ice: count representor stats

Removing control plane VSI result in no information about slow-path
statistic. In current solution statistics need to be counted in driver.

Patch is based on similar implementation done by Simon Horman in nfp:
commit eadfa4c3 ("nfp: add stats and xmit helpers for representors")

Add const modifier to netdev parameter in ice_netdev_to_repr(). It isn't
(and shouldn't be) modified in the function.
Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

drivers/net/ethernet/intel/ice/ice_eswitch.c

@@ -192,13 +192,18 @@ netdev_tx_t
 ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 {
 	struct ice_repr *repr = ice_netdev_to_repr(netdev);
+	unsigned int len = skb->len;
+	int ret;
 
 	skb_dst_drop(skb);
 	dst_hold((struct dst_entry *)repr->dst);
 	skb_dst_set(skb, (struct dst_entry *)repr->dst);
 	skb->dev = repr->dst->u.port_info.lower_dev;
 
-	return dev_queue_xmit(skb);
+	ret = dev_queue_xmit(skb);
+	ice_repr_inc_tx_stats(repr, len, ret);
+
+	return ret;
 }
 
 /**

drivers/net/ethernet/intel/ice/ice_repr.c

@@ -41,6 +41,47 @@ ice_repr_get_phys_port_name(struct net_device *netdev, char *buf, size_t len)
 	return 0;
 }
 
+/**
+ * ice_repr_inc_tx_stats - increment Tx statistic by one packet
+ * @repr: repr to increment stats on
+ * @len: length of the packet
+ * @xmit_status: value returned by xmit function
+ */
+void ice_repr_inc_tx_stats(struct ice_repr *repr, unsigned int len,
+			   int xmit_status)
+{
+	struct ice_repr_pcpu_stats *stats;
+
+	if (unlikely(xmit_status != NET_XMIT_SUCCESS &&
+		     xmit_status != NET_XMIT_CN)) {
+		this_cpu_inc(repr->stats->tx_drops);
+		return;
+	}
+
+	stats = this_cpu_ptr(repr->stats);
+	u64_stats_update_begin(&stats->syncp);
+	stats->tx_packets++;
+	stats->tx_bytes += len;
+	u64_stats_update_end(&stats->syncp);
+}
+
+/**
+ * ice_repr_inc_rx_stats - increment Rx statistic by one packet
+ * @netdev: repr netdev to increment stats on
+ * @len: length of the packet
+ */
+void ice_repr_inc_rx_stats(struct net_device *netdev, unsigned int len)
+{
+	struct ice_repr *repr = ice_netdev_to_repr(netdev);
+	struct ice_repr_pcpu_stats *stats;
+
+	stats = this_cpu_ptr(repr->stats);
+	u64_stats_update_begin(&stats->syncp);
+	stats->rx_packets++;
+	stats->rx_bytes += len;
+	u64_stats_update_end(&stats->syncp);
+}
+
 /**
  * ice_repr_get_stats64 - get VF stats for VFPR use
  * @netdev: pointer to port representor netdev
@@ -76,7 +117,7 @@ ice_repr_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
  * ice_netdev_to_repr - Get port representor for given netdevice
  * @netdev: pointer to port representor netdev
  */
-struct ice_repr *ice_netdev_to_repr(struct net_device *netdev)
+struct ice_repr *ice_netdev_to_repr(const struct net_device *netdev)
 {
 	struct ice_netdev_priv *np = netdev_priv(netdev);
 
@@ -139,38 +180,35 @@ static int ice_repr_stop(struct net_device *netdev)
  * ice_repr_sp_stats64 - get slow path stats for port representor
  * @dev: network interface device structure
  * @stats: netlink stats structure
- *
- * RX/TX stats are being swapped here to be consistent with VF stats. In slow
- * path, port representor receives data when the corresponding VF is sending it
- * (and vice versa), TX and RX bytes/packets are effectively swapped on port
- * representor.
  */
 static int
 ice_repr_sp_stats64(const struct net_device *dev,
 		    struct rtnl_link_stats64 *stats)
 {
-	struct ice_netdev_priv *np = netdev_priv(dev);
-	int vf_id = np->repr->vf->vf_id;
-	struct ice_tx_ring *tx_ring;
-	struct ice_rx_ring *rx_ring;
-	u64 pkts, bytes;
-
-	tx_ring = np->vsi->tx_rings[vf_id];
-	ice_fetch_u64_stats_per_ring(&tx_ring->ring_stats->syncp,
-				     tx_ring->ring_stats->stats,
-				     &pkts, &bytes);
-	stats->rx_packets = pkts;
-	stats->rx_bytes = bytes;
-
-	rx_ring = np->vsi->rx_rings[vf_id];
-	ice_fetch_u64_stats_per_ring(&rx_ring->ring_stats->syncp,
-				     rx_ring->ring_stats->stats,
-				     &pkts, &bytes);
-	stats->tx_packets = pkts;
-	stats->tx_bytes = bytes;
-	stats->tx_dropped = rx_ring->ring_stats->rx_stats.alloc_page_failed +
-			    rx_ring->ring_stats->rx_stats.alloc_buf_failed;
-
+	struct ice_repr *repr = ice_netdev_to_repr(dev);
+	int i;
+
+	for_each_possible_cpu(i) {
+		u64 tbytes, tpkts, tdrops, rbytes, rpkts;
+		struct ice_repr_pcpu_stats *repr_stats;
+		unsigned int start;
+
+		repr_stats = per_cpu_ptr(repr->stats, i);
+		do {
+			start = u64_stats_fetch_begin(&repr_stats->syncp);
+			tbytes = repr_stats->tx_bytes;
+			tpkts = repr_stats->tx_packets;
+			tdrops = repr_stats->tx_drops;
+			rbytes = repr_stats->rx_bytes;
+			rpkts = repr_stats->rx_packets;
+		} while (u64_stats_fetch_retry(&repr_stats->syncp, start));
+
+		stats->tx_bytes += tbytes;
+		stats->tx_packets += tpkts;
+		stats->tx_dropped += tdrops;
+		stats->rx_bytes += rbytes;
+		stats->rx_packets += rpkts;
+	}
 	return 0;
 }
 
@@ -291,6 +329,7 @@ static void ice_repr_remove_node(struct devlink_port *devlink_port)
  */
 static void ice_repr_rem(struct ice_repr *repr)
 {
+	free_percpu(repr->stats);
 	free_netdev(repr->netdev);
 	kfree(repr);
 }
@@ -344,6 +383,12 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 		goto err_alloc;
 	}
 
+	repr->stats = netdev_alloc_pcpu_stats(struct ice_repr_pcpu_stats);
+	if (!repr->stats) {
+		err = -ENOMEM;
+		goto err_stats;
+	}
+
 	repr->src_vsi = src_vsi;
 	repr->id = src_vsi->vsi_num;
 	np = netdev_priv(repr->netdev);
@@ -353,6 +398,8 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 
 	return repr;
 
+err_stats:
+	free_netdev(repr->netdev);
 err_alloc:
 	kfree(repr);
 	return ERR_PTR(err);

drivers/net/ethernet/intel/ice/ice_repr.h

@@ -6,12 +6,22 @@
 
 #include <net/dst_metadata.h>
 
+struct ice_repr_pcpu_stats {
+	struct u64_stats_sync syncp;
+	u64 rx_packets;
+	u64 rx_bytes;
+	u64 tx_packets;
+	u64 tx_bytes;
+	u64 tx_drops;
+};
+
 struct ice_repr {
 	struct ice_vsi *src_vsi;
 	struct ice_vf *vf;
 	struct net_device *netdev;
 	struct metadata_dst *dst;
 	struct ice_esw_br_port *br_port;
+	struct ice_repr_pcpu_stats __percpu *stats;
 	u32 id;
 	u8 parent_mac[ETH_ALEN];
 };
@@ -22,8 +32,12 @@ void ice_repr_rem_vf(struct ice_repr *repr);
 void ice_repr_start_tx_queues(struct ice_repr *repr);
 void ice_repr_stop_tx_queues(struct ice_repr *repr);
 
-struct ice_repr *ice_netdev_to_repr(struct net_device *netdev);
+struct ice_repr *ice_netdev_to_repr(const struct net_device *netdev);
 bool ice_is_port_repr_netdev(const struct net_device *netdev);
 
 struct ice_repr *ice_repr_get_by_vsi(struct ice_vsi *vsi);
+
+void ice_repr_inc_tx_stats(struct ice_repr *repr, unsigned int len,
+			   int xmit_status);
+void ice_repr_inc_rx_stats(struct net_device *netdev, unsigned int len);
 #endif

drivers/net/ethernet/intel/ice/ice_txrx_lib.c

@@ -240,6 +240,8 @@ ice_process_skb_fields(struct ice_rx_ring *rx_ring,
 		struct net_device *netdev = ice_eswitch_get_target(rx_ring,
 								   rx_desc);
 
+		if (ice_is_port_repr_netdev(netdev))
+			ice_repr_inc_rx_stats(netdev, skb->len);
 		skb->protocol = eth_type_trans(skb, netdev);
 	} else {
 		skb->protocol = eth_type_trans(skb, rx_ring->netdev);