[PATCH net-next 4/8] net: make softnet_data.dropped an atomic_t

[Eric Dumazet <edumazet@google.com>]

If under extreme cpu backlog pressure enqueue_to_backlog() has
to drop a packet, it could do this without dirtying a cache line
and potentially slowing down the target cpu.

Move sd->dropped into a separate cache line, and make it atomic.

In non pressure mode, this field is not touched, no need to consume
valuable space in a hot cache line.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/netdevice.h |  3 ++-
 net/core/dev.c            | 13 +++++++++----
 net/core/net-procfs.c     |  3 ++-
 3 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 70775021cc269e0983f538619115237b0067d408..1c31cd2691d32064613836141fbdeeebc831b21f 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3236,10 +3236,11 @@ struct softnet_data {
 	unsigned int		input_queue_tail;
 #endif
 	unsigned int		received_rps;
-	unsigned int		dropped;
 	struct sk_buff_head	input_pkt_queue;
 	struct napi_struct	backlog;
 
+	atomic_t		dropped ____cacheline_aligned_in_smp;
+
 	/* Another possibly contended cache line */
 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
 	int			defer_count;
diff --git a/net/core/dev.c b/net/core/dev.c
index af7a34b0a7d6683c6ffb21dd3388ed678473d95e..be2392ba56bc57bed456e2748b332d4971c83a4e 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4790,17 +4790,22 @@ static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
 	struct softnet_data *sd;
 	unsigned long flags;
 	unsigned int qlen;
+	int max_backlog;
 
 	reason = SKB_DROP_REASON_DEV_READY;
 	if (!netif_running(skb->dev))
 		goto bad_dev;
 
+	reason = SKB_DROP_REASON_CPU_BACKLOG;
 	sd = &per_cpu(softnet_data, cpu);
 
+	qlen = skb_queue_len_lockless(&sd->input_pkt_queue);
+	max_backlog = READ_ONCE(net_hotdata.max_backlog);
+	if (unlikely(qlen > max_backlog))
+		goto cpu_backlog_drop;
 	backlog_lock_irq_save(sd, &flags);
 	qlen = skb_queue_len(&sd->input_pkt_queue);
-	if (qlen <= READ_ONCE(net_hotdata.max_backlog) &&
-	    !skb_flow_limit(skb, qlen)) {
+	if (qlen <= max_backlog && !skb_flow_limit(skb, qlen)) {
 		if (qlen) {
 enqueue:
 			__skb_queue_tail(&sd->input_pkt_queue, skb);
@@ -4816,11 +4821,11 @@ static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
 			napi_schedule_rps(sd);
 		goto enqueue;
 	}
-	reason = SKB_DROP_REASON_CPU_BACKLOG;
 
-	sd->dropped++;
 	backlog_unlock_irq_restore(sd, &flags);
 
+cpu_backlog_drop:
+	atomic_inc(&sd->dropped);
 bad_dev:
 	dev_core_stats_rx_dropped_inc(skb->dev);
 	kfree_skb_reason(skb, reason);
diff --git a/net/core/net-procfs.c b/net/core/net-procfs.c
index a97eceb84e61ec347ad132ff0f22c8ce82f12e90..fa6d3969734a6ec154c3444d1b25ee93edfc5588 100644
--- a/net/core/net-procfs.c
+++ b/net/core/net-procfs.c
@@ -144,7 +144,8 @@ static int softnet_seq_show(struct seq_file *seq, void *v)
 	seq_printf(seq,
 		   "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x "
 		   "%08x %08x\n",
-		   sd->processed, sd->dropped, sd->time_squeeze, 0,
+		   sd->processed, atomic_read(&sd->dropped),
+		   sd->time_squeeze, 0,
 		   0, 0, 0, 0, /* was fastroute */
 		   0,	/* was cpu_collision */
 		   sd->received_rps, flow_limit_count,
-- 
2.44.0.478.gd926399ef9-goog