RDS: Don't destroy the rdma id until after we're done using it
authorSantosh Shilimkar <ssantosh@kernel.org>
Sat, 22 Aug 2015 22:45:32 +0000 (15:45 -0700)
committerDavid S. Miller <davem@davemloft.net>
Tue, 25 Aug 2015 20:35:31 +0000 (13:35 -0700)
During connection resets, we are destroying the rdma id too soon. We can't
destroy it when it is still in use. So lets move rdma_destroy_id() after
we clear the rings.

Reviewed-by: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/rds/ib_cm.c

index 04243dd1c2eae813d515288fff5eb4c277576e27..d150bb4aa3cb913510f704959e4aac4004710909 100644 (file)
@@ -647,7 +647,6 @@ void rds_ib_conn_shutdown(struct rds_connection *conn)
                        ib_destroy_cq(ic->i_send_cq);
                if (ic->i_recv_cq)
                        ib_destroy_cq(ic->i_recv_cq);
-               rdma_destroy_id(ic->i_cm_id);
 
                /* then free the resources that ib callbacks use */
                if (ic->i_send_hdrs)
@@ -673,6 +672,8 @@ void rds_ib_conn_shutdown(struct rds_connection *conn)
                if (ic->i_recvs)
                        rds_ib_recv_clear_ring(ic);
 
+               rdma_destroy_id(ic->i_cm_id);
+
                /*
                 * Move connection back to the nodev list.
                 */