Merge branch 'net-atomic-dev-stats'

Eric Dumazet says:

====================
net: add atomic dev->stats infra

Long standing KCSAN issues are caused by data-race around
some dev->stats changes.

Most performance critical paths already use per-cpu
variables, or per-queue ones.

It is reasonable (and more correct) to use atomic operations
for the slow paths.

First patch adds the infrastructure, then three patches address
the most common paths that syzbot is playing with.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 37 insertions, 26 deletions

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 02a2318da7c7..23b3903b0678 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -171,31 +171,38 @@ static inline bool dev_xmit_complete(int rc)
  *	(unsigned long) so they can be read and written atomically.
  */
 
+#define NET_DEV_STAT(FIELD)			\
+	union {					\
+		unsigned long FIELD;		\
+		atomic_long_t __##FIELD;	\
+	}
+
 struct net_device_stats {
-	unsigned long	rx_packets;
-	unsigned long	tx_packets;
-	unsigned long	rx_bytes;
-	unsigned long	tx_bytes;
-	unsigned long	rx_errors;
-	unsigned long	tx_errors;
-	unsigned long	rx_dropped;
-	unsigned long	tx_dropped;
-	unsigned long	multicast;
-	unsigned long	collisions;
-	unsigned long	rx_length_errors;
-	unsigned long	rx_over_errors;
-	unsigned long	rx_crc_errors;
-	unsigned long	rx_frame_errors;
-	unsigned long	rx_fifo_errors;
-	unsigned long	rx_missed_errors;
-	unsigned long	tx_aborted_errors;
-	unsigned long	tx_carrier_errors;
-	unsigned long	tx_fifo_errors;
-	unsigned long	tx_heartbeat_errors;
-	unsigned long	tx_window_errors;
-	unsigned long	rx_compressed;
-	unsigned long	tx_compressed;
+	NET_DEV_STAT(rx_packets);
+	NET_DEV_STAT(tx_packets);
+	NET_DEV_STAT(rx_bytes);
+	NET_DEV_STAT(tx_bytes);
+	NET_DEV_STAT(rx_errors);
+	NET_DEV_STAT(tx_errors);
+	NET_DEV_STAT(rx_dropped);
+	NET_DEV_STAT(tx_dropped);
+	NET_DEV_STAT(multicast);
+	NET_DEV_STAT(collisions);
+	NET_DEV_STAT(rx_length_errors);
+	NET_DEV_STAT(rx_over_errors);
+	NET_DEV_STAT(rx_crc_errors);
+	NET_DEV_STAT(rx_frame_errors);
+	NET_DEV_STAT(rx_fifo_errors);
+	NET_DEV_STAT(rx_missed_errors);
+	NET_DEV_STAT(tx_aborted_errors);
+	NET_DEV_STAT(tx_carrier_errors);
+	NET_DEV_STAT(tx_fifo_errors);
+	NET_DEV_STAT(tx_heartbeat_errors);
+	NET_DEV_STAT(tx_window_errors);
+	NET_DEV_STAT(rx_compressed);
+	NET_DEV_STAT(tx_compressed);
 };
+#undef NET_DEV_STAT
 
 /* per-cpu stats, allocated on demand.
  * Try to fit them in a single cache line, for dev_get_stats() sake.
@@ -5171,4 +5178,9 @@ extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
 
 extern struct net_device *blackhole_netdev;
 
+/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
+#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
+#define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
+		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
+
 #endif	/* _LINUX_NETDEVICE_H */
diff --git a/include/net/dst.h b/include/net/dst.h
index 00b479ce6b99..d67fda89cd0f 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -356,9 +356,8 @@ static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 				 struct net *net)
 {
-	/* TODO : stats should be SMP safe */
-	dev->stats.rx_packets++;
-	dev->stats.rx_bytes += skb->len;
+	DEV_STATS_INC(dev, rx_packets);
+	DEV_STATS_ADD(dev, rx_bytes, skb->len);
 	__skb_tunnel_rx(skb, dev, net);
 }