCPY( &ei.bei_id, &d->entryID, sizeof(ID) );
+					AC_MEMCPY( &nrlen, &d->nrdnlen, sizeof(d->nrdnlen) );
+					ei.bei_nrdn.bv_len = nrlen;
+					/* nrdn/rdn are set in-place.
+					 * hdb_cache_load will copy them as needed
+					 */
+					ei.bei_nrdn.bv_val = d->nrdn;
+					ei.bei_rdn.bv_len = len - sizeof(diskNode) - ei.bei_nrdn.bv_len;
+					ei.bei_rdn.bv_val = d->nrdn + ei.bei_nrdn.bv_len + 1;
+					bdb_idl_insert( cx->tmp, ei.bei_id );
+					hdb_cache_load( cx->bdb, &ei, &ei2 );
+				}
+			}
+		}
+		cx->rc = cx->dbc->c_close( cx->dbc );
 	} else {
-		ids[0] = 1;
-		ids[1] = id;
-		if (n->i_kids)
-			rc = avl_apply(n->i_kids, insert_sub, ids, -1, AVL_INORDER);
+		/* The in-memory cache is in sync with the on-disk data.
+		 * do we have any kids?
+		 */
+		cx->rc = 0;
+		if ( cx->ei->bei_ckids > 0 ) {
+
+			/* Walk the kids tree; order is irrelevant since bdb_idl_insert
+			 * will insert in sorted order.
+			 */
+			avl_apply( cx->ei->bei_kids, apply_func, cx->tmp, -1, AVL_POSTORDER );
+		}
+		bdb_cache_entryinfo_unlock( cx->ei );
 	}
-	ldap_pvt_thread_rdwr_runlock(&n->i_kids_rdwr);
-	return rc;
+
+saveit:
+#ifdef SLAP_IDL_CACHE
+	if ( cx->bdb->bi_idl_cache_max_size ) {
+		bdb_idl_cache_put( cx->bdb, cx->db, &cx->key, cx->tmp, cx->rc );
+	}
+#endif
+	;
+gotit:
+	if ( !BDB_IDL_IS_ZERO( cx->tmp )) {
+		if ( cx->prefix == DN_SUBTREE_PREFIX ) {
+			if (cx->ei->bei_state & CACHE_ENTRY_NO_GRANDKIDS) {
+				bdb_idl_union( cx->ids, cx->tmp );
+			} else {
+				ID *save, idcurs;
+				EntryInfo *ei = cx->ei;
+				int nokids = 1;
+				save = cx->op->o_tmpalloc( BDB_IDL_SIZEOF( cx->tmp ),
+					cx->op->o_tmpmemctx );
+				BDB_IDL_CPY( save, cx->tmp );
+				bdb_idl_union( cx->ids, cx->tmp );
+
+				idcurs = 0;
+				for ( cx->id = bdb_idl_first( save, &idcurs );
+					cx->id != NOID;
+					cx->id = bdb_idl_next( save, &idcurs )) {
+					cx->ei = NULL;
+					hdb_dn2idl_internal( cx );
+					if ( !BDB_IDL_IS_ZERO( cx->tmp ))
+						nokids = 0;
+				}
+				cx->op->o_tmpfree( save, cx->op->o_tmpmemctx );
+				if ( nokids ) ei->bei_state |= CACHE_ENTRY_NO_GRANDKIDS;
+			}
+			/* Make sure caller knows it had kids! */
+			cx->tmp[0]=1;
+
+			cx->rc = 0;
+		} else {
+			BDB_IDL_CPY( cx->ids, cx->tmp );
+		}
+	}
+	return cx->rc;
+}
+
+int
+hdb_dn2idl(
+	Operation	*op,
+	Entry		*e,
+	ID *ids,
+	ID *stack )
+{
+	struct bdb_info *bdb = (struct bdb_info *)op->o_bd->be_private;
+	struct dn2id_cookie cx;
+
+#ifdef NEW_LOGGING
+	LDAP_LOG ( INDEX, ARGS, 
+		"=> hdb_dn2ididl( \"%s\" )\n", e->e_nname.bv_val, 0, 0 );
+#else
+	Debug( LDAP_DEBUG_TRACE, "=> hdb_dn2idl( \"%s\" )\n", e->e_nname.bv_val, 0, 0 );
+#endif
+
+#ifndef BDB_MULTIPLE_SUFFIXES
+	if ( op->ors_scope != LDAP_SCOPE_ONELEVEL && 
+		BEI(e)->bei_parent->bei_id == 0 )
+	{
+		BDB_IDL_ALL( bdb, ids );
+		return 0;
+	}
+#endif
+
+	cx.id = e->e_id;
+	cx.ei = e->e_id ? BEI(e) : &bdb->bi_cache.c_dntree;
+	cx.bdb = bdb;
+	cx.db = cx.bdb->bi_dn2id->bdi_db;
+	cx.prefix = op->ors_scope == LDAP_SCOPE_ONELEVEL
+		? DN_ONE_PREFIX : DN_SUBTREE_PREFIX;
+	cx.ids = ids;
+	cx.buf = stack;
+	cx.op = op;
+
+	BDB_IDL_ZERO( ids );
+	if ( cx.prefix == DN_SUBTREE_PREFIX ) {
+		bdb_idl_insert( ids, cx.id );
+	}
+
+	DBTzero(&cx.key);
+	cx.key.data = &cx.id;
+	cx.key.ulen = sizeof(ID);
+	cx.key.size = sizeof(ID);
+	cx.key.flags = DB_DBT_USERMEM;
+
+	DBTzero(&cx.data);
+
+	return hdb_dn2idl_internal(&cx);
 }
 #endif	/* BDB_HIER */