Adding upstream version 2.5.13+dfsg.upstream/2.5.13+dfsgupstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 1151 insertions, 0 deletions

diff --git a/servers/slapd/back-mdb/id2entry.c b/servers/slapd/back-mdb/id2entry.c
new file mode 100644
index 0000000..aa6067a
--- /dev/null
+++ b/servers/slapd/back-mdb/id2entry.c
@@ -0,0 +1,1151 @@
+/* id2entry.c - routines to deal with the id2entry database */
+/* $OpenLDAP$ */
+/* This work is part of OpenLDAP Software <http://www.openldap.org/>.
+ *
+ * Copyright 2000-2022 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"
+
+#include <stdio.h>
+#include <ac/string.h>
+#include <ac/errno.h>
+
+#include "back-mdb.h"
+
+typedef struct Ecount {
+	ber_len_t len;	/* total entry size */
+	ber_len_t dlen;	/* contiguous data size */
+	int nattrs;
+	int nvals;
+	int offset;
+	Attribute *multi;
+} Ecount;
+
+static int mdb_entry_partsize(struct mdb_info *mdb, MDB_txn *txn, Entry *e,
+	Ecount *eh);
+static int mdb_entry_encode(Operation *op, Entry *e, MDB_val *data,
+	Ecount *ec);
+static Entry *mdb_entry_alloc( Operation *op, int nattrs, int nvals );
+
+#define ID2VKSZ	(sizeof(ID)+2)
+
+int
+mdb_id2v_compare(
+	const MDB_val *usrkey,
+	const MDB_val *curkey
+)
+{
+	unsigned short *uv, *cv;
+	ID ui, ci;
+	int rc;
+
+	memcpy(&ui, usrkey->mv_data, sizeof(ID));
+	memcpy(&ci, curkey->mv_data, sizeof(ID));
+	if (ui < ci)
+		return -1;
+	if (ui > ci)
+		return 1;
+	uv = usrkey->mv_data;
+	cv = curkey->mv_data;
+	return uv[sizeof(ID)/2] - cv[sizeof(ID)/2];
+}
+
+/* usrkey[0] is the key in DB format, as described at mdb_mval_put.
+ * usrkey[1] is the value we'll actually match against.
+ * usrkey[2] is the attributeDescription for this value.
+ */
+int
+mdb_id2v_dupsort(
+	const MDB_val *usrkey,
+	const MDB_val *curkey
+)
+{
+	AttributeDescription *ad = usrkey[2].mv_data;
+	struct berval bv1, bv2;
+	int rc, match, olen;
+	unsigned short s;
+	char *ptr;
+
+	ptr = curkey->mv_data + curkey->mv_size - 2;
+	memcpy(&s, ptr, 2);
+	bv2.bv_val = curkey->mv_data;
+	bv2.bv_len = curkey->mv_size - 3;
+	if (s)
+		bv2.bv_len -= (s+1);
+
+	bv1.bv_val = usrkey[1].mv_data;
+	bv1.bv_len = usrkey[1].mv_size;
+
+	if (ad && ad->ad_type->sat_equality) {
+		MatchingRule *mr = ad->ad_type->sat_equality;
+		rc = mr->smr_match(&match, SLAP_MR_EQUALITY
+		| SLAP_MR_VALUE_OF_ASSERTION_SYNTAX
+		| SLAP_MR_ASSERTED_VALUE_NORMALIZED_MATCH
+		| SLAP_MR_ATTRIBUTE_VALUE_NORMALIZED_MATCH,
+		ad->ad_type->sat_syntax, mr, &bv1, &bv2);
+	} else {
+		match = ber_bvcmp(&bv1, &bv2);
+	}
+
+	return match;
+}
+
+/* Values are stored as
+ * [normalized-value NUL ] original-value NUL 2-byte-len
+ * The trailing 2-byte-len is zero if there is no normalized value.
+ * Otherwise, it is the length of the original-value.
+ */
+int mdb_mval_put(Operation *op, MDB_cursor *mc, ID id, Attribute *a)
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	MDB_val key, data[3];
+	char *buf;
+	char ivk[ID2VKSZ];
+	unsigned i;
+	unsigned short s;
+	int rc, len;
+
+	memcpy(ivk, &id, sizeof(id));
+	s = mdb->mi_adxs[a->a_desc->ad_index];
+	memcpy(ivk+sizeof(ID), &s, 2);
+	key.mv_data = &ivk;
+	key.mv_size = sizeof(ivk);
+	if ((a->a_desc->ad_type->sat_flags & SLAP_AT_ORDERED) || a->a_desc == slap_schema.si_ad_objectClass)
+		data[2].mv_data = NULL;
+	else
+		data[2].mv_data = a->a_desc;
+
+	for (i=0; i<a->a_numvals; i++) {
+		len = a->a_nvals[i].bv_len + 1 + 2;
+		if (a->a_nvals != a->a_vals) {
+			len += a->a_vals[i].bv_len + 1;
+			data[1].mv_data = a->a_nvals[i].bv_val;
+			data[1].mv_size = a->a_nvals[i].bv_len;
+		} else {
+			data[1].mv_data = a->a_vals[i].bv_val;
+			data[1].mv_size = a->a_vals[i].bv_len;
+		}
+		data[0].mv_size = len;
+		buf = op->o_tmpalloc( len, op->o_tmpmemctx );
+		data[0].mv_data = buf;
+		memcpy(buf, a->a_nvals[i].bv_val, a->a_nvals[i].bv_len);
+		buf += a->a_nvals[i].bv_len;
+		*buf++ = 0;
+		if (a->a_nvals != a->a_vals) {
+			s = a->a_vals[i].bv_len;
+			memcpy(buf, a->a_vals[i].bv_val, a->a_vals[i].bv_len);
+			buf += a->a_vals[i].bv_len;
+			*buf++ = 0;
+			memcpy(buf, &s, 2);
+		} else {
+			*buf++ = 0;
+			*buf++ = 0;
+		}
+		rc = mdb_cursor_put(mc, &key, data, 0);
+		op->o_tmpfree( data[0].mv_data, op->o_tmpmemctx );
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
+int mdb_mval_del(Operation *op, MDB_cursor *mc, ID id, Attribute *a)
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	MDB_val key, data[3];
+	char *ptr;
+	char ivk[ID2VKSZ];
+	unsigned i;
+	int rc;
+	unsigned short s;
+
+	memcpy(ivk, &id, sizeof(id));
+	s = mdb->mi_adxs[a->a_desc->ad_index];
+	memcpy(ivk+sizeof(ID), &s, 2);
+	key.mv_data = &ivk;
+	key.mv_size = sizeof(ivk);
+	if ((a->a_desc->ad_type->sat_flags & SLAP_AT_ORDERED) || a->a_desc == slap_schema.si_ad_objectClass)
+		data[2].mv_data = NULL;
+	else
+		data[2].mv_data = a->a_desc;
+
+	if (a->a_numvals) {
+		for (i=0; i<a->a_numvals; i++) {
+			data[0].mv_data = a->a_nvals[i].bv_val;
+			data[0].mv_size = a->a_nvals[i].bv_len+1;
+			if (a->a_nvals != a->a_vals) {
+				data[1].mv_data = a->a_nvals[i].bv_val;
+				data[1].mv_size = a->a_nvals[i].bv_len;
+			} else {
+				data[1].mv_data = a->a_vals[i].bv_val;
+				data[1].mv_size = a->a_vals[i].bv_len;
+			}
+			rc = mdb_cursor_get(mc, &key, data, MDB_GET_BOTH_RANGE);
+			if (rc)
+				return rc;
+			rc = mdb_cursor_del(mc, 0);
+			if (rc)
+				return rc;
+		}
+	} else {
+		rc = mdb_cursor_get(mc, &key, data, MDB_SET);
+		if (rc)
+			return rc;
+		rc = mdb_cursor_del(mc, MDB_NODUPDATA);
+	}
+	return rc;
+}
+
+static int mdb_mval_get(Operation *op, MDB_cursor *mc, ID id, Attribute *a, int have_nvals)
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	MDB_val key, data[3];
+	char *ptr;
+	char ivk[ID2VKSZ];
+	unsigned i;
+	int rc = 0;
+	unsigned short s;
+
+	memcpy(ivk, &id, sizeof(id));
+	s = mdb->mi_adxs[a->a_desc->ad_index];
+	memcpy(ivk+sizeof(ID), &s, 2);
+	key.mv_data = &ivk;
+	key.mv_size = sizeof(ivk);
+
+	/* not needed */
+	if ((a->a_desc->ad_type->sat_flags & SLAP_AT_ORDERED) || a->a_desc == slap_schema.si_ad_objectClass)
+		data[2].mv_data = NULL;
+	else
+		data[2].mv_data = a->a_desc;
+
+	if (have_nvals)
+		a->a_nvals = a->a_vals + a->a_numvals + 1;
+	else
+		a->a_nvals = a->a_vals;
+	for (i=0; i<a->a_numvals; i++) {
+		if (!i)
+			rc = mdb_cursor_get(mc, &key, data, MDB_SET);
+		else
+			rc = mdb_cursor_get(mc, &key, data, MDB_NEXT_DUP);
+		if (rc)
+			break;
+		ptr = (char*)data[0].mv_data + data[0].mv_size - 2;
+		memcpy(&s, ptr, 2);
+		if (have_nvals) {
+			a->a_nvals[i].bv_val = data[0].mv_data;
+			a->a_vals[i].bv_len = s;
+			a->a_vals[i].bv_val = ptr - a->a_vals[i].bv_len - 1;
+			a->a_nvals[i].bv_len = a->a_vals[i].bv_val - a->a_nvals[i].bv_val - 1;
+		} else {
+			assert(!s);
+			a->a_vals[i].bv_val = data[0].mv_data;
+			a->a_vals[i].bv_len = data[0].mv_size - 3;
+		}
+	}
+	a->a_numvals = i;
+	BER_BVZERO(&a->a_vals[i]);
+	if (have_nvals) {
+		BER_BVZERO(&a->a_nvals[i]);
+	}
+	return rc;
+}
+
+#define ADD_FLAGS	(MDB_NOOVERWRITE|MDB_APPEND)
+
+static int mdb_id2entry_put(
+	Operation *op,
+	MDB_txn *txn,
+	MDB_cursor *mc,
+	Entry *e,
+	int flag )
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	Ecount ec;
+	MDB_val key, data;
+	int rc, adding = flag, prev_ads = mdb->mi_numads;
+
+	/* We only store rdns, and they go in the dn2id database. */
+
+	key.mv_data = &e->e_id;
+	key.mv_size = sizeof(ID);
+
+	rc = mdb_entry_partsize( mdb, txn, e, &ec );
+	if (rc) {
+		rc = LDAP_OTHER;
+		goto fail;
+	}
+
+	flag |= MDB_RESERVE;
+
+	if (e->e_id < mdb->mi_nextid)
+		flag &= ~MDB_APPEND;
+
+	if (mdb->mi_maxentrysize && ec.len > mdb->mi_maxentrysize) {
+		rc = LDAP_ADMINLIMIT_EXCEEDED;
+		goto fail;
+	}
+
+again:
+	data.mv_size = ec.dlen;
+	if ( mc )
+		rc = mdb_cursor_put( mc, &key, &data, flag );
+	else
+		rc = mdb_put( txn, mdb->mi_id2entry, &key, &data, flag );
+	if (rc == MDB_SUCCESS) {
+		rc = mdb_entry_encode( op, e, &data, &ec );
+		if( rc != LDAP_SUCCESS )
+			goto fail;
+		/* Handle adds of large multi-valued attrs here.
+		 * Modifies handle them directly.
+		 */
+		if (adding && ec.multi) {
+			MDB_cursor *mvc;
+			Attribute *a;
+			rc = mdb_cursor_open( txn, mdb->mi_dbis[MDB_ID2VAL], &mvc );
+			if( !rc ) {
+				for ( a = ec.multi; a; a=a->a_next ) {
+					if (!(a->a_flags & SLAP_ATTR_BIG_MULTI))
+						continue;
+					rc = mdb_mval_put( op, mvc, e->e_id, a );
+					if( rc )
+						break;
+				}
+				mdb_cursor_close( mvc );
+			}
+			if ( rc ) {
+				Debug( LDAP_DEBUG_ANY,
+					"mdb_id2entry_put: mdb_mval_put failed: %s(%d) \"%s\"\n",
+					mdb_strerror(rc), rc,
+					e->e_nname.bv_val );
+				rc = LDAP_OTHER;
+				goto fail;
+			}
+		}
+	}
+	if (rc) {
+		/* Was there a hole from slapadd? */
+		if ( (flag & MDB_NOOVERWRITE) && data.mv_size == 0 ) {
+			flag &= ~ADD_FLAGS;
+			goto again;
+		}
+		Debug( LDAP_DEBUG_ANY,
+			"mdb_id2entry_put: mdb_put failed: %s(%d) \"%s\"\n",
+			mdb_strerror(rc), rc,
+			e->e_nname.bv_val );
+		if ( rc != MDB_KEYEXIST )
+			rc = LDAP_OTHER;
+	}
+fail:
+	if (rc) {
+		mdb_ad_unwind( mdb, prev_ads );
+	}
+	return rc;
+}
+
+/*
+ * This routine adds (or updates) an entry on disk.
+ */
+int mdb_id2entry_add(
+	Operation *op,
+	MDB_txn *txn,
+	MDB_cursor *mc,
+	Entry *e )
+{
+	return mdb_id2entry_put(op, txn, mc, e, ADD_FLAGS);
+}
+
+int mdb_id2entry_update(
+	Operation *op,
+	MDB_txn *txn,
+	MDB_cursor *mc,
+	Entry *e )
+{
+	return mdb_id2entry_put(op, txn, mc, e, 0);
+}
+
+int mdb_id2edata(
+	Operation *op,
+	MDB_cursor *mc,
+	ID id,
+	MDB_val *data )
+{
+	MDB_val key;
+	int rc;
+
+	key.mv_data = &id;
+	key.mv_size = sizeof(ID);
+
+	/* fetch it */
+	rc = mdb_cursor_get( mc, &key, data, MDB_SET );
+	/* stubs from missing parents - DB is actually invalid */
+	if ( rc == MDB_SUCCESS && !data->mv_size )
+		rc = MDB_NOTFOUND;
+	return rc;
+}
+
+int mdb_id2entry(
+	Operation *op,
+	MDB_cursor *mc,
+	ID id,
+	Entry **e )
+{
+	MDB_val key, data;
+	int rc = 0;
+
+	*e = NULL;
+
+	key.mv_data = &id;
+	key.mv_size = sizeof(ID);
+
+	/* fetch it */
+	rc = mdb_cursor_get( mc, &key, &data, MDB_SET );
+	if ( rc == MDB_NOTFOUND ) {
+		/* Looking for root entry on an empty-dn suffix? */
+		if ( !id && BER_BVISEMPTY( &op->o_bd->be_nsuffix[0] )) {
+			struct berval gluebv = BER_BVC("glue");
+			Entry *r = mdb_entry_alloc(op, 2, 4);
+			Attribute *a = r->e_attrs;
+			struct berval *bptr;
+
+			r->e_id = 0;
+			r->e_ocflags = SLAP_OC_GLUE|SLAP_OC__END;
+			bptr = a->a_vals;
+			a->a_flags = SLAP_ATTR_DONT_FREE_DATA | SLAP_ATTR_DONT_FREE_VALS;
+			a->a_desc = slap_schema.si_ad_objectClass;
+			a->a_nvals = a->a_vals;
+			a->a_numvals = 1;
+			*bptr++ = gluebv;
+			BER_BVZERO(bptr);
+			bptr++;
+			a->a_next = a+1;
+			a = a->a_next;
+			a->a_flags = SLAP_ATTR_DONT_FREE_DATA | SLAP_ATTR_DONT_FREE_VALS;
+			a->a_desc = slap_schema.si_ad_structuralObjectClass;
+			a->a_vals = bptr;
+			a->a_nvals = a->a_vals;
+			a->a_numvals = 1;
+			*bptr++ = gluebv;
+			BER_BVZERO(bptr);
+			a->a_next = NULL;
+			*e = r;
+			return MDB_SUCCESS;
+		}
+	}
+	/* stubs from missing parents - DB is actually invalid */
+	if ( rc == MDB_SUCCESS && !data.mv_size )
+		rc = MDB_NOTFOUND;
+	if ( rc ) return rc;
+
+	rc = mdb_entry_decode( op, mdb_cursor_txn( mc ), &data, id, e );
+	if ( rc ) return rc;
+
+	(*e)->e_id = id;
+	(*e)->e_name.bv_val = NULL;
+	(*e)->e_nname.bv_val = NULL;
+
+	return rc;
+}
+
+int mdb_id2entry_delete(
+	BackendDB *be,
+	MDB_txn *tid,
+	Entry *e )
+{
+	struct mdb_info *mdb = (struct mdb_info *) be->be_private;
+	MDB_dbi dbi = mdb->mi_id2entry;
+	MDB_val key;
+	MDB_cursor *mvc;
+	char kbuf[sizeof(ID) + sizeof(unsigned short)];
+	int rc;
+
+	memcpy( kbuf, &e->e_id, sizeof(ID) );
+	memset( kbuf+sizeof(ID), 0, sizeof(unsigned short) );
+	key.mv_data = kbuf;
+	key.mv_size = sizeof(kbuf);
+
+	/* delete from database */
+	rc = mdb_del( tid, dbi, &key, NULL );
+	if (rc)
+		return rc;
+	rc = mdb_cursor_open( tid, mdb->mi_dbis[MDB_ID2VAL], &mvc );
+	if (rc)
+		return rc;
+
+	rc = mdb_cursor_get( mvc, &key, NULL, MDB_SET_RANGE );
+	if (rc) {
+		if (rc == MDB_NOTFOUND)
+			rc = MDB_SUCCESS;
+		return rc;
+	}
+	while (*(ID *)key.mv_data == e->e_id ) {
+		rc = mdb_cursor_del( mvc, MDB_NODUPDATA );
+		if (rc)
+			return rc;
+		rc = mdb_cursor_get( mvc, &key, NULL, MDB_GET_CURRENT );
+		if (rc) {
+			/* no record or DB is empty */
+			if (rc == MDB_NOTFOUND || rc == EINVAL)
+				rc = MDB_SUCCESS;
+			break;
+		}
+	}
+	return rc;
+}
+
+static Entry * mdb_entry_alloc(
+	Operation *op,
+	int nattrs,
+	int nvals )
+{
+	Entry *e = op->o_tmpalloc( sizeof(Entry) +
+		nattrs * sizeof(Attribute) +
+		nvals * sizeof(struct berval), op->o_tmpmemctx );
+	BER_BVZERO(&e->e_bv);
+	e->e_private = e;
+	if (nattrs) {
+		e->e_attrs = (Attribute *)(e+1);
+		e->e_attrs->a_vals = (struct berval *)(e->e_attrs+nattrs);
+	} else {
+		e->e_attrs = NULL;
+	}
+
+	return e;
+}
+
+int mdb_entry_return(
+	Operation *op,
+	Entry *e
+)
+{
+	if ( !e )
+		return 0;
+	if ( e->e_private ) {
+		if ( op->o_hdr && op->o_tmpmfuncs ) {
+			op->o_tmpfree( e->e_nname.bv_val, op->o_tmpmemctx );
+			op->o_tmpfree( e->e_name.bv_val, op->o_tmpmemctx );
+			op->o_tmpfree( e, op->o_tmpmemctx );
+		} else {
+			ch_free( e->e_nname.bv_val );
+			ch_free( e->e_name.bv_val );
+			ch_free( e );
+		}
+	} else {
+		entry_free( e );
+	}
+	return 0;
+}
+
+int mdb_entry_release(
+	Operation *op,
+	Entry *e,
+	int rw )
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	struct mdb_op_info *moi = NULL;
+ 
+	/* slapMode : SLAP_SERVER_MODE, SLAP_TOOL_MODE,
+			SLAP_TRUNCATE_MODE, SLAP_UNDEFINED_MODE */
+ 
+	int release = 1;
+	if ( slapMode & SLAP_SERVER_MODE ) {
+		OpExtra *oex;
+		LDAP_SLIST_FOREACH( oex, &op->o_extra, oe_next ) {
+			release = 0;
+			if ( oex->oe_key == mdb ) {
+				mdb_entry_return( op, e );
+				moi = (mdb_op_info *)oex;
+				/* If it was setup by entry_get we should probably free it */
+				if (( moi->moi_flag & (MOI_FREEIT|MOI_KEEPER)) == MOI_FREEIT ) {
+					moi->moi_ref--;
+					if ( moi->moi_ref < 1 ) {
+						mdb_txn_reset( moi->moi_txn );
+						moi->moi_ref = 0;
+						LDAP_SLIST_REMOVE( &op->o_extra, &moi->moi_oe, OpExtra, oe_next );
+						op->o_tmpfree( moi, op->o_tmpmemctx );
+					}
+				}
+				break;
+			}
+		}
+	}
+
+	if (release)
+		mdb_entry_return( op, e );
+ 
+	return 0;
+}
+
+/* return LDAP_SUCCESS IFF we can retrieve the specified entry.
+ */
+int mdb_entry_get(
+	Operation *op,
+	struct berval *ndn,
+	ObjectClass *oc,
+	AttributeDescription *at,
+	int rw,
+	Entry **ent )
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	struct mdb_op_info *moi = NULL;
+	MDB_txn *txn = NULL;
+	Entry *e = NULL;
+	int	rc;
+	const char *at_name = at ? at->ad_cname.bv_val : "(null)";
+
+	Debug( LDAP_DEBUG_ARGS,
+		"=> mdb_entry_get: ndn: \"%s\"\n", ndn->bv_val );
+	Debug( LDAP_DEBUG_ARGS,
+		"=> mdb_entry_get: oc: \"%s\", at: \"%s\"\n",
+		oc ? oc->soc_cname.bv_val : "(null)", at_name );
+
+	rc = mdb_opinfo_get( op, mdb, rw == 0, &moi );
+	if ( rc )
+		return LDAP_OTHER;
+	txn = moi->moi_txn;
+
+	/* can we find entry */
+	rc = mdb_dn2entry( op, txn, NULL, ndn, &e, NULL, 0 );
+	switch( rc ) {
+	case MDB_NOTFOUND:
+	case 0:
+		break;
+	default:
+		return (rc != LDAP_BUSY) ? LDAP_OTHER : LDAP_BUSY;
+	}
+	if (e == NULL) {
+		Debug( LDAP_DEBUG_ACL,
+			"=> mdb_entry_get: cannot find entry: \"%s\"\n",
+				ndn->bv_val );
+		rc = LDAP_NO_SUCH_OBJECT;
+		goto return_results;
+	}
+	
+	Debug( LDAP_DEBUG_ACL,
+		"=> mdb_entry_get: found entry: \"%s\"\n",
+		ndn->bv_val );
+
+	if ( oc && !is_entry_objectclass( e, oc, 0 )) {
+		Debug( LDAP_DEBUG_ACL,
+			"<= mdb_entry_get: failed to find objectClass %s\n",
+			oc->soc_cname.bv_val );
+		rc = LDAP_NO_SUCH_ATTRIBUTE;
+		goto return_results;
+	}
+
+	/* NOTE: attr_find() or attrs_find()? */
+	if ( at && attr_find( e->e_attrs, at ) == NULL ) {
+		Debug( LDAP_DEBUG_ACL,
+			"<= mdb_entry_get: failed to find attribute %s\n",
+			at->ad_cname.bv_val );
+		rc = LDAP_NO_SUCH_ATTRIBUTE;
+		goto return_results;
+	}
+
+return_results:
+	if( rc != LDAP_SUCCESS ) {
+		/* free entry */
+		mdb_entry_release( op, e, rw );
+	} else {
+		*ent = e;
+	}
+
+	Debug( LDAP_DEBUG_TRACE,
+		"mdb_entry_get: rc=%d\n",
+		rc );
+	return(rc);
+}
+
+static void
+mdb_reader_free( void *key, void *data )
+{
+	MDB_txn *txn = data;
+
+	if ( txn ) mdb_txn_abort( txn );
+}
+
+/* free up any keys used by the main thread */
+void
+mdb_reader_flush( MDB_env *env )
+{
+	void *data;
+	void *ctx = ldap_pvt_thread_pool_context();
+
+	if ( !ldap_pvt_thread_pool_getkey( ctx, env, &data, NULL ) ) {
+		ldap_pvt_thread_pool_setkey( ctx, env, NULL, 0, NULL, NULL );
+		mdb_reader_free( env, data );
+	}
+}
+
+extern MDB_txn *mdb_tool_txn;
+
+int
+mdb_opinfo_get( Operation *op, struct mdb_info *mdb, int rdonly, mdb_op_info **moip )
+{
+	int rc, renew = 0;
+	void *data;
+	void *ctx;
+	mdb_op_info *moi = NULL;
+	OpExtra *oex;
+
+	assert( op != NULL );
+
+	if ( !mdb || !moip ) return -1;
+
+	/* If no op was provided, try to find the ctx anyway... */
+	if ( op ) {
+		ctx = op->o_threadctx;
+	} else {
+		ctx = ldap_pvt_thread_pool_context();
+	}
+
+	if ( op ) {
+		LDAP_SLIST_FOREACH( oex, &op->o_extra, oe_next ) {
+			if ( oex->oe_key == mdb ) break;
+		}
+		moi = (mdb_op_info *)oex;
+	}
+
+	if ( !moi ) {
+		moi = *moip;
+
+		if ( !moi ) {
+			if ( op ) {
+				moi = op->o_tmpalloc(sizeof(struct mdb_op_info),op->o_tmpmemctx);
+			} else {
+				moi = ch_malloc(sizeof(mdb_op_info));
+			}
+			moi->moi_flag = MOI_FREEIT;
+			*moip = moi;
+		}
+		LDAP_SLIST_INSERT_HEAD( &op->o_extra, &moi->moi_oe, oe_next );
+		moi->moi_oe.oe_key = mdb;
+		moi->moi_ref = 0;
+		moi->moi_txn = NULL;
+	}
+
+	if ( !rdonly ) {
+		/* This op started as a reader, but now wants to write. */
+		if ( moi->moi_flag & MOI_READER ) {
+			moi = *moip;
+			LDAP_SLIST_INSERT_HEAD( &op->o_extra, &moi->moi_oe, oe_next );
+		} else {
+		/* This op is continuing an existing write txn */
+			*moip = moi;
+		}
+		moi->moi_ref++;
+		if ( !moi->moi_txn ) {
+			if (( slapMode & SLAP_TOOL_MODE ) && mdb_tool_txn ) {
+				moi->moi_txn = mdb_tool_txn;
+			} else {
+				int flag = 0;
+#ifdef SLAP_CONTROL_X_LAZY_COMMIT
+				if ( get_lazyCommit( op ))
+					flag |= MDB_NOMETASYNC;
+#endif
+				rc = mdb_txn_begin( mdb->mi_dbenv, NULL, flag, &moi->moi_txn );
+				if (rc) {
+					Debug( LDAP_DEBUG_ANY, "mdb_opinfo_get: err %s(%d)\n",
+						mdb_strerror(rc), rc );
+				}
+				return rc;
+			}
+		}
+		return 0;
+	}
+
+	/* OK, this is a reader */
+	if ( !moi->moi_txn ) {
+		if (( slapMode & SLAP_TOOL_MODE ) && mdb_tool_txn ) {
+			moi->moi_txn = mdb_tool_txn;
+			goto ok;
+		}
+		if ( !ctx ) {
+			/* Shouldn't happen unless we're single-threaded */
+			rc = mdb_txn_begin( mdb->mi_dbenv, NULL, MDB_RDONLY, &moi->moi_txn );
+			if (rc) {
+				Debug( LDAP_DEBUG_ANY, "mdb_opinfo_get: err %s(%d)\n",
+					mdb_strerror(rc), rc );
+			}
+			return rc;
+		}
+		if ( ldap_pvt_thread_pool_getkey( ctx, mdb->mi_dbenv, &data, NULL ) ) {
+			int retried = 0;
+retry:
+			rc = mdb_txn_begin( mdb->mi_dbenv, NULL, MDB_RDONLY, &moi->moi_txn );
+			if (rc == MDB_READERS_FULL && !retried) {
+				int dead;
+				/* if any stale readers were cleared, a slot should be available */
+				if (!mdb_reader_check( mdb->mi_dbenv, &dead ) && dead) {
+					retried = 1;
+					goto retry;
+				}
+			}
+			if (rc) {
+				Debug( LDAP_DEBUG_ANY, "mdb_opinfo_get: err %s(%d)\n",
+					mdb_strerror(rc), rc );
+				return rc;
+			}
+			data = moi->moi_txn;
+			if ( ( rc = ldap_pvt_thread_pool_setkey( ctx, mdb->mi_dbenv,
+				data, mdb_reader_free, NULL, NULL ) ) ) {
+				mdb_txn_abort( moi->moi_txn );
+				moi->moi_txn = NULL;
+				Debug( LDAP_DEBUG_ANY, "mdb_opinfo_get: thread_pool_setkey failed err (%d)\n",
+					rc );
+				return rc;
+			}
+		} else {
+			moi->moi_txn = data;
+			renew = 1;
+		}
+		moi->moi_flag |= MOI_READER;
+	}
+ok:
+	if ( moi->moi_ref < 1 ) {
+		moi->moi_ref = 0;
+	}
+	if ( renew ) {
+		rc = mdb_txn_renew( moi->moi_txn );
+		assert(!rc);
+	}
+	moi->moi_ref++;
+	if ( *moip != moi )
+		*moip = moi;
+
+	return 0;
+}
+
+int mdb_txn( Operation *op, int txnop, OpExtra **ptr )
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	mdb_op_info **moip = (mdb_op_info **)ptr, *moi = *moip;
+	int rc;
+
+	switch( txnop ) {
+	case SLAP_TXN_BEGIN:
+		rc = mdb_opinfo_get( op, mdb, 0, moip );
+		if ( !rc ) {
+			moi = *moip;
+			moi->moi_flag |= MOI_KEEPER;
+		}
+		return rc;
+	case SLAP_TXN_COMMIT:
+		rc = mdb_txn_commit( moi->moi_txn );
+		if ( rc )
+			mdb->mi_numads = 0;
+		op->o_tmpfree( moi, op->o_tmpmemctx );
+		return rc;
+	case SLAP_TXN_ABORT:
+		mdb->mi_numads = 0;
+		mdb_txn_abort( moi->moi_txn );
+		op->o_tmpfree( moi, op->o_tmpmemctx );
+		return 0;
+	}
+	return LDAP_OTHER;
+}
+
+/* Count up the sizes of the components of an entry */
+static int mdb_entry_partsize(struct mdb_info *mdb, MDB_txn *txn, Entry *e,
+	Ecount *eh)
+{
+	ber_len_t len, dlen;
+	int i, nat = 0, nval = 0, nnval = 0, doff = 0;
+	Attribute *a;
+	unsigned hi;
+
+	eh->multi = NULL;
+	len = 4*sizeof(int);	/* nattrs, nvals, ocflags, offset */
+	dlen = len;
+	for (a=e->e_attrs; a; a=a->a_next) {
+		/* For AttributeDesc, we only store the attr index */
+		nat++;
+		if (a->a_desc->ad_index >= MDB_MAXADS) {
+			Debug( LDAP_DEBUG_ANY, "mdb_entry_partsize: too many AttributeDescriptions used\n" );
+			return LDAP_OTHER;
+		}
+		if (!mdb->mi_adxs[a->a_desc->ad_index]) {
+			int rc = mdb_ad_get(mdb, txn, a->a_desc);
+			if (rc)
+				return rc;
+		}
+		len += 2*sizeof(int);	/* AD index, numvals */
+		dlen += 2*sizeof(int);
+		nval += a->a_numvals + 1;	/* empty berval at end */
+		mdb_attr_multi_thresh( mdb, a->a_desc, &hi, NULL );
+		if (a->a_numvals > hi)
+			a->a_flags |= SLAP_ATTR_BIG_MULTI;
+		if (a->a_flags & SLAP_ATTR_BIG_MULTI)
+			doff += a->a_numvals;
+		for (i=0; i<a->a_numvals; i++) {
+			int alen = a->a_vals[i].bv_len + 1 + sizeof(int);	/* len */
+			len += alen;
+			if (a->a_flags & SLAP_ATTR_BIG_MULTI) {
+				if (!eh->multi)
+					eh->multi = a;
+			} else {
+				dlen += alen;
+			}
+		}
+		if (a->a_nvals != a->a_vals) {
+			nval += a->a_numvals + 1;
+			nnval++;
+			if (a->a_flags & SLAP_ATTR_BIG_MULTI)
+				doff += a->a_numvals;
+			for (i=0; i<a->a_numvals; i++) {
+				int alen = a->a_nvals[i].bv_len + 1 + sizeof(int);
+				len += alen;
+				if (!(a->a_flags & SLAP_ATTR_BIG_MULTI))
+					dlen += alen;
+			}
+		}
+	}
+	/* padding */
+	dlen = (dlen + sizeof(ID)-1) & ~(sizeof(ID)-1);
+	eh->len = len;
+	eh->dlen = dlen;
+	eh->nattrs = nat;
+	eh->nvals = nval;
+	eh->offset = nat + nval - nnval - doff;
+	return 0;
+}
+
+/* Flag bits for an encoded attribute */
+#define MDB_AT_SORTED	(1U<<(sizeof(unsigned int)*CHAR_BIT-1))
+	/* the values are in sorted order */
+#define MDB_AT_MULTI	(1<<(sizeof(unsigned int)*CHAR_BIT-2))
+	/* the values of this multi-valued attr are stored separately */
+
+#define MDB_AT_NVALS	(1U<<(sizeof(unsigned int)*CHAR_BIT-1))
+	/* this attribute has normalized values */
+
+/* Flatten an Entry into a buffer. The buffer starts with the count of the
+ * number of attributes in the entry, the total number of values in the
+ * entry, and the e_ocflags. It then contains a list of integers for each
+ * attribute. For each attribute the first integer gives the index of the
+ * matching AttributeDescription, followed by the number of values in the
+ * attribute. If the MDB_AT_SORTED bit of the attr index is set, the
+ * attribute's values are already sorted. If the MDB_AT_MULTI bit of the
+ * attr index is set, the values are stored separately.
+ *
+ * If the MDB_AT_NVALS bit of numvals is set, the attribute also has
+ * normalized values present. (Note - a_numvals is an unsigned int, so this
+ * means it's possible to receive an attribute that we can't encode due
+ * to size overflow. In practice, this should not be an issue.)
+ *
+ * Then the length of each value is listed. If there are normalized values,
+ * their lengths come next. This continues for each attribute. After all
+ * of the lengths for the last attribute, the actual values are copied,
+ * with a NUL terminator after each value.
+ * The buffer is padded to the sizeof(ID). The entire buffer size is
+ * precomputed so that a single malloc can be performed.
+ */
+static int mdb_entry_encode(Operation *op, Entry *e, MDB_val *data, Ecount *eh)
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	ber_len_t i;
+	Attribute *a;
+	unsigned char *ptr;
+	unsigned int *lp, l;
+
+	Debug( LDAP_DEBUG_TRACE, "=> mdb_entry_encode(0x%08lx): %s\n",
+		(long) e->e_id, e->e_dn );
+
+	/* make sure e->e_ocflags is set */
+	if (is_entry_referral(e))
+		;	/* empty */
+
+	lp = (unsigned int *)data->mv_data;
+	*lp++ = eh->nattrs;
+	*lp++ = eh->nvals;
+	*lp++ = (unsigned int)e->e_ocflags;
+	*lp++ = eh->offset;
+	ptr = (unsigned char *)(lp + eh->offset);
+
+	for (a=e->e_attrs; a; a=a->a_next) {
+		if (!a->a_desc->ad_index)
+			return LDAP_UNDEFINED_TYPE;
+		l = mdb->mi_adxs[a->a_desc->ad_index];
+		if (a->a_flags & SLAP_ATTR_BIG_MULTI)
+			l |= MDB_AT_MULTI;
+		if (a->a_flags & SLAP_ATTR_SORTED_VALS)
+			l |= MDB_AT_SORTED;
+		*lp++ = l;
+		l = a->a_numvals;
+		if (a->a_nvals != a->a_vals)
+			l |= MDB_AT_NVALS;
+		*lp++ = l;
+		if (a->a_flags & SLAP_ATTR_BIG_MULTI) {
+			continue;
+		} else {
+			if (a->a_vals) {
+				for (i=0; a->a_vals[i].bv_val; i++);
+				assert( i == a->a_numvals );
+				for (i=0; i<a->a_numvals; i++) {
+					*lp++ = a->a_vals[i].bv_len;
+					memcpy(ptr, a->a_vals[i].bv_val,
+						a->a_vals[i].bv_len);
+					ptr += a->a_vals[i].bv_len;
+					*ptr++ = '\0';
+				}
+				if (a->a_nvals != a->a_vals) {
+					for (i=0; i<a->a_numvals; i++) {
+						*lp++ = a->a_nvals[i].bv_len;
+						memcpy(ptr, a->a_nvals[i].bv_val,
+							a->a_nvals[i].bv_len);
+						ptr += a->a_nvals[i].bv_len;
+						*ptr++ = '\0';
+					}
+				}
+			}
+		}
+	}
+
+	Debug( LDAP_DEBUG_TRACE, "<= mdb_entry_encode(0x%08lx): %s\n",
+		(long) e->e_id, e->e_dn );
+
+	return 0;
+}
+
+/* Retrieve an Entry that was stored using entry_encode above.
+ *
+ * Note: everything is stored in a single contiguous block, so
+ * you can not free individual attributes or names from this
+ * structure. Attempting to do so will likely corrupt memory.
+ */
+
+int mdb_entry_decode(Operation *op, MDB_txn *txn, MDB_val *data, ID id, Entry **e)
+{
+	struct mdb_info *mdb = (struct mdb_info *) op->o_bd->be_private;
+	int i, j, nattrs, nvals;
+	int rc;
+	Attribute *a;
+	Entry *x;
+	const char *text;
+	unsigned int *lp = (unsigned int *)data->mv_data;
+	unsigned char *ptr;
+	BerVarray bptr;
+	MDB_cursor *mvc = NULL;
+
+	Debug( LDAP_DEBUG_TRACE,
+		"=> mdb_entry_decode:\n" );
+
+	nattrs = *lp++;
+	nvals = *lp++;
+	x = mdb_entry_alloc(op, nattrs, nvals);
+	x->e_ocflags = *lp++;
+	if (!nvals) {
+		goto done;
+	}
+	a = x->e_attrs;
+	bptr = a->a_vals;
+	i = *lp++;
+	ptr = (unsigned char *)(lp + i);
+
+	for (;nattrs>0; nattrs--) {
+		int have_nval = 0, multi = 0;
+		a->a_flags = SLAP_ATTR_DONT_FREE_DATA | SLAP_ATTR_DONT_FREE_VALS;
+		i = *lp++;
+		if (i & MDB_AT_SORTED) {
+			i ^= MDB_AT_SORTED;
+			a->a_flags |= SLAP_ATTR_SORTED_VALS;
+		}
+		if (i & MDB_AT_MULTI) {
+			i ^= MDB_AT_MULTI;
+			a->a_flags |= SLAP_ATTR_BIG_MULTI;
+			multi = 1;
+		}
+		if (i > mdb->mi_numads) {
+			rc = mdb_ad_read(mdb, txn);
+			if (rc)
+				goto leave;
+			if (i > mdb->mi_numads) {
+				Debug( LDAP_DEBUG_ANY,
+					"mdb_entry_decode: attribute index %d not recognized\n",
+					i );
+				rc = LDAP_OTHER;
+				goto leave;
+			}
+		}
+		a->a_desc = mdb->mi_ads[i];
+		a->a_numvals = *lp++;
+		if (a->a_numvals & MDB_AT_NVALS) {
+			a->a_numvals ^= MDB_AT_NVALS;
+			have_nval = 1;
+		}
+		a->a_vals = bptr;
+		if (multi) {
+			if (!mvc) {
+				rc = mdb_cursor_open(txn, mdb->mi_dbis[MDB_ID2VAL], &mvc);
+				if (rc)
+					goto leave;
+			}
+			i = a->a_numvals;
+			mdb_mval_get(op, mvc, id, a, have_nval);
+			bptr += i + 1;
+			if (have_nval)
+				bptr += i + 1;
+		} else {
+			for (i=0; i<a->a_numvals; i++) {
+				bptr->bv_len = *lp++;
+				bptr->bv_val = (char *)ptr;
+				ptr += bptr->bv_len+1;
+				bptr++;
+			}
+			bptr->bv_val = NULL;
+			bptr->bv_len = 0;
+			bptr++;
+
+			if (have_nval) {
+				a->a_nvals = bptr;
+				for (i=0; i<a->a_numvals; i++) {
+					bptr->bv_len = *lp++;
+					bptr->bv_val = (char *)ptr;
+					ptr += bptr->bv_len+1;
+					bptr++;
+				}
+				bptr->bv_val = NULL;
+				bptr->bv_len = 0;
+				bptr++;
+			} else {
+				a->a_nvals = a->a_vals;
+			}
+		}
+
+		/* FIXME: This is redundant once a sorted entry is saved into the DB */
+		if (( a->a_desc->ad_type->sat_flags & SLAP_AT_SORTED_VAL )
+			&& !(a->a_flags & SLAP_ATTR_SORTED_VALS)) {
+			rc = slap_sort_vals( (Modifications *)a, &text, &j, NULL );
+			if ( rc == LDAP_SUCCESS ) {
+				a->a_flags |= SLAP_ATTR_SORTED_VALS;
+			} else if ( rc == LDAP_TYPE_OR_VALUE_EXISTS ) {
+				/* should never happen */
+				Debug( LDAP_DEBUG_ANY,
+					"mdb_entry_decode: attributeType %s value #%d provided more than once\n",
+					a->a_desc->ad_cname.bv_val, j );
+				goto leave;
+			}
+		}
+		a->a_next = a+1;
+		a = a->a_next;
+	}
+	a[-1].a_next = NULL;
+done:
+	Debug(LDAP_DEBUG_TRACE, "<= mdb_entry_decode\n" );
+	*e = x;
+	rc = 0;
+
+leave:
+	if (mvc)
+		mdb_cursor_close(mvc);
+	return rc;
+}