protocol/rpc/transport: bring in one more event for 'TRANSPORT-DESTROY'

needed because, a RPC disconnect doesn't mean that a RPC transport/listener
is dead. With this, the race in server protocol cleaning up the lock table /
fd table when some frames are in transit will be handled properly.

Signed-off-by: Amar Tumballi <amar@gluster.com>
Signed-off-by: Vijay Bellur <vijay@dev.gluster.com>

BUG: 1843 ()
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=1843

1 files changed, 5 insertions, 1 deletions

diff --git a/xlators/protocol/server/src/server-helpers.c b/xlators/protocol/server/src/server-helpers.c
index 5ea6f5619..69fb5f6a7 100644
--- a/xlators/protocol/server/src/server-helpers.c
+++ b/xlators/protocol/server/src/server-helpers.c
@@ -102,6 +102,10 @@ free_state (server_state_t *state)
                 state->conn = NULL;
         }
 
+        if (state->xprt) {
+                rpc_transport_unref (state->xprt);
+                state->xprt = NULL;
+        }
         if (state->fd) {
                 fd_unref (state->fd);
                 state->fd = NULL;
@@ -776,7 +780,7 @@ server_alloc_frame (rpcsvc_request_t *req)
         if (conn->bound_xl)
                 state->itable = conn->bound_xl->itable;
 
-        state->xprt  = req->trans;
+        state->xprt  = rpc_transport_ref (req->trans);
         state->conn  = conn;
 
         state->resolve.fd_no = -1;