[Cluster-devel] [PATCH] libdlm: fix dlm_controld skipping lock spaces

[Lars Marowsky-Bree <lmb@suse.de>]

With several lock spaces, there was a chance that poll_fencing,
poll_quorum, or poll_fs were reset after the earlier lock spaces had
been iterated over by process_lockspace_changes(). This would cause the
earlier lock spaces to not be recovered, since
process_lockspace_changes() would not be called again from the main
loop.

This fixes bugzilla.novell.com #528425.

Acked-by: teigland at redhat.com

Index: libdlm/group/dlm_controld/cpg.c
===================================================================
--- libdlm.orig/group/dlm_controld/cpg.c
+++ libdlm/group/dlm_controld/cpg.c
@@ -653,26 +653,23 @@ static int wait_conditions_done(struct l
 	   just the latest change */
 
 	if (!check_fencing_done(ls)) {
-		poll_fencing = 1;
+		poll_fencing++;
 		return 0;
 	}
-	poll_fencing = 0;
 
 	/* even though fencing also waits for quorum, checking fencing isn't
 	   sufficient because we don't want to start new lockspaces in an
 	   inquorate cluster */
 
 	if (!check_quorum_done(ls)) {
-		poll_quorum = 1;
+		poll_quorum++;
 		return 0;
 	}
-	poll_quorum = 0;
 
 	if (!check_fs_done(ls)) {
-		poll_fs = 1;
+		poll_fs++;
 		return 0;
 	}
-	poll_fs = 0;
 
 	return 1;
 }
@@ -1168,6 +1165,9 @@ void process_lockspace_changes(void)
 {
 	struct lockspace *ls, *safe;
 
+	poll_fencing = 0;
+	poll_quorum = 0;
+	poll_fs = 0;
 	list_for_each_entry_safe(ls, safe, &lockspaces, list) {
 		if (!list_empty(&ls->changes))
 			apply_changes(ls);

-- 
Architect Storage/HA, OPS Engineering, Novell, Inc.
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG N?rnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde