summaryrefslogtreecommitdiffstats
path: root/servers/slapd/back-mdb/config.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:35:32 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:35:32 +0000
commit5ea77a75dd2d2158401331879f3c8f47940a732c (patch)
treed89dc06e9f4850a900f161e25f84e922c4f86cc8 /servers/slapd/back-mdb/config.c
parentInitial commit. (diff)
downloadopenldap-upstream.tar.xz
openldap-upstream.zip
Adding upstream version 2.5.13+dfsg.upstream/2.5.13+dfsgupstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'servers/slapd/back-mdb/config.c')
-rw-r--r--servers/slapd/back-mdb/config.c828
1 files changed, 828 insertions, 0 deletions
diff --git a/servers/slapd/back-mdb/config.c b/servers/slapd/back-mdb/config.c
new file mode 100644
index 0000000..ce6d78a
--- /dev/null
+++ b/servers/slapd/back-mdb/config.c
@@ -0,0 +1,828 @@
+/* config.c - mdb backend configuration file routine */
+/* $OpenLDAP$ */
+/* 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/ctype.h>
+#include <ac/string.h>
+#include <ac/errno.h>
+
+#include "back-mdb.h"
+#include "idl.h"
+
+#include "slap-config.h"
+
+#include "lutil.h"
+#include "ldap_rq.h"
+
+
+static ConfigDriver mdb_cf_gen;
+static ConfigDriver mdb_bk_cfg;
+
+enum {
+ MDB_CHKPT = 1,
+ MDB_DIRECTORY,
+ MDB_DBNOSYNC,
+ MDB_ENVFLAGS,
+ MDB_INDEX,
+ MDB_MAXREADERS,
+ MDB_MAXSIZE,
+ MDB_MODE,
+ MDB_SSTACK,
+ MDB_MULTIVAL,
+ MDB_IDLEXP,
+};
+
+static ConfigTable mdbcfg[] = {
+ { "idlexp", "log", 2, 2, 0, ARG_UINT|ARG_MAGIC|MDB_IDLEXP,
+ mdb_bk_cfg, "( OLcfgBkAt:12.1 NAME 'olcBkMdbIdlExp' "
+ "DESC 'Power of 2 used to set IDL size' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL, NULL },
+ { "directory", "dir", 2, 2, 0, ARG_STRING|ARG_MAGIC|MDB_DIRECTORY,
+ mdb_cf_gen, "( OLcfgDbAt:0.1 NAME 'olcDbDirectory' "
+ "DESC 'Directory for database content' "
+ "EQUALITY caseExactMatch "
+ "SYNTAX OMsDirectoryString SINGLE-VALUE )", NULL, NULL },
+ { "checkpoint", "kbyte> <min", 3, 3, 0, ARG_MAGIC|MDB_CHKPT,
+ mdb_cf_gen, "( OLcfgDbAt:1.2 NAME 'olcDbCheckpoint' "
+ "DESC 'Database checkpoint interval in kbytes and minutes' "
+ "EQUALITY caseIgnoreMatch "
+ "SYNTAX OMsDirectoryString SINGLE-VALUE )",NULL, NULL },
+ { "dbnosync", NULL, 1, 2, 0, ARG_ON_OFF|ARG_MAGIC|MDB_DBNOSYNC,
+ mdb_cf_gen, "( OLcfgDbAt:1.4 NAME 'olcDbNoSync' "
+ "DESC 'Disable synchronous database writes' "
+ "EQUALITY booleanMatch "
+ "SYNTAX OMsBoolean SINGLE-VALUE )", NULL, NULL },
+ { "envflags", "flags", 2, 0, 0, ARG_MAGIC|MDB_ENVFLAGS,
+ mdb_cf_gen, "( OLcfgDbAt:12.3 NAME 'olcDbEnvFlags' "
+ "DESC 'Database environment flags' "
+ "EQUALITY caseIgnoreMatch "
+ "SYNTAX OMsDirectoryString )", NULL, NULL },
+ { "index", "attr> <[pres,eq,approx,sub]", 2, 3, 0, ARG_MAGIC|MDB_INDEX,
+ mdb_cf_gen, "( OLcfgDbAt:0.2 NAME 'olcDbIndex' "
+ "DESC 'Attribute index parameters' "
+ "EQUALITY caseIgnoreMatch "
+ "SYNTAX OMsDirectoryString )", NULL, NULL },
+ { "maxentrysize", "size", 2, 2, 0, ARG_ULONG|ARG_OFFSET,
+ (void *)offsetof(struct mdb_info, mi_maxentrysize),
+ "( OLcfgDbAt:12.4 NAME 'olcDbMaxEntrySize' "
+ "DESC 'Maximum size of an entry in bytes' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL, NULL },
+ { "maxreaders", "num", 2, 2, 0, ARG_UINT|ARG_MAGIC|MDB_MAXREADERS,
+ mdb_cf_gen, "( OLcfgDbAt:12.1 NAME 'olcDbMaxReaders' "
+ "DESC 'Maximum number of threads that may access the DB concurrently' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL, NULL },
+ { "maxsize", "size", 2, 2, 0, ARG_ULONG|ARG_MAGIC|MDB_MAXSIZE,
+ mdb_cf_gen, "( OLcfgDbAt:12.2 NAME 'olcDbMaxSize' "
+ "DESC 'Maximum size of DB in bytes' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL, NULL },
+ { "mode", "mode", 2, 2, 0, ARG_MAGIC|MDB_MODE,
+ mdb_cf_gen, "( OLcfgDbAt:0.3 NAME 'olcDbMode' "
+ "DESC 'Unix permissions of database files' "
+ "EQUALITY caseIgnoreMatch "
+ "SYNTAX OMsDirectoryString SINGLE-VALUE )", NULL, NULL },
+ { "multival", "attr> <hi,lo", 3, 3, 0, ARG_MAGIC|MDB_MULTIVAL,
+ mdb_cf_gen,
+ "( OLcfgDbAt:12.6 NAME 'olcDbMultival' "
+ "DESC 'Hi/Lo thresholds for splitting multivalued attr out of main blob' "
+ "EQUALITY caseIgnoreMatch "
+ "SYNTAX OMsDirectoryString )", NULL, NULL },
+ { "rtxnsize", "entries", 2, 2, 0, ARG_UINT|ARG_OFFSET,
+ (void *)offsetof(struct mdb_info, mi_rtxn_size),
+ "( OLcfgDbAt:12.5 NAME 'olcDbRtxnSize' "
+ "DESC 'Number of entries to process in one read transaction' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL,
+ { .v_uint = DEFAULT_RTXN_SIZE } },
+ { "searchstack", "depth", 2, 2, 0, ARG_INT|ARG_MAGIC|MDB_SSTACK,
+ mdb_cf_gen, "( OLcfgDbAt:1.9 NAME 'olcDbSearchStack' "
+ "DESC 'Depth of search stack in IDLs' "
+ "EQUALITY integerMatch "
+ "SYNTAX OMsInteger SINGLE-VALUE )", NULL, NULL },
+ { NULL, NULL, 0, 0, 0, ARG_IGNORED,
+ NULL, NULL, NULL, NULL }
+};
+
+static ConfigOCs mdbocs[] = {
+ {
+ "( OLcfgBkOc:12.1 "
+ "NAME 'olcMdbBkConfig' "
+ "DESC 'MDB backend configuration' "
+ "SUP olcBackendConfig "
+ "MAY olcBkMdbIdlExp )",
+ Cft_Backend, mdbcfg },
+ {
+ "( OLcfgDbOc:12.1 "
+ "NAME 'olcMdbConfig' "
+ "DESC 'MDB database configuration' "
+ "SUP olcDatabaseConfig "
+ "MUST olcDbDirectory "
+ "MAY ( olcDbCheckpoint $ olcDbEnvFlags $ "
+ "olcDbNoSync $ olcDbIndex $ olcDbMaxReaders $ olcDbMaxSize $ "
+ "olcDbMode $ olcDbSearchStack $ olcDbMaxEntrySize $ olcDbRtxnSize $ "
+ "olcDbMultival ) )",
+ Cft_Database, mdbcfg+1 },
+ { NULL, 0, NULL }
+};
+
+static slap_verbmasks mdb_envflags[] = {
+ { BER_BVC("nosync"), MDB_NOSYNC },
+ { BER_BVC("nometasync"), MDB_NOMETASYNC },
+ { BER_BVC("writemap"), MDB_WRITEMAP },
+ { BER_BVC("mapasync"), MDB_MAPASYNC },
+ { BER_BVC("nordahead"), MDB_NORDAHEAD },
+ { BER_BVNULL, 0 }
+};
+
+static int
+mdb_bk_cfg( ConfigArgs *c )
+{
+ int rc = 0;
+ if ( c->op == SLAP_CONFIG_EMIT ) {
+ if ( MDB_idl_logn != MDB_IDL_LOGN )
+ c->value_int = MDB_idl_logn;
+ else
+ rc = 1;
+ } else if ( c->op == LDAP_MOD_DELETE ) {
+ /* We expect to immediately be followed by an Add, but */
+ MDB_idl_logn = MDB_IDL_LOGN; /* return to default for safety */
+ mdb_idl_reset();
+ c->bi->bi_private = 0;
+ } else {
+ /* with 32 bit ints, db_size max is 2^30 and um_size max is 2^31 */
+ if ( c->value_int >= MDB_IDL_LOGN && ( c->value_int < sizeof(int) * CHAR_BIT - 1 )) {
+ MDB_idl_logn = c->value_int;
+ mdb_idl_reset();
+ c->bi->bi_private = (void *)8; /* non-NULL to show we're using it */
+ } else {
+ rc = 1;
+ }
+ }
+ return rc;
+}
+
+/* perform periodic syncs */
+static void *
+mdb_checkpoint( void *ctx, void *arg )
+{
+ struct re_s *rtask = arg;
+ struct mdb_info *mdb = rtask->arg;
+
+ mdb_env_sync( mdb->mi_dbenv, 1 );
+ ldap_pvt_thread_mutex_lock( &slapd_rq.rq_mutex );
+ ldap_pvt_runqueue_stoptask( &slapd_rq, rtask );
+ ldap_pvt_thread_mutex_unlock( &slapd_rq.rq_mutex );
+ return NULL;
+}
+
+/* reindex entries on the fly */
+static void *
+mdb_online_index( void *ctx, void *arg )
+{
+ struct re_s *rtask = arg;
+ BackendDB *be = rtask->arg;
+ struct mdb_info *mdb = be->be_private;
+
+ Connection conn = {0};
+ OperationBuffer opbuf;
+ Operation *op;
+
+ MDB_cursor *curs;
+ MDB_val key, data;
+ MDB_txn *txn;
+ ID id;
+ Entry *e;
+ int rc, getnext = 1;
+ int i;
+
+ connection_fake_init( &conn, &opbuf, ctx );
+ op = &opbuf.ob_op;
+
+ op->o_bd = be;
+
+ id = 1;
+ key.mv_size = sizeof(ID);
+
+ while ( 1 ) {
+ if ( slapd_shutdown )
+ break;
+
+ rc = mdb_txn_begin( mdb->mi_dbenv, NULL, 0, &txn );
+ if ( rc )
+ break;
+ rc = mdb_cursor_open( txn, mdb->mi_id2entry, &curs );
+ if ( rc ) {
+ mdb_txn_abort( txn );
+ break;
+ }
+ if ( getnext ) {
+ getnext = 0;
+ key.mv_data = &id;
+ rc = mdb_cursor_get( curs, &key, &data, MDB_SET_RANGE );
+ if ( rc ) {
+ mdb_txn_abort( txn );
+ if ( rc == MDB_NOTFOUND )
+ rc = 0;
+ break;
+ }
+ memcpy( &id, key.mv_data, sizeof( id ));
+ }
+
+ rc = mdb_id2entry( op, curs, id, &e );
+ mdb_cursor_close( curs );
+ if ( rc ) {
+ mdb_txn_abort( txn );
+ if ( rc == MDB_NOTFOUND ) {
+ id++;
+ getnext = 1;
+ continue;
+ }
+ break;
+ }
+ rc = mdb_index_entry( op, txn, MDB_INDEX_UPDATE_OP, e );
+ mdb_entry_return( op, e );
+ if ( rc == 0 ) {
+ rc = mdb_txn_commit( txn );
+ txn = NULL;
+ } else {
+ mdb_txn_abort( txn );
+ txn = NULL;
+ }
+ if ( rc ) {
+ Debug( LDAP_DEBUG_ANY,
+ LDAP_XSTRING(mdb_online_index) ": database %s: "
+ "txn_commit failed: %s (%d)\n",
+ be->be_suffix[0].bv_val, mdb_strerror(rc), rc );
+ break;
+ }
+ id++;
+ getnext = 1;
+ }
+
+ for ( i = 0; i < mdb->mi_nattrs; i++ ) {
+ if ( mdb->mi_attrs[ i ]->ai_indexmask & MDB_INDEX_DELETING
+ || mdb->mi_attrs[ i ]->ai_newmask == 0 )
+ {
+ continue;
+ }
+ mdb->mi_attrs[ i ]->ai_indexmask = mdb->mi_attrs[ i ]->ai_newmask;
+ mdb->mi_attrs[ i ]->ai_newmask = 0;
+ }
+
+ ldap_pvt_thread_mutex_lock( &slapd_rq.rq_mutex );
+ ldap_pvt_runqueue_stoptask( &slapd_rq, rtask );
+ mdb->mi_index_task = NULL;
+ ldap_pvt_runqueue_remove( &slapd_rq, rtask );
+ ldap_pvt_thread_mutex_unlock( &slapd_rq.rq_mutex );
+
+ return NULL;
+}
+
+/* Cleanup loose ends after Modify completes */
+static int
+mdb_cf_cleanup( ConfigArgs *c )
+{
+ struct mdb_info *mdb = c->be->be_private;
+ int rc = 0;
+
+ if ( mdb->mi_flags & MDB_DEL_INDEX ) {
+ mdb_attr_flush( mdb );
+ mdb->mi_flags ^= MDB_DEL_INDEX;
+ }
+
+ if ( mdb->mi_flags & MDB_RE_OPEN ) {
+ mdb->mi_flags ^= MDB_RE_OPEN;
+ rc = c->be->bd_info->bi_db_close( c->be, &c->reply );
+ if ( rc == 0 )
+ rc = c->be->bd_info->bi_db_open( c->be, &c->reply );
+ /* If this fails, we need to restart */
+ if ( rc ) {
+ slapd_shutdown = 2;
+ snprintf( c->cr_msg, sizeof( c->cr_msg ),
+ "failed to reopen database, rc=%d", rc );
+ Debug( LDAP_DEBUG_ANY, LDAP_XSTRING(mdb_cf_cleanup)
+ ": %s\n", c->cr_msg );
+ rc = LDAP_OTHER;
+ }
+ }
+
+ if ( mdb->mi_flags & MDB_OPEN_INDEX ) {
+ mdb->mi_flags ^= MDB_OPEN_INDEX;
+ rc = mdb_attr_dbs_open( c->be, NULL, &c->reply );
+ if ( rc )
+ rc = LDAP_OTHER;
+ }
+ return rc;
+}
+
+static int
+mdb_cf_gen( ConfigArgs *c )
+{
+ struct mdb_info *mdb = c->be->be_private;
+ int rc;
+
+ if ( c->op == SLAP_CONFIG_EMIT ) {
+ rc = 0;
+ switch( c->type ) {
+ case MDB_MODE: {
+ char buf[64];
+ struct berval bv;
+ bv.bv_len = snprintf( buf, sizeof(buf), "0%o", mdb->mi_dbenv_mode );
+ if ( bv.bv_len > 0 && bv.bv_len < sizeof(buf) ) {
+ bv.bv_val = buf;
+ value_add_one( &c->rvalue_vals, &bv );
+ } else {
+ rc = 1;
+ }
+ } break;
+
+ case MDB_CHKPT:
+ if ( mdb->mi_txn_cp ) {
+ char buf[64];
+ struct berval bv;
+ bv.bv_len = snprintf( buf, sizeof(buf), "%ld %ld",
+ (long) mdb->mi_txn_cp_kbyte, (long) mdb->mi_txn_cp_min );
+ if ( bv.bv_len > 0 && bv.bv_len < sizeof(buf) ) {
+ bv.bv_val = buf;
+ value_add_one( &c->rvalue_vals, &bv );
+ } else {
+ rc = 1;
+ }
+ } else {
+ rc = 1;
+ }
+ break;
+
+ case MDB_DIRECTORY:
+ if ( mdb->mi_dbenv_home ) {
+ c->value_string = ch_strdup( mdb->mi_dbenv_home );
+ } else {
+ rc = 1;
+ }
+ break;
+
+ case MDB_DBNOSYNC:
+ if ( mdb->mi_dbenv_flags & MDB_NOSYNC )
+ c->value_int = 1;
+ break;
+
+ case MDB_ENVFLAGS:
+ if ( mdb->mi_dbenv_flags ) {
+ mask_to_verbs( mdb_envflags, mdb->mi_dbenv_flags, &c->rvalue_vals );
+ }
+ if ( !c->rvalue_vals ) rc = 1;
+ break;
+
+ case MDB_INDEX:
+ mdb_attr_index_unparse( mdb, &c->rvalue_vals );
+ if ( !c->rvalue_vals ) rc = 1;
+ break;
+
+ case MDB_SSTACK:
+ c->value_int = mdb->mi_search_stack_depth;
+ break;
+
+ case MDB_MAXREADERS:
+ c->value_int = mdb->mi_readers;
+ break;
+
+ case MDB_MAXSIZE:
+ c->value_ulong = mdb->mi_mapsize;
+ break;
+
+ case MDB_MULTIVAL:
+ mdb_attr_multi_unparse( mdb, &c->rvalue_vals );
+ if ( !c->rvalue_vals ) rc = 1;
+ break;
+ }
+ return rc;
+ } else if ( c->op == LDAP_MOD_DELETE ) {
+ rc = 0;
+ switch( c->type ) {
+ case MDB_MODE:
+#if 0
+ /* FIXME: does it make any sense to change the mode,
+ * if we don't exec a chmod()? */
+ mdb->bi_dbenv_mode = SLAPD_DEFAULT_DB_MODE;
+ break;
+#endif
+
+ /* single-valued no-ops */
+ case MDB_SSTACK:
+ case MDB_MAXREADERS:
+ case MDB_MAXSIZE:
+ break;
+
+ case MDB_CHKPT:
+ if ( mdb->mi_txn_cp_task ) {
+ struct re_s *re = mdb->mi_txn_cp_task;
+ mdb->mi_txn_cp_task = NULL;
+ ldap_pvt_thread_mutex_lock( &slapd_rq.rq_mutex );
+ if ( ldap_pvt_runqueue_isrunning( &slapd_rq, re ) )
+ ldap_pvt_runqueue_stoptask( &slapd_rq, re );
+ ldap_pvt_runqueue_remove( &slapd_rq, re );
+ ldap_pvt_thread_mutex_unlock( &slapd_rq.rq_mutex );
+ }
+ mdb->mi_txn_cp = 0;
+ break;
+ case MDB_DIRECTORY:
+ mdb->mi_flags |= MDB_RE_OPEN;
+ ch_free( mdb->mi_dbenv_home );
+ mdb->mi_dbenv_home = NULL;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ ldap_pvt_thread_pool_purgekey( mdb->mi_dbenv );
+ break;
+ case MDB_DBNOSYNC:
+ mdb_env_set_flags( mdb->mi_dbenv, MDB_NOSYNC, 0 );
+ mdb->mi_dbenv_flags &= ~MDB_NOSYNC;
+ break;
+
+ case MDB_ENVFLAGS:
+ if ( c->valx == -1 ) {
+ int i;
+ for ( i=0; mdb_envflags[i].mask; i++) {
+ if ( mdb->mi_dbenv_flags & mdb_envflags[i].mask ) {
+ /* not all flags are runtime resettable */
+ rc = mdb_env_set_flags( mdb->mi_dbenv, mdb_envflags[i].mask, 0 );
+ if ( rc ) {
+ mdb->mi_flags |= MDB_RE_OPEN;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ rc = 0;
+ }
+ mdb->mi_dbenv_flags ^= mdb_envflags[i].mask;
+ }
+ }
+ } else {
+ int i = verb_to_mask( c->line, mdb_envflags );
+ if ( mdb_envflags[i].mask & mdb->mi_dbenv_flags ) {
+ rc = mdb_env_set_flags( mdb->mi_dbenv, mdb_envflags[i].mask, 0 );
+ if ( rc ) {
+ mdb->mi_flags |= MDB_RE_OPEN;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ rc = 0;
+ }
+ mdb->mi_dbenv_flags ^= mdb_envflags[i].mask;
+ } else {
+ /* unknown keyword */
+ snprintf( c->cr_msg, sizeof( c->cr_msg ), "%s: unknown keyword \"%s\"",
+ c->argv[0], c->argv[i] );
+ Debug( LDAP_DEBUG_CONFIG, "%s %s\n", c->log, c->cr_msg );
+ rc = 1;
+ }
+ }
+ break;
+
+ case MDB_INDEX:
+ if ( c->valx == -1 ) {
+ int i;
+
+ /* delete all */
+ for ( i = 0; i < mdb->mi_nattrs; i++ ) {
+ mdb->mi_attrs[i]->ai_indexmask |= MDB_INDEX_DELETING;
+ }
+ mdb->mi_defaultmask = 0;
+ mdb->mi_flags |= MDB_DEL_INDEX;
+ config_push_cleanup( c, mdb_cf_cleanup );
+
+ } else {
+ struct berval bv, def = BER_BVC("default");
+ char *ptr;
+
+ for (ptr = c->line; !isspace( (unsigned char) *ptr ); ptr++);
+
+ bv.bv_val = c->line;
+ bv.bv_len = ptr - bv.bv_val;
+ if ( bvmatch( &bv, &def )) {
+ mdb->mi_defaultmask = 0;
+
+ } else {
+ int i;
+ char **attrs;
+ char sep;
+
+ sep = bv.bv_val[ bv.bv_len ];
+ bv.bv_val[ bv.bv_len ] = '\0';
+ attrs = ldap_str2charray( bv.bv_val, "," );
+
+ for ( i = 0; attrs[ i ]; i++ ) {
+ AttributeDescription *ad = NULL;
+ const char *text;
+ AttrInfo *ai;
+
+ slap_str2ad( attrs[ i ], &ad, &text );
+ /* if we got here... */
+ assert( ad != NULL );
+
+ ai = mdb_attr_mask( mdb, ad );
+ /* if we got here... */
+ assert( ai != NULL );
+
+ ai->ai_indexmask |= MDB_INDEX_DELETING;
+ mdb->mi_flags |= MDB_DEL_INDEX;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ }
+
+ bv.bv_val[ bv.bv_len ] = sep;
+ ldap_charray_free( attrs );
+ }
+ }
+ break;
+ case MDB_MULTIVAL:
+ if ( c->valx == -1 ) {
+ int i;
+
+ /* delete all */
+ for ( i = 0; i < mdb->mi_nattrs; i++ ) {
+ mdb->mi_attrs[i]->ai_multi_hi = UINT_MAX;
+ mdb->mi_attrs[i]->ai_multi_lo = UINT_MAX;
+ }
+ mdb->mi_multi_hi = UINT_MAX;
+ mdb->mi_multi_lo = UINT_MAX;
+
+ } else {
+ struct berval bv, def = BER_BVC("default");
+ char *ptr;
+
+ for (ptr = c->line; !isspace( (unsigned char) *ptr ); ptr++);
+
+ bv.bv_val = c->line;
+ bv.bv_len = ptr - bv.bv_val;
+ if ( bvmatch( &bv, &def )) {
+ mdb->mi_multi_hi = UINT_MAX;
+ mdb->mi_multi_lo = UINT_MAX;
+
+ } else {
+ int i;
+ char **attrs;
+ char sep;
+
+ sep = bv.bv_val[ bv.bv_len ];
+ bv.bv_val[ bv.bv_len ] = '\0';
+ attrs = ldap_str2charray( bv.bv_val, "," );
+
+ for ( i = 0; attrs[ i ]; i++ ) {
+ AttributeDescription *ad = NULL;
+ const char *text;
+ AttrInfo *ai;
+
+ slap_str2ad( attrs[ i ], &ad, &text );
+ /* if we got here... */
+ assert( ad != NULL );
+
+ ai = mdb_attr_mask( mdb, ad );
+ /* if we got here... */
+ assert( ai != NULL );
+
+ ai->ai_multi_hi = UINT_MAX;
+ ai->ai_multi_lo = UINT_MAX;
+ }
+
+ bv.bv_val[ bv.bv_len ] = sep;
+ ldap_charray_free( attrs );
+ }
+ }
+ break;
+ }
+ return rc;
+ }
+
+ switch( c->type ) {
+ case MDB_MODE:
+ if ( ASCII_DIGIT( c->argv[1][0] ) ) {
+ long mode;
+ char *next;
+ errno = 0;
+ mode = strtol( c->argv[1], &next, 0 );
+ if ( errno != 0 || next == c->argv[1] || next[0] != '\0' ) {
+ fprintf( stderr, "%s: "
+ "unable to parse mode=\"%s\".\n",
+ c->log, c->argv[1] );
+ return 1;
+ }
+ mdb->mi_dbenv_mode = mode;
+
+ } else {
+ char *m = c->argv[1];
+ int who, what, mode = 0;
+
+ if ( strlen( m ) != STRLENOF("-rwxrwxrwx") ) {
+ return 1;
+ }
+
+ if ( m[0] != '-' ) {
+ return 1;
+ }
+
+ m++;
+ for ( who = 0; who < 3; who++ ) {
+ for ( what = 0; what < 3; what++, m++ ) {
+ if ( m[0] == '-' ) {
+ continue;
+ } else if ( m[0] != "rwx"[what] ) {
+ return 1;
+ }
+ mode += ((1 << (2 - what)) << 3*(2 - who));
+ }
+ }
+ mdb->mi_dbenv_mode = mode;
+ }
+ break;
+ case MDB_CHKPT: {
+ unsigned cp_kbyte, cp_min;
+ if ( lutil_atoux( &cp_kbyte, c->argv[1], 0 ) != 0 ) {
+ fprintf( stderr, "%s: "
+ "invalid kbyte \"%s\" in \"checkpoint\".\n",
+ c->log, c->argv[1] );
+ return 1;
+ }
+ if ( lutil_atoux( &cp_min, c->argv[2], 0 ) != 0 ) {
+ fprintf( stderr, "%s: "
+ "invalid minutes \"%s\" in \"checkpoint\".\n",
+ c->log, c->argv[2] );
+ return 1;
+ }
+ mdb->mi_txn_cp = 1;
+ mdb->mi_txn_cp_kbyte = cp_kbyte;
+ mdb->mi_txn_cp_min = cp_min;
+ /* If we're in server mode and time-based checkpointing is enabled,
+ * submit a task to perform periodic checkpoints.
+ */
+ if ((slapMode & SLAP_SERVER_MODE) && mdb->mi_txn_cp_min ) {
+ struct re_s *re = mdb->mi_txn_cp_task;
+ if ( re ) {
+ re->interval.tv_sec = mdb->mi_txn_cp_min * 60;
+ } else {
+ if ( c->be->be_suffix == NULL || BER_BVISNULL( &c->be->be_suffix[0] ) ) {
+ fprintf( stderr, "%s: "
+ "\"checkpoint\" must occur after \"suffix\".\n",
+ c->log );
+ return 1;
+ }
+ ldap_pvt_thread_mutex_lock( &slapd_rq.rq_mutex );
+ mdb->mi_txn_cp_task = ldap_pvt_runqueue_insert( &slapd_rq,
+ mdb->mi_txn_cp_min * 60, mdb_checkpoint, mdb,
+ LDAP_XSTRING(mdb_checkpoint), c->be->be_suffix[0].bv_val );
+ ldap_pvt_thread_mutex_unlock( &slapd_rq.rq_mutex );
+ }
+ }
+ } break;
+
+ case MDB_DIRECTORY: {
+ FILE *f;
+ char *ptr, *testpath;
+ int len;
+
+ len = strlen( c->value_string );
+ testpath = ch_malloc( len + STRLENOF(LDAP_DIRSEP) + STRLENOF("DUMMY") + 1 );
+ ptr = lutil_strcopy( testpath, c->value_string );
+ *ptr++ = LDAP_DIRSEP[0];
+ strcpy( ptr, "DUMMY" );
+ f = fopen( testpath, "w" );
+ if ( f ) {
+ fclose( f );
+ unlink( testpath );
+ }
+ ch_free( testpath );
+ if ( !f ) {
+ char ebuf[128];
+ int saved_errno = errno;
+ snprintf( c->cr_msg, sizeof( c->cr_msg ), "%s: invalid path: %s",
+ c->log, AC_STRERROR_R( saved_errno, ebuf, sizeof(ebuf) ) );
+ Debug( LDAP_DEBUG_ANY, "%s\n", c->cr_msg );
+ return -1;
+ }
+
+ if ( mdb->mi_dbenv_home )
+ ch_free( mdb->mi_dbenv_home );
+ mdb->mi_dbenv_home = c->value_string;
+
+ }
+ break;
+
+ case MDB_DBNOSYNC:
+ if ( c->value_int )
+ mdb->mi_dbenv_flags |= MDB_NOSYNC;
+ else
+ mdb->mi_dbenv_flags &= ~MDB_NOSYNC;
+ if ( mdb->mi_flags & MDB_IS_OPEN ) {
+ mdb_env_set_flags( mdb->mi_dbenv, MDB_NOSYNC,
+ c->value_int );
+ }
+ break;
+
+ case MDB_ENVFLAGS: {
+ int i, j;
+ for ( i=1; i<c->argc; i++ ) {
+ j = verb_to_mask( c->argv[i], mdb_envflags );
+ if ( mdb_envflags[j].mask ) {
+ if ( mdb->mi_flags & MDB_IS_OPEN )
+ rc = mdb_env_set_flags( mdb->mi_dbenv, mdb_envflags[j].mask, 1 );
+ else
+ rc = 0;
+ if ( rc ) {
+ mdb->mi_flags |= MDB_RE_OPEN;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ rc = 0;
+ }
+ mdb->mi_dbenv_flags |= mdb_envflags[j].mask;
+ } else {
+ /* unknown keyword */
+ snprintf( c->cr_msg, sizeof( c->cr_msg ), "%s: unknown keyword \"%s\"",
+ c->argv[0], c->argv[i] );
+ Debug( LDAP_DEBUG_ANY, "%s %s\n", c->log, c->cr_msg );
+ return 1;
+ }
+ }
+ }
+ break;
+
+ case MDB_INDEX:
+ rc = mdb_attr_index_config( mdb, c->fname, c->lineno,
+ c->argc - 1, &c->argv[1], &c->reply);
+
+ if( rc != LDAP_SUCCESS ) return 1;
+ if ( mdb->mi_flags & MDB_IS_OPEN ) {
+ mdb->mi_flags |= MDB_OPEN_INDEX;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ if ( !mdb->mi_index_task ) {
+ /* Start the task as soon as we finish here. Set a long
+ * interval (10 hours) so that it only gets scheduled once.
+ */
+ if ( c->be->be_suffix == NULL || BER_BVISNULL( &c->be->be_suffix[0] ) ) {
+ fprintf( stderr, "%s: "
+ "\"index\" must occur after \"suffix\".\n",
+ c->log );
+ return 1;
+ }
+ ldap_pvt_thread_mutex_lock( &slapd_rq.rq_mutex );
+ mdb->mi_index_task = ldap_pvt_runqueue_insert( &slapd_rq, 36000,
+ mdb_online_index, c->be,
+ LDAP_XSTRING(mdb_online_index), c->be->be_suffix[0].bv_val );
+ ldap_pvt_thread_mutex_unlock( &slapd_rq.rq_mutex );
+ }
+ }
+ break;
+
+ case MDB_SSTACK:
+ if ( c->value_int < MINIMUM_SEARCH_STACK_DEPTH ) {
+ fprintf( stderr,
+ "%s: depth %d too small, using %d\n",
+ c->log, c->value_int, MINIMUM_SEARCH_STACK_DEPTH );
+ c->value_int = MINIMUM_SEARCH_STACK_DEPTH;
+ }
+ mdb->mi_search_stack_depth = c->value_int;
+ break;
+
+ case MDB_MAXREADERS:
+ mdb->mi_readers = c->value_int;
+ if ( mdb->mi_flags & MDB_IS_OPEN ) {
+ mdb->mi_flags |= MDB_RE_OPEN;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ }
+ break;
+
+ case MDB_MAXSIZE:
+ mdb->mi_mapsize = c->value_ulong;
+ if ( mdb->mi_flags & MDB_IS_OPEN ) {
+ mdb->mi_flags |= MDB_RE_OPEN;
+ config_push_cleanup( c, mdb_cf_cleanup );
+ }
+ break;
+
+ case MDB_MULTIVAL:
+ rc = mdb_attr_multi_config( mdb, c->fname, c->lineno,
+ c->argc - 1, &c->argv[1], &c->reply);
+
+ if( rc != LDAP_SUCCESS ) return 1;
+ break;
+ }
+ return 0;
+}
+
+int mdb_back_init_cf( BackendInfo *bi )
+{
+ int rc;
+ bi->bi_cf_ocs = mdbocs;
+
+ rc = config_register_schema( mdbcfg, mdbocs );
+ if ( rc ) return rc;
+ return 0;
+}