Rev 439: fixed a race condition in the handling of the recovery lock in http://samba.org/~tridge/ctdb

[tridge at samba.org]

------------------------------------------------------------
revno: 439
revision-id: tridge at samba.org-20070603002914-oycc73iztdrrq6g9
parent: tridge at samba.org-20070602094506-384lkqipi0x10k6c
committer: Andrew Tridgell <tridge at samba.org>
branch nick: tridge
timestamp: Sun 2007-06-03 10:29:14 +1000
message:
  fixed a race condition in the handling of the recovery lock
modified:
  common/ctdb_call.c             ctdb_call.c-20061128065342-to93h6eejj5kon81-1
  common/ctdb_recover.c          ctdb_recover.c-20070503002147-admmfgt1oj6gexfo-1
  common/ctdb_recoverd.c         recoverd.c-20070503213540-bvxuyd9jm1f7ig90-1
=== modified file 'common/ctdb_call.c'
--- a/common/ctdb_call.c	2007-05-31 03:50:53 +0000
+++ b/common/ctdb_call.c	2007-06-03 00:29:14 +0000
@@ -394,8 +394,9 @@
 
 	/* its a protocol error if the sending node is not the current dmaster */
 	if (header.dmaster != hdr->srcnode) {
-		DEBUG(0,("vnn %u dmaster request non-master %u dmaster=%u key %08x\n",
-			 ctdb->vnn, hdr->srcnode, header.dmaster, ctdb_hash(&key)));
+		DEBUG(0,("vnn %u dmaster request non-master %u dmaster=%u key %08x dbid 0x%08x\n",
+			 ctdb->vnn, hdr->srcnode, header.dmaster, ctdb_hash(&key),
+			 ctdb_db->db_id));
 		ctdb_fatal(ctdb, "ctdb_req_dmaster from non-master");
 		return;
 	}

=== modified file 'common/ctdb_recover.c'
--- a/common/ctdb_recover.c	2007-06-02 03:16:11 +0000
+++ b/common/ctdb_recover.c	2007-06-03 00:29:14 +0000
@@ -487,8 +487,8 @@
 	/* we should not be able to get the lock on the nodes list, as it should be
 	   held by the recovery master */
 	if (ctdb_recovery_lock(ctdb, false)) {
-		DEBUG(0,("ERROR: node list not locked when recovering!\n"));
-		ctdb_fatal(ctdb, "node list not locked - make sure it is on shared storage");
+		DEBUG(0,("ERROR: recovery lock file %s not locked when recovering!\n",
+			 ctdb->recovery_lock_file));
 		return -1;
 	}	
 

=== modified file 'common/ctdb_recoverd.c'
--- a/common/ctdb_recoverd.c	2007-06-02 01:36:42 +0000
+++ b/common/ctdb_recoverd.c	2007-06-03 00:29:14 +0000
@@ -614,7 +614,8 @@
 	}
 
 	/* release the recmaster lock */
-	if (ctdb->recovery_lock_fd != -1) {
+	if (em->vnn != ctdb_get_vnn(ctdb) &&
+	    ctdb->recovery_lock_fd != -1) {
 		close(ctdb->recovery_lock_fd);
 		ctdb->recovery_lock_fd = -1;
 	}