mds: properly return log replay error

The only context that should see this is
C_MDS_BootStart, and it is prepared for failures,
so there's no reason to always return 0.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Greg Farnum <greg@inktank.com>

diff --git a/src/mds/MDLog.cc b/src/mds/MDLog.cc
index 1ace72e0ac378a14e2dbffd69997040de2d043f9..bd89da714952866fbc79d5f3991a2e872680417b 100644 (file)
--- a/src/mds/MDLog.cc
+++ b/src/mds/MDLog.cc
@@ -605,7 +605,7 @@ void MDLog::_replay_thread()
   }
 
   dout(10) << "_replay_thread kicking waiters" << dendl;
-  finish_contexts(g_ceph_context, waitfor_replay, 0);  
+  finish_contexts(g_ceph_context, waitfor_replay, r);  
 
   dout(10) << "_replay_thread finish" << dendl;
   mds->mds_lock.Unlock();