[PATCH 10/11] drbd_transport_rdma: introduce timeout for rdma_connect

[zhengbing.huang]

From: Dongsheng Yang <dongsheng.yang at easystack.cn>

Signed-off-by: Dongsheng Yang <dongsheng.yang at easystack.cn>
---
 drbd/drbd_transport_rdma.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drbd/drbd_transport_rdma.c b/drbd/drbd_transport_rdma.c
index 811f1a20a..0cd639254 100644
--- a/drbd/drbd_transport_rdma.c
+++ b/drbd/drbd_transport_rdma.c
@@ -140,11 +140,13 @@ union dtr_immediate {
 
 enum dtr_state_bits {
 	DSB_CONNECT_REQ,
+	DSB_CONNECTING,
 	DSB_CONNECTED,
 	DSB_ERROR,
 };
 
 #define DSM_CONNECT_REQ   (1 << DSB_CONNECT_REQ)
+#define DSM_CONNECTING    (1 << DSB_CONNECTING)
 #define DSM_CONNECTED     (1 << DSB_CONNECTED)
 #define DSM_ERROR         (1 << DSB_ERROR)
 
@@ -1033,6 +1035,7 @@ static int dtr_cma_accept(struct dtr_listener *listener, struct rdma_cm_id *new_
 		return -EAGAIN;
 	}
 
+	set_bit(DSB_CONNECTING, &cm->state);
 	err = rdma_accept(new_cm_id, &dtr_conn_param);
 	if (err)
 		kref_put(&cm->kref, dtr_destroy_cm);
@@ -1163,6 +1166,7 @@ static void dtr_cma_connect_work_fn(struct work_struct *work)
 	}
 
 	kref_get(&cm->kref); /* Expecting RDMA_CM_EVENT_ESTABLISHED */
+	set_bit(DSB_CONNECTING, &cm->state);
 	err = rdma_connect(cm->id, &dtr_conn_param);
 	if (err) {
 		kref_put(&cm->kref, dtr_destroy_cm); /* no RDMA_CM_EVENT_ESTABLISHED */
@@ -1170,6 +1174,15 @@ static void dtr_cma_connect_work_fn(struct work_struct *work)
 		goto out;
 	}
 
+	err = wait_event_timeout(cm->state_wq,
+			   !test_bit(DSB_CONNECTING, &cm->state), 20*HZ);
+
+	if (err == 0 && test_and_clear_bit(DSB_CONNECTING, &cm->state)) {
+		kref_put(&cm->kref, dtr_destroy_cm);
+		tr_err(transport, "rdma_connect timeout\n");
+		goto out;
+	}
+
 	kref_put(&cm->kref, dtr_destroy_cm); /* for work */
 	return;
 out:
@@ -1293,6 +1306,9 @@ static int dtr_cma_event_handler(struct rdma_cm_id *cm_id, struct rdma_cm_event
 
 	case RDMA_CM_EVENT_ESTABLISHED:
 		// pr_info("%s: RDMA_CM_EVENT_ESTABLISHED\n", cm->name);
+		if (!test_and_clear_bit(DSB_CONNECTING, &cm->state))
+			return 0;
+		wake_up(&cm->state_wq);
 		/* cm->state = DSM_CONNECTED; is set later in the work item */
 		/* This is called for active and passive connections */
 
@@ -1313,6 +1329,8 @@ static int dtr_cma_event_handler(struct rdma_cm_id *cm_id, struct rdma_cm_event
 		// pr_info("%s: RDMA_CM_EVENT_REJECTED\n", cm->name);
 		// pr_info("event = %d, status = %d\n", event->event, event->status);
 		set_bit(DSB_ERROR, &cm->state);
+		if (!test_and_clear_bit(DSB_CONNECTING, &cm->state))
+			return 0;
 
 		dtr_cma_retry_connect(cm->path, cm);
 		break;
-- 
2.27.0