be2net: create/destroy rx-queues on interface open/close

On some skews, the BE card sends pause frames (and not drop pkts) if there are
no more posted buffers available for packet reception.  This behaviour has a
side effect: When an interface is disabled, buffers are no longer posted on the
corresponding RX rings. All broadcast and multicast traffic being received on
the port will quickly fill up the PMEM and cause pause push back. As the PMEM
is shared by both the ports, all traffic being received on the other (enabled)
port also gets stalled.
The fix is to destroy RX rings when the interface is disabled. If there is no
RX ring match in the RXF lookup, the packets are discarded and so don't hog the
PMEM.
The RXQ creation cmd must now use MCC instead of MBOX as they are are called
post MCC queue creation.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 68 insertions, 37 deletions

diff --git a/drivers/net/benet/be_main.c b/drivers/net/benet/be_main.c
index ae281de94b99..5b319d78f355 100644
--- a/drivers/net/benet/be_main.c
+++ b/drivers/net/benet/be_main.c
@@ -1572,6 +1572,7 @@ static void be_rx_q_clean(struct be_adapter *adapter, struct be_rx_obj *rxo)
 		memset(page_info, 0, sizeof(*page_info));
 	}
 	BUG_ON(atomic_read(&rxq->used));
+	rxq->tail = rxq->head = 0;
 }
 
 static void be_tx_compl_clean(struct be_adapter *adapter,
@@ -1752,29 +1753,16 @@ static void be_rx_queues_destroy(struct be_adapter *adapter)
 	int i;
 
 	for_all_rx_queues(adapter, rxo, i) {
-		q = &rxo->q;
-		if (q->created) {
-			be_cmd_q_destroy(adapter, q, QTYPE_RXQ);
-			/* After the rxq is invalidated, wait for a grace time
-			 * of 1ms for all dma to end and the flush compl to
-			 * arrive
-			 */
-			mdelay(1);
-			be_rx_q_clean(adapter, rxo);
-		}
-		be_queue_free(adapter, q);
+		be_queue_free(adapter, &rxo->q);
 
 		q = &rxo->cq;
 		if (q->created)
 			be_cmd_q_destroy(adapter, q, QTYPE_CQ);
 		be_queue_free(adapter, q);
 
-		/* Clear any residual events */
 		q = &rxo->rx_eq.q;
-		if (q->created) {
-			be_eq_clean(adapter, &rxo->rx_eq);
+		if (q->created)
 			be_cmd_q_destroy(adapter, q, QTYPE_EQ);
-		}
 		be_queue_free(adapter, q);
 	}
 }
@@ -1833,30 +1821,14 @@ static int be_rx_queues_create(struct be_adapter *adapter)
 		rc = be_cmd_cq_create(adapter, cq, eq, false, false, 3);
 		if (rc)
 			goto err;
-		/* Rx Q */
+
+		/* Rx Q - will be created in be_open() */
 		q = &rxo->q;
 		rc = be_queue_alloc(adapter, q, RX_Q_LEN,
 				sizeof(struct be_eth_rx_d));
 		if (rc)
 			goto err;
 
-		rc = be_cmd_rxq_create(adapter, q, cq->id, rx_frag_size,
-			BE_MAX_JUMBO_FRAME_SIZE, adapter->if_handle,
-			(i > 0) ? 1 : 0/* rss enable */, &rxo->rss_id);
-		if (rc)
-			goto err;
-	}
-
-	if (be_multi_rxq(adapter)) {
-		u8 rsstable[MAX_RSS_QS];
-
-		for_all_rss_queues(adapter, rxo, i)
-			rsstable[i] = rxo->rss_id;
-
-		rc = be_cmd_rss_config(adapter, rsstable,
-			adapter->num_rx_qs - 1);
-		if (rc)
-			goto err;
 	}
 
 	return 0;
@@ -2302,6 +2274,31 @@ done:
 	adapter->isr_registered = false;
 }
 
+static void be_rx_queues_clear(struct be_adapter *adapter)
+{
+	struct be_queue_info *q;
+	struct be_rx_obj *rxo;
+	int i;
+
+	for_all_rx_queues(adapter, rxo, i) {
+		q = &rxo->q;
+		if (q->created) {
+			be_cmd_rxq_destroy(adapter, q);
+			/* After the rxq is invalidated, wait for a grace time
+			 * of 1ms for all dma to end and the flush compl to
+			 * arrive
+			 */
+			mdelay(1);
+			be_rx_q_clean(adapter, rxo);
+		}
+
+		/* Clear any residual events */
+		q = &rxo->rx_eq.q;
+		if (q->created)
+			be_eq_clean(adapter, &rxo->rx_eq);
+	}
+}
+
 static int be_close(struct net_device *netdev)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
@@ -2350,6 +2347,40 @@ static int be_close(struct net_device *netdev)
 	for_all_tx_queues(adapter, txo, i)
 		be_tx_compl_clean(adapter, txo);
 
+	be_rx_queues_clear(adapter);
+	return 0;
+}
+
+static int be_rx_queues_setup(struct be_adapter *adapter)
+{
+	struct be_rx_obj *rxo;
+	int rc, i;
+	u8 rsstable[MAX_RSS_QS];
+
+	for_all_rx_queues(adapter, rxo, i) {
+		rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
+			rx_frag_size, BE_MAX_JUMBO_FRAME_SIZE,
+			adapter->if_handle,
+			(i > 0) ? 1 : 0/* rss enable */, &rxo->rss_id);
+		if (rc)
+			return rc;
+	}
+
+	if (be_multi_rxq(adapter)) {
+		for_all_rss_queues(adapter, rxo, i)
+			rsstable[i] = rxo->rss_id;
+
+		rc = be_cmd_rss_config(adapter, rsstable,
+			adapter->num_rx_qs - 1);
+		if (rc)
+			return rc;
+	}
+
+	/* First time posting */
+	for_all_rx_queues(adapter, rxo, i) {
+		be_post_rx_frags(rxo, GFP_KERNEL);
+		napi_enable(&rxo->rx_eq.napi);
+	}
 	return 0;
 }
 
@@ -2363,10 +2394,10 @@ static int be_open(struct net_device *netdev)
 	u8 mac_speed;
 	u16 link_speed;
 
-	for_all_rx_queues(adapter, rxo, i) {
-		be_post_rx_frags(rxo, GFP_KERNEL);
-		napi_enable(&rxo->rx_eq.napi);
-	}
+	status = be_rx_queues_setup(adapter);
+	if (status)
+		goto err;
+
 	napi_enable(&tx_eq->napi);
 
 	be_irq_register(adapter);