liquidio: xmit_more support

Defer ringing the Tx doorbell if skb->xmit_more is set unless the Tx queue
is full or stopped.  To keep latency low, use a deferral limit of 8
packets.  We chose 8 because Octeon can fetch at most 8 packets in a single
PCI read, and our tests show that 8 results in low latency.

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/cavium/liquidio/lio_core.c

@@ -91,7 +91,7 @@ void octeon_update_tx_completion_counters(void *buf, int reqtype,
 	*bytes_compl += skb->len;
 }
 
-void octeon_report_sent_bytes_to_bql(void *buf, int reqtype)
+int octeon_report_sent_bytes_to_bql(void *buf, int reqtype)
 {
 	struct octnet_buf_free_info *finfo;
 	struct sk_buff *skb;
@@ -112,11 +112,13 @@ void octeon_report_sent_bytes_to_bql(void *buf, int reqtype)
 		break;
 
 	default:
-		return;
+		return 0;
 	}
 
 	txq = netdev_get_tx_queue(skb->dev, skb_get_queue_mapping(skb));
 	netdev_tx_sent_queue(txq, skb->len);
+
+	return netif_xmit_stopped(txq);
 }
 
 void liquidio_link_ctrl_cmd_completion(void *nctrl_ptr)

drivers/net/ethernet/cavium/liquidio/lio_main.c

@@ -2594,7 +2594,8 @@ static void handle_timestamp(struct octeon_device *oct,
  */
 static inline int send_nic_timestamp_pkt(struct octeon_device *oct,
 					 struct octnic_data_pkt *ndata,
-					 struct octnet_buf_free_info *finfo)
+					 struct octnet_buf_free_info *finfo,
+					 int xmit_more)
 {
 	int retval;
 	struct octeon_soft_command *sc;
@@ -2629,7 +2630,7 @@ static inline int send_nic_timestamp_pkt(struct octeon_device *oct,
 		len = (u32)((struct octeon_instr_ih2 *)
 			    (&sc->cmd.cmd2.ih2))->dlengsz;
 
-	ring_doorbell = 1;
+	ring_doorbell = !xmit_more;
 
 	retval = octeon_send_command(oct, sc->iq_no, ring_doorbell, &sc->cmd,
 				     sc, len, ndata->reqtype);
@@ -2663,7 +2664,7 @@ static int liquidio_xmit(struct sk_buff *skb, struct net_device *netdev)
 	union tx_info *tx_info;
 	int status = 0;
 	int q_idx = 0, iq_no = 0;
-	int j;
+	int j, xmit_more = 0;
 	u64 dptr = 0;
 	u32 tag = 0;
 
@@ -2868,17 +2869,19 @@ static int liquidio_xmit(struct sk_buff *skb, struct net_device *netdev)
 		irh->vlan = skb_vlan_tag_get(skb) & 0xfff;
 	}
 
+	xmit_more = skb->xmit_more;
+
 	if (unlikely(cmdsetup.s.timestamp))
-		status = send_nic_timestamp_pkt(oct, &ndata, finfo);
+		status = send_nic_timestamp_pkt(oct, &ndata, finfo, xmit_more);
 	else
-		status = octnet_send_nic_data_pkt(oct, &ndata);
+		status = octnet_send_nic_data_pkt(oct, &ndata, xmit_more);
 	if (status == IQ_SEND_FAILED)
 		goto lio_xmit_failed;
 
 	netif_info(lio, tx_queued, lio->netdev, "Transmit queued successfully\n");
 
 	if (status == IQ_SEND_STOP)
-		stop_q(lio->netdev, q_idx);
+		stop_q(netdev, q_idx);
 
 	netif_trans_update(netdev);
 
@@ -2897,6 +2900,9 @@ lio_xmit_failed:
 	if (dptr)
 		dma_unmap_single(&oct->pci_dev->dev, dptr,
 				 ndata.datasize, DMA_TO_DEVICE);
+
+	octeon_ring_doorbell_locked(oct, iq_no);
+
 	tx_buffer_free(skb);
 	return NETDEV_TX_OK;
 }

drivers/net/ethernet/cavium/liquidio/lio_vf_main.c

@@ -1690,7 +1690,8 @@ static void handle_timestamp(struct octeon_device *oct, u32 status, void *buf)
  */
 static int send_nic_timestamp_pkt(struct octeon_device *oct,
 				  struct octnic_data_pkt *ndata,
-				  struct octnet_buf_free_info *finfo)
+				  struct octnet_buf_free_info *finfo,
+				  int xmit_more)
 {
 	struct octeon_soft_command *sc;
 	int ring_doorbell;
@@ -1720,7 +1721,7 @@ static int send_nic_timestamp_pkt(struct octeon_device *oct,
 
 	len = (u32)((struct octeon_instr_ih3 *)(&sc->cmd.cmd3.ih3))->dlengsz;
 
-	ring_doorbell = 1;
+	ring_doorbell = !xmit_more;
 
 	retval = octeon_send_command(oct, sc->iq_no, ring_doorbell, &sc->cmd,
 				     sc, len, ndata->reqtype);
@@ -1752,6 +1753,7 @@ static int liquidio_xmit(struct sk_buff *skb, struct net_device *netdev)
 	struct octeon_device *oct;
 	int q_idx = 0, iq_no = 0;
 	union tx_info *tx_info;
+	int xmit_more = 0;
 	struct lio *lio;
 	int status = 0;
 	u64 dptr = 0;
@@ -1940,10 +1942,12 @@ static int liquidio_xmit(struct sk_buff *skb, struct net_device *netdev)
 		irh->vlan = skb_vlan_tag_get(skb) & VLAN_VID_MASK;
 	}
 
+	xmit_more = skb->xmit_more;
+
 	if (unlikely(cmdsetup.s.timestamp))
-		status = send_nic_timestamp_pkt(oct, &ndata, finfo);
+		status = send_nic_timestamp_pkt(oct, &ndata, finfo, xmit_more);
 	else
-		status = octnet_send_nic_data_pkt(oct, &ndata);
+		status = octnet_send_nic_data_pkt(oct, &ndata, xmit_more);
 	if (status == IQ_SEND_FAILED)
 		goto lio_xmit_failed;
 
@@ -1952,7 +1956,7 @@ static int liquidio_xmit(struct sk_buff *skb, struct net_device *netdev)
 	if (status == IQ_SEND_STOP) {
 		dev_err(&oct->pci_dev->dev, "Rcvd IQ_SEND_STOP signal; stopping IQ-%d\n",
 			iq_no);
-		stop_q(lio->netdev, q_idx);
+		stop_q(netdev, q_idx);
 	}
 
 	netif_trans_update(netdev);
@@ -1972,6 +1976,9 @@ lio_xmit_failed:
 	if (dptr)
 		dma_unmap_single(&oct->pci_dev->dev, dptr,
 				 ndata.datasize, DMA_TO_DEVICE);
+
+	octeon_ring_doorbell_locked(oct, iq_no);
+
 	tx_buffer_free(skb);
 	return NETDEV_TX_OK;
 }

drivers/net/ethernet/cavium/liquidio/octeon_config.h

@@ -37,6 +37,8 @@
 #define   MAX_OCTEON_LINKS	       MAX_OCTEON_NICIF
 #define   MAX_OCTEON_MULTICAST_ADDR    32
 
+#define   MAX_OCTEON_FILL_COUNT        8
+
 /* CN6xxx IQ configuration macros */
 #define   CN6XXX_MAX_INPUT_QUEUES      32
 #define   CN6XXX_MAX_IQ_DESCRIPTORS    2048

drivers/net/ethernet/cavium/liquidio/octeon_iq.h

@@ -343,6 +343,9 @@ int octeon_delete_instr_queue(struct octeon_device *octeon_dev, u32 iq_no);
 
 int lio_wait_for_instr_fetch(struct octeon_device *oct);
 
+void
+octeon_ring_doorbell_locked(struct octeon_device *oct, u32 iq_no);
+
 int
 octeon_register_reqtype_free_fn(struct octeon_device *oct, int reqtype,
 				void (*fn)(void *));

drivers/net/ethernet/cavium/liquidio/octeon_main.h

@@ -63,7 +63,7 @@ struct octnet_buf_free_info {
 };
 
 /* BQL-related functions */
-void octeon_report_sent_bytes_to_bql(void *buf, int reqtype);
+int octeon_report_sent_bytes_to_bql(void *buf, int reqtype);
 void octeon_update_tx_completion_counters(void *buf, int reqtype,
 					  unsigned int *pkts_compl,
 					  unsigned int *bytes_compl);

drivers/net/ethernet/cavium/liquidio/octeon_nic.c

@@ -82,9 +82,10 @@ octeon_alloc_soft_command_resp(struct octeon_device    *oct,
 }
 
 int octnet_send_nic_data_pkt(struct octeon_device *oct,
-			     struct octnic_data_pkt *ndata)
+			     struct octnic_data_pkt *ndata,
+			     int xmit_more)
 {
-	int ring_doorbell = 1;
+	int ring_doorbell = !xmit_more;
 
 	return octeon_send_command(oct, ndata->q_no, ring_doorbell, &ndata->cmd,
 				   ndata->buf, ndata->datasize,

drivers/net/ethernet/cavium/liquidio/octeon_nic.h

@@ -279,7 +279,8 @@ octeon_alloc_soft_command_resp(struct octeon_device    *oct,
  * queue should be stopped, and IQ_SEND_OK if it sent okay.
  */
 int octnet_send_nic_data_pkt(struct octeon_device *oct,
-			     struct octnic_data_pkt *ndata);
+			     struct octnic_data_pkt *ndata,
+			     int xmit_more);
 
 /** Send a NIC control packet to the device
  * @param oct - octeon device pointer

drivers/net/ethernet/cavium/liquidio/request_manager.c

@@ -278,6 +278,18 @@ ring_doorbell(struct octeon_device *oct, struct octeon_instr_queue *iq)
 	}
 }
 
+void
+octeon_ring_doorbell_locked(struct octeon_device *oct, u32 iq_no)
+{
+	struct octeon_instr_queue *iq;
+
+	iq = oct->instr_queue[iq_no];
+	spin_lock(&iq->post_lock);
+	if (iq->fill_cnt)
+		ring_doorbell(oct, iq);
+	spin_unlock(&iq->post_lock);
+}
+
 static inline void __copy_cmd_into_iq(struct octeon_instr_queue *iq,
 				      u8 *cmd)
 {
@@ -543,6 +555,7 @@ octeon_send_command(struct octeon_device *oct, u32 iq_no,
 		    u32 force_db, void *cmd, void *buf,
 		    u32 datasize, u32 reqtype)
 {
+	int xmit_stopped;
 	struct iq_post_status st;
 	struct octeon_instr_queue *iq = oct->instr_queue[iq_no];
 
@@ -554,12 +567,13 @@ octeon_send_command(struct octeon_device *oct, u32 iq_no,
 	st = __post_command2(iq, cmd);
 
 	if (st.status != IQ_SEND_FAILED) {
-		octeon_report_sent_bytes_to_bql(buf, reqtype);
+		xmit_stopped = octeon_report_sent_bytes_to_bql(buf, reqtype);
 		__add_to_request_list(iq, st.index, buf, reqtype);
 		INCR_INSTRQUEUE_PKT_COUNT(oct, iq_no, bytes_sent, datasize);
 		INCR_INSTRQUEUE_PKT_COUNT(oct, iq_no, instr_posted, 1);
 
-		if (force_db)
+		if (iq->fill_cnt >= MAX_OCTEON_FILL_COUNT || force_db ||
+		    xmit_stopped || st.status == IQ_SEND_STOP)
 			ring_doorbell(oct, iq);
 	} else {
 		INCR_INSTRQUEUE_PKT_COUNT(oct, iq_no, instr_dropped, 1);