[Ocfs2-commits] smushran commits r2845 - branches/ocfs2-1.2/fs/ocfs2/dlm

[svn-commits@oss.oracle.com]

Author: smushran
Signed-off-by: khackel
Signed-off-by: mfasheh
Date: 2006-04-19 15:38:02 -0500 (Wed, 19 Apr 2006)
New Revision: 2845

Modified:
   branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c
Log:
DLM: Catches bad recovery state
Ensures the cluster is recovering only one dead node at a time
Signed-off-by: khackel
Signed-off-by: mfasheh

Modified: branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c
===================================================================
--- branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c	2006-04-19 20:30:46 UTC (rev 2844)
+++ branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c	2006-04-19 20:38:02 UTC (rev 2845)
@@ -709,6 +709,14 @@
 	if (!dlm_grab(dlm))
 		return -EINVAL;
 
+	if (lr->dead_node != dlm->reco.dead_node) {
+		mlog(ML_ERROR, "%s: node %u sent dead_node=%u, but local "
+		     "dead_node is %u\n", dlm->name, lr->node_idx,
+		     lr->dead_node, dlm->reco.dead_node);
+		/* this is a hack */
+		dlm_put(dlm);
+		return -ENOMEM;
+	}
 	BUG_ON(lr->dead_node != dlm->reco.dead_node);
 
 	item = kcalloc(1, sizeof(*item), GFP_KERNEL);
@@ -1502,7 +1510,7 @@
 	struct dlm_lock *newlock = NULL;
 	struct dlm_lockstatus *lksb = NULL;
 	int ret = 0;
-	int i;
+	int i, bad;
 	struct list_head *iter;
 	struct dlm_lock *lock = NULL;
 
@@ -1612,9 +1620,33 @@
 		 * relative to each other, but clearly *not*
 		 * preserved relative to locks from other nodes.
 		 */
+		bad = 0;
 		spin_lock(&res->spinlock);
-		dlm_lock_get(newlock);
-		list_add_tail(&newlock->list, queue);
+		list_for_each_entry(lock, queue, list) {
+			if (lock->ml.cookie == ml->cookie) {
+				u64 c = lock->ml.cookie;
+				mlog(ML_ERROR, "%s:%.*s: %u:%llu: lock already "
+				     "exists on this lockres!\n", dlm->name,
+				     res->lockname.len, res->lockname.name,
+				     dlm_get_lock_cookie_node(c),
+				     dlm_get_lock_cookie_seq(c));
+
+				mlog(ML_NOTICE, "sent lock: type=%d, conv=%d, "
+				     "node=%u, cookie=%u:%llu, queue=%d\n",
+	      			     ml->type, ml->convert_type, ml->node,
+				     dlm_get_lock_cookie_node(ml->cookie),
+				     dlm_get_lock_cookie_seq(ml->cookie),
+				     ml->list);
+
+				__dlm_print_one_lock_resource(res);
+				bad = 1;
+				break;
+			}
+		}
+		if (!bad) {
+			dlm_lock_get(newlock);
+			list_add_tail(&newlock->list, queue);
+		}
 		spin_unlock(&res->spinlock);
 	}
 	mlog(0, "done running all the locks\n");