[Ocfs2-commits] smushran commits r2894 - branches/ocfs2-1.2/fs/ocfs2/dlm

[svn-commits@oss.oracle.com]

Author: smushran
Signed-off-by: mfasheh
Signed-off-by: khackel
Date: 2006-04-20 13:21:07 -0500 (Thu, 20 Apr 2006)
New Revision: 2894

Modified:
   branches/ocfs2-1.2/fs/ocfs2/dlm/dlmcommon.h
   branches/ocfs2-1.2/fs/ocfs2/dlm/dlmmaster.c
   branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c
Log:
Wait for any nodes to be recovered when starting mastery.
Signed-off-by: mfasheh
Signed-off-by: khackel

Modified: branches/ocfs2-1.2/fs/ocfs2/dlm/dlmcommon.h
===================================================================
--- branches/ocfs2-1.2/fs/ocfs2/dlm/dlmcommon.h	2006-04-20 18:10:28 UTC (rev 2893)
+++ branches/ocfs2-1.2/fs/ocfs2/dlm/dlmcommon.h	2006-04-20 18:21:07 UTC (rev 2894)
@@ -700,6 +700,7 @@
 void dlm_kick_recovery_thread(struct dlm_ctxt *dlm);
 int dlm_is_node_dead(struct dlm_ctxt *dlm, u8 node);
 int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout);
+int dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout);
 
 void dlm_get(struct dlm_ctxt *dlm);
 void dlm_put(struct dlm_ctxt *dlm);

Modified: branches/ocfs2-1.2/fs/ocfs2/dlm/dlmmaster.c
===================================================================
--- branches/ocfs2-1.2/fs/ocfs2/dlm/dlmmaster.c	2006-04-20 18:10:28 UTC (rev 2893)
+++ branches/ocfs2-1.2/fs/ocfs2/dlm/dlmmaster.c	2006-04-20 18:21:07 UTC (rev 2894)
@@ -897,6 +897,9 @@
 		} else
 			wait_on_recovery = 0;
 		spin_unlock(&dlm->spinlock);
+
+		if (wait_on_recovery)
+			dlm_wait_for_node_recovery(dlm, bit, 10000);
 	}
 
 	/* must wait for lock to be mastered elsewhere */

Modified: branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c
===================================================================
--- branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c	2006-04-20 18:10:28 UTC (rev 2893)
+++ branches/ocfs2-1.2/fs/ocfs2/dlm/dlmrecovery.c	2006-04-20 18:21:07 UTC (rev 2894)
@@ -343,6 +343,18 @@
 	return dead;
 }
 
+/* returns true if node is no longer in the domain
+ * could be dead or just not joined */
+int dlm_is_node_recovered(struct dlm_ctxt *dlm, u8 node)
+{
+	int recovered;
+	spin_lock(&dlm->spinlock);
+	recovered = !test_bit(node, dlm->recovery_map);
+	spin_unlock(&dlm->spinlock);
+	return recovered;
+}
+
+
 int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout)
 {
 	if (timeout) {
@@ -361,6 +373,24 @@
 	return 0;
 }
 
+int dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout)
+{
+	if (timeout) {
+		mlog(0, "%s: waiting %dms for notification of "
+		     "recovery of node %u\n", dlm->name, timeout, node);
+		wait_event_timeout(dlm->dlm_reco_thread_wq,
+			   dlm_is_node_recovered(dlm, node),
+			   msecs_to_jiffies(timeout));
+	} else {
+		mlog(0, "%s: waiting indefinitely for notification "
+		     "of recovery of node %u\n", dlm->name, node);
+		wait_event(dlm->dlm_reco_thread_wq,
+			   dlm_is_node_recovered(dlm, node));
+	}
+	/* for now, return 0 */
+	return 0;
+}
+
 /* callers of the top-level api calls (dlmlock/dlmunlock) should
  * block on the dlm->reco.event when recovery is in progress.
  * the dlm recovery thread will set this state when it begins