[DRBD-cvs] svn commit by simon - r2428 - trunk/drbd - Fixes a crash if you try to connect two diskless nodes.

[drbd-cvs at lists.linbit.com]

Author: simon
Date: 2006-09-18 13:58:41 +0200 (Mon, 18 Sep 2006)
New Revision: 2428

Modified:
   trunk/drbd/drbd_int.h
   trunk/drbd/drbd_main.c
   trunk/drbd/drbd_receiver.c
   trunk/drbd/drbd_worker.c
Log:
Fixes a crash if you try to connect two diskless nodes.

* Made sure that the exiting worker calls w_disconnect as well.
* The very last action of the exiting worker is the 
  module_put(THIS_MODULE).
* Changed (nearly) all calls to drbd_thread_stop() in after_state_ch()
  to the asynchronous variant.



Modified: trunk/drbd/drbd_int.h
===================================================================
--- trunk/drbd/drbd_int.h	2006-09-15 13:14:25 UTC (rev 2427)
+++ trunk/drbd/drbd_int.h	2006-09-18 11:58:41 UTC (rev 2428)
@@ -711,7 +711,7 @@
 	CLUSTER_ST_CHANGE,      // Cluster wide state change going on...
 	CL_ST_CHG_SUCCESS,
 	CL_ST_CHG_FAIL,
-	CRASHED_PRIMARY         // This node was a crashed primary
+	CRASHED_PRIMARY		// This node was a crashed primary
 };
 
 struct drbd_bitmap; // opaque for Drbd_Conf

Modified: trunk/drbd/drbd_main.c
===================================================================
--- trunk/drbd/drbd_main.c	2006-09-15 13:14:25 UTC (rev 2427)
+++ trunk/drbd/drbd_main.c	2006-09-18 11:58:41 UTC (rev 2428)
@@ -987,7 +987,7 @@
 	}
 
 	if ( os.conn != StandAlone && ns.conn == StandAlone ) {
-		drbd_thread_stop(&mdev->receiver);
+		drbd_thread_stop_nowait(&mdev->receiver);
 	}
 
 	if ( os.conn != Unconnected && ns.conn == Unconnected ) {
@@ -1003,9 +1003,7 @@
 	/* it feels better to have the module_put last ... */
 	if ( (os.disk > Diskless || os.conn > StandAlone) &&
 	     ns.disk == Diskless && ns.conn == StandAlone ) {
-		drbd_thread_stop(&mdev->worker);
-		drbd_mdev_cleanup(mdev);
-		module_put(THIS_MODULE);
+		drbd_thread_stop_nowait(&mdev->worker);
 	}
 }
 
@@ -2011,7 +2009,7 @@
 	mdev->rs_total     =
 	mdev->rs_mark_left =
 	mdev->rs_mark_time = 0;
-	mdev->net_conf     = NULL;
+	D_ASSERT(mdev->net_conf == NULL);
 	drbd_set_my_capacity(mdev,0);
 	drbd_bm_resize(mdev,0);
 

Modified: trunk/drbd/drbd_receiver.c
===================================================================
--- trunk/drbd/drbd_receiver.c	2006-09-15 13:14:25 UTC (rev 2427)
+++ trunk/drbd/drbd_receiver.c	2006-09-18 11:58:41 UTC (rev 2428)
@@ -1431,7 +1431,7 @@
 	if(mdev->state.pdsk <= Inconsistent) {
 		// In case we have the only disk of the cluster, 
 		drbd_set_out_of_sync(mdev,e->sector,e->size);
-		e->flags >= CALL_AL_COMPLETE_IO;
+		e->flags |= CALL_AL_COMPLETE_IO;
 		drbd_al_begin_io(mdev, e->sector);
 	}
 

Modified: trunk/drbd/drbd_worker.c
===================================================================
--- trunk/drbd/drbd_worker.c	2006-09-15 13:14:25 UTC (rev 2427)
+++ trunk/drbd/drbd_worker.c	2006-09-18 11:58:41 UTC (rev 2428)
@@ -680,6 +680,11 @@
 
 	if(w) kfree(w);
 
+	if(mdev->state.conn == StandAlone && mdev->net_conf ) {
+		kfree(mdev->net_conf);
+		mdev->net_conf = NULL;
+	}
+
 	return 1;
 }
 
@@ -1082,8 +1087,10 @@
 		i++; /* dead debugging code */
 	}
 
+	drbd_thread_stop(&mdev->receiver);
+
 	spin_lock_irq(&mdev->data.work.q_lock);
-	ERR_IF(!list_empty(&mdev->data.work.q))
+	if(!list_empty(&mdev->data.work.q))
 		goto again;
 	sema_init(&mdev->data.work.s,0);
 	/* DANGEROUS race: if someone did queue his work within the spinlock,
@@ -1093,6 +1100,10 @@
 	 */
 	spin_unlock_irq(&mdev->data.work.q_lock);
 
+	D_ASSERT( mdev->state.disk == Diskless && mdev->state.conn == StandAlone );
+	drbd_mdev_cleanup(mdev);
+	module_put(THIS_MODULE);
+
 	INFO("worker terminated\n");
 
 	return 0;