[openib-general] [PATCH 2/2] 2.6.19 ib_cm: send DREP in response to unmatched DREQ

[Sean Hefty]

From: Sean Hefty <sean.hefty at intel.com>

Currently a DREP is only sent in response to a DREQ if a connection
has been found matching the DREQ, and it is in the proper state.  Once
a DREP is sent, the local connection moves into timewait.  Duplicate
DREQs received while in this state result in re-sending the DREP.

However, it's likely that the local connection will enter and exit
timewait before the remote side times out a lost DREP and resends a DREQ.
To handle this, we send a DREP in response to a DREQ, even if a local
connection is not found.  This avoids maintaining disconnected
id's in timewait states for excessively long times, just to handle a
lost DREP.

Signed-off-by: Sean Hefty <sean.hefty at intel.com>
---
This addresses a problem experienced by MPI during scale-up testing.

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 470c482..25b1018 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -1902,6 +1902,32 @@ out:	spin_unlock_irqrestore(&cm_id_priv-
 }
 EXPORT_SYMBOL(ib_send_cm_drep);
 
+static int cm_issue_drep(struct cm_port *port,
+			 struct ib_mad_recv_wc *mad_recv_wc)
+{
+	struct ib_mad_send_buf *msg = NULL;
+	struct cm_dreq_msg *dreq_msg;
+	struct cm_drep_msg *drep_msg;
+	int ret;
+
+	ret = cm_alloc_response_msg(port, mad_recv_wc, &msg);
+	if (ret)
+		return ret;
+
+	dreq_msg = (struct cm_dreq_msg *) mad_recv_wc->recv_buf.mad;
+	drep_msg = (struct cm_drep_msg *) msg->mad;
+
+	cm_format_mad_hdr(&drep_msg->hdr, CM_DREP_ATTR_ID, dreq_msg->hdr.tid);
+	drep_msg->remote_comm_id = dreq_msg->local_comm_id;
+	drep_msg->local_comm_id = dreq_msg->remote_comm_id;
+
+	ret = ib_post_send_mad(msg, NULL);
+	if (ret)
+		cm_free_msg(msg);
+
+	return ret;
+}
+
 static int cm_dreq_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
@@ -1913,8 +1939,10 @@ static int cm_dreq_handler(struct cm_wor
 	dreq_msg = (struct cm_dreq_msg *)work->mad_recv_wc->recv_buf.mad;
 	cm_id_priv = cm_acquire_id(dreq_msg->remote_comm_id,
 				   dreq_msg->local_comm_id);
-	if (!cm_id_priv)
+	if (!cm_id_priv) {
+		cm_issue_drep(work->port, work->mad_recv_wc);
 		return -EINVAL;
+	}
 
 	work->cm_event.private_data = &dreq_msg->private_data;