cluster/ec: Correctly cleanup delayed locks

When a delayed lock is pending, a graph switch doesn't correctly
terminate it. This means that the update of version and size xattrs
is lost, causing EIO errors.

This patch handles GF_EVENT_PARENT_DOWN event to correctly finish
pending udpdates before completing the graph switch.

Change-Id: I394f3b8d41df8d83cdd36636aeb62330f30a66d5
BUG: 1188145
Signed-off-by: Xavier Hernandez <xhernandez@datalab.es>
Reviewed-on: http://review.gluster.org/10787
Tested-by: NetBSD Build System
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>

1 files changed, 7 insertions, 6 deletions

diff --git a/xlators/cluster/ec/src/ec-data.c b/xlators/cluster/ec/src/ec-data.c
index fb47aea90a8..b747fc42348 100644
--- a/xlators/cluster/ec/src/ec-data.c
+++ b/xlators/cluster/ec/src/ec-data.c
@@ -165,17 +165,18 @@ ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
         parent = frame->local;
         if (parent != NULL)
         {
-            LOCK(&parent->lock);
-
-            parent->jobs++;
-            parent->refs++;
-
-            UNLOCK(&parent->lock);
+            ec_sleep(parent);
         }
 
         fop->parent = parent;
     }
 
+    LOCK(&ec->lock);
+
+    list_add_tail(&fop->pending_list, &ec->pending_fops);
+
+    UNLOCK(&ec->lock);
+
     return fop;
 }