net: add per-cpu storage and net->core_stats

Before adding yet another possibly contended atomic_long_t,
it is time to add per-cpu storage for existing ones:
 dev->tx_dropped, dev->rx_dropped, and dev->rx_nohandler

Because many devices do not have to increment such counters,
allocate the per-cpu storage on demand, so that dev_get_stats()
does not have to spend considerable time folding zero counters.

Note that some drivers have abused these counters which
were supposed to be only used by core networking stack.

v4: should use per_cpu_ptr() in dev_get_stats() (Jakub)
v3: added a READ_ONCE() in netdev_core_stats_alloc() (Paolo)
v2: add a missing include (reported by kernel test robot <lkp@intel.com>)
    Change in netdev_core_stats_alloc() (Jakub)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: jeffreyji <jeffreyji@google.com>
Reviewed-by: Brian Vazquez <brianvv@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20220311051420.2608812-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

9 files changed, 18 insertions, 18 deletions

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 55e0ba2a163d..15eddca7b4b6 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -5120,7 +5120,7 @@ static netdev_tx_t bond_xmit_broadcast(struct sk_buff *skb,
 	if (xmit_suc)
 		return NETDEV_TX_OK;
 
-	atomic_long_inc(&bond_dev->tx_dropped);
+	dev_core_stats_tx_dropped_inc(bond_dev);
 	return NET_XMIT_DROP;
 }
 
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 2de02950086f..92a1a43b3bee 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -370,7 +370,7 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	i = skb_get_queue_mapping(skb);
 	if (unlikely(i >= bp->tx_nr_rings)) {
 		dev_kfree_skb_any(skb);
-		atomic_long_inc(&dev->tx_dropped);
+		dev_core_stats_tx_dropped_inc(dev);
 		return NETDEV_TX_OK;
 	}
 
@@ -646,7 +646,7 @@ tx_kick_pending:
 	if (txr->kick_pending)
 		bnxt_txr_db_kick(bp, txr, txr->tx_prod);
 	txr->tx_buf_ring[txr->tx_prod].skb = NULL;
-	atomic_long_inc(&dev->tx_dropped);
+	dev_core_stats_tx_dropped_inc(dev);
 	return NETDEV_TX_OK;
 }
 
diff --git a/drivers/net/ethernet/hisilicon/hns/hns_ethtool.c b/drivers/net/ethernet/hisilicon/hns/hns_ethtool.c
index d7a27c244d48..54faf0f2d1d8 100644
--- a/drivers/net/ethernet/hisilicon/hns/hns_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns/hns_ethtool.c
@@ -887,8 +887,8 @@ static void hns_get_ethtool_stats(struct net_device *netdev,
 	p[21] = net_stats->rx_compressed;
 	p[22] = net_stats->tx_compressed;
 
-	p[23] = netdev->rx_dropped.counter;
-	p[24] = netdev->tx_dropped.counter;
+	p[23] = 0; /* was netdev->rx_dropped.counter */
+	p[24] = 0; /* was netdev->tx_dropped.counter */
 
 	p[25] = priv->tx_timeout_count;
 
diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
index bfbd7847f946..a313242a762e 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
@@ -207,7 +207,7 @@ rx_handler_result_t rmnet_rx_handler(struct sk_buff **pskb)
 	dev = skb->dev;
 	port = rmnet_get_port_rcu(dev);
 	if (unlikely(!port)) {
-		atomic_long_inc(&skb->dev->rx_nohandler);
+		dev_core_stats_rx_nohandler_inc(skb->dev);
 		kfree_skb(skb);
 		goto done;
 	}
diff --git a/drivers/net/ipvlan/ipvlan_core.c b/drivers/net/ipvlan/ipvlan_core.c
index c613900c3811..6ffb27419e64 100644
--- a/drivers/net/ipvlan/ipvlan_core.c
+++ b/drivers/net/ipvlan/ipvlan_core.c
@@ -555,7 +555,7 @@ static void ipvlan_multicast_enqueue(struct ipvl_port *port,
 		schedule_work(&port->wq);
 	} else {
 		spin_unlock(&port->backlog.lock);
-		atomic_long_inc(&skb->dev->rx_dropped);
+		dev_core_stats_rx_dropped_inc(skb->dev);
 		kfree_skb(skb);
 	}
 }
diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
index 33753a2fde29..4b77819e9328 100644
--- a/drivers/net/macvlan.c
+++ b/drivers/net/macvlan.c
@@ -371,7 +371,7 @@ static void macvlan_broadcast_enqueue(struct macvlan_port *port,
 free_nskb:
 	kfree_skb(nskb);
 err:
-	atomic_long_inc(&skb->dev->rx_dropped);
+	dev_core_stats_rx_dropped_inc(skb->dev);
 }
 
 static void macvlan_flush_sources(struct macvlan_port *port,
diff --git a/drivers/net/net_failover.c b/drivers/net/net_failover.c
index 86ec5aae4289..21a0435c02de 100644
--- a/drivers/net/net_failover.c
+++ b/drivers/net/net_failover.c
@@ -89,7 +89,7 @@ static int net_failover_close(struct net_device *dev)
 static netdev_tx_t net_failover_drop_xmit(struct sk_buff *skb,
 					  struct net_device *dev)
 {
-	atomic_long_inc(&dev->tx_dropped);
+	dev_core_stats_tx_dropped_inc(dev);
 	dev_kfree_skb_any(skb);
 	return NETDEV_TX_OK;
 }
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 2b9a22669a12..276a0e42ca8e 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -1135,7 +1135,7 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
 	return NETDEV_TX_OK;
 
 drop:
-	atomic_long_inc(&dev->tx_dropped);
+	dev_core_stats_tx_dropped_inc(dev);
 	skb_tx_error(skb);
 	kfree_skb_reason(skb, drop_reason);
 	rcu_read_unlock();
@@ -1291,7 +1291,7 @@ resample:
 		void *frame = tun_xdp_to_ptr(xdp);
 
 		if (__ptr_ring_produce(&tfile->tx_ring, frame)) {
-			atomic_long_inc(&dev->tx_dropped);
+			dev_core_stats_tx_dropped_inc(dev);
 			break;
 		}
 		nxmit++;
@@ -1626,7 +1626,7 @@ static int tun_xdp_act(struct tun_struct *tun, struct bpf_prog *xdp_prog,
 		trace_xdp_exception(tun->dev, xdp_prog, act);
 		fallthrough;
 	case XDP_DROP:
-		atomic_long_inc(&tun->dev->rx_dropped);
+		dev_core_stats_rx_dropped_inc(tun->dev);
 		break;
 	}
 
@@ -1797,7 +1797,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 		 */
 		skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp);
 		if (IS_ERR(skb)) {
-			atomic_long_inc(&tun->dev->rx_dropped);
+			dev_core_stats_rx_dropped_inc(tun->dev);
 			return PTR_ERR(skb);
 		}
 		if (!skb)
@@ -1826,7 +1826,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 
 		if (IS_ERR(skb)) {
 			if (PTR_ERR(skb) != -EAGAIN)
-				atomic_long_inc(&tun->dev->rx_dropped);
+				dev_core_stats_rx_dropped_inc(tun->dev);
 			if (frags)
 				mutex_unlock(&tfile->napi_mutex);
 			return PTR_ERR(skb);
@@ -1841,7 +1841,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 			err = -EFAULT;
 			drop_reason = SKB_DROP_REASON_SKB_UCOPY_FAULT;
 drop:
-			atomic_long_inc(&tun->dev->rx_dropped);
+			dev_core_stats_rx_dropped_inc(tun->dev);
 			kfree_skb_reason(skb, drop_reason);
 			if (frags) {
 				tfile->napi.skb = NULL;
@@ -1876,7 +1876,7 @@ drop:
 				pi.proto = htons(ETH_P_IPV6);
 				break;
 			default:
-				atomic_long_inc(&tun->dev->rx_dropped);
+				dev_core_stats_rx_dropped_inc(tun->dev);
 				kfree_skb(skb);
 				return -EINVAL;
 			}
@@ -1956,7 +1956,7 @@ drop:
 					  skb_headlen(skb));
 
 		if (unlikely(headlen > skb_headlen(skb))) {
-			atomic_long_inc(&tun->dev->rx_dropped);
+			dev_core_stats_rx_dropped_inc(tun->dev);
 			napi_free_frags(&tfile->napi);
 			rcu_read_unlock();
 			mutex_unlock(&tfile->napi_mutex);
diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c
index 3872f76ea1d3..de97ff98d36e 100644
--- a/drivers/net/vxlan/vxlan_core.c
+++ b/drivers/net/vxlan/vxlan_core.c
@@ -1760,7 +1760,7 @@ static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
 
 	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
 		rcu_read_unlock();
-		atomic_long_inc(&vxlan->dev->rx_dropped);
+		dev_core_stats_rx_dropped_inc(vxlan->dev);
 		vxlan_vnifilter_count(vxlan, vni, vninode,
 				      VXLAN_VNI_STATS_RX_DROPS, 0);
 		goto drop;