[PATCH net-next 6/8] net: rps: change input_queue_tail_incr_save()

From: Eric Dumazet <edumazet@google.com>
To: "David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	 Paolo Abeni <pabeni@redhat.com>
Cc: netdev@vger.kernel.org, eric.dumazet@gmail.com,
	 Eric Dumazet <edumazet@google.com>
Subject: [PATCH net-next 6/8] net: rps: change input_queue_tail_incr_save()
Date: Thu, 28 Mar 2024 17:03:07 +0000	[thread overview]
Message-ID: <20240328170309.2172584-7-edumazet@google.com> (raw)
In-Reply-To: <20240328170309.2172584-1-edumazet@google.com>

input_queue_tail_incr_save() is incrementing the sd queue_tail
and save it in the flow last_qtail.

Two issues here :

- no lock protects the write on last_qtail, we should use appropriate
  annotations.

- We can perform this write after releasing the per-cpu backlog lock,
  to decrease this lock hold duration (move away the cache line miss)

Also move input_queue_head_incr() and rps helpers to include/net/rps.h,
while adding rps_ prefix to better reflect their role.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/netdevice.h | 15 ---------------
 include/net/rps.h         | 23 +++++++++++++++++++++++
 net/core/dev.c            | 20 ++++++++++++--------
 3 files changed, 35 insertions(+), 23 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 1c31cd2691d32064613836141fbdeeebc831b21f..14f19cc2616452d7e6afbbaa52f8ad3e61a419e9 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3249,21 +3249,6 @@ struct softnet_data {
 	call_single_data_t	defer_csd;
 };
 
-static inline void input_queue_head_incr(struct softnet_data *sd)
-{
-#ifdef CONFIG_RPS
-	sd->input_queue_head++;
-#endif
-}
-
-static inline void input_queue_tail_incr_save(struct softnet_data *sd,
-					      unsigned int *qtail)
-{
-#ifdef CONFIG_RPS
-	*qtail = ++sd->input_queue_tail;
-#endif
-}
-
 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
 
 static inline int dev_recursion_level(void)
diff --git a/include/net/rps.h b/include/net/rps.h
index 7660243e905b92651a41292e04caf72c5f12f26e..c13f829b8556fda63e76544c332f2c089f0d6ea4 100644
--- a/include/net/rps.h
+++ b/include/net/rps.h
@@ -35,6 +35,29 @@ struct rps_dev_flow {
 };
 #define RPS_NO_FILTER 0xffff
 
+static inline u32 rps_input_queue_tail_incr(struct softnet_data *sd)
+{
+#ifdef CONFIG_RPS
+	return ++sd->input_queue_tail;
+#else
+	return 0;
+#endif
+}
+
+static inline void rps_input_queue_tail_save(u32 *dest, u32 tail)
+{
+#ifdef CONFIG_RPS
+	WRITE_ONCE(*dest, tail);
+#endif
+}
+
+static inline void rps_input_queue_head_incr(struct softnet_data *sd)
+{
+#ifdef CONFIG_RPS
+	sd->input_queue_head++;
+#endif
+}
+
 /*
  * The rps_dev_flow_table structure contains a table of flow mappings.
  */
diff --git a/net/core/dev.c b/net/core/dev.c
index 4e52745f23412bac6d3ff1b9f4d9f2ce4a2eb666..1fe7c6b10793d45a03461ee581d240d2442f9e17 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4601,7 +4601,7 @@ static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
 		if (unlikely(tcpu != next_cpu) &&
 		    (tcpu >= nr_cpu_ids || !cpu_online(tcpu) ||
 		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
-		      rflow->last_qtail)) >= 0)) {
+		      READ_ONCE(rflow->last_qtail))) >= 0)) {
 			tcpu = next_cpu;
 			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
 		}
@@ -4656,7 +4656,7 @@ bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
 		cpu = READ_ONCE(rflow->cpu);
 		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&
 		    ((int)(per_cpu(softnet_data, cpu).input_queue_head -
-			   rflow->last_qtail) <
+			   READ_ONCE(rflow->last_qtail)) <
 		     (int)(10 * flow_table->mask)))
 			expire = false;
 	}
@@ -4791,6 +4791,7 @@ static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
 	unsigned long flags;
 	unsigned int qlen;
 	int max_backlog;
+	u32 tail;
 
 	reason = SKB_DROP_REASON_DEV_READY;
 	if (!netif_running(skb->dev))
@@ -4815,8 +4816,11 @@ static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
 				napi_schedule_rps(sd);
 		}
 		__skb_queue_tail(&sd->input_pkt_queue, skb);
-		input_queue_tail_incr_save(sd, qtail);
+		tail = rps_input_queue_tail_incr(sd);
 		backlog_unlock_irq_restore(sd, &flags);
+
+		/* save the tail outside of the critical section */
+		rps_input_queue_tail_save(qtail, tail);
 		return NET_RX_SUCCESS;
 	}
 
@@ -5894,7 +5898,7 @@ static void flush_backlog(struct work_struct *work)
 		if (skb->dev->reg_state == NETREG_UNREGISTERING) {
 			__skb_unlink(skb, &sd->input_pkt_queue);
 			dev_kfree_skb_irq(skb);
-			input_queue_head_incr(sd);
+			rps_input_queue_head_incr(sd);
 		}
 	}
 	backlog_unlock_irq_enable(sd);
@@ -5903,7 +5907,7 @@ static void flush_backlog(struct work_struct *work)
 		if (skb->dev->reg_state == NETREG_UNREGISTERING) {
 			__skb_unlink(skb, &sd->process_queue);
 			kfree_skb(skb);
-			input_queue_head_incr(sd);
+			rps_input_queue_head_incr(sd);
 		}
 	}
 	local_bh_enable();
@@ -6031,7 +6035,7 @@ static int process_backlog(struct napi_struct *napi, int quota)
 			rcu_read_lock();
 			__netif_receive_skb(skb);
 			rcu_read_unlock();
-			input_queue_head_incr(sd);
+			rps_input_queue_head_incr(sd);
 			if (++work >= quota)
 				return work;
 
@@ -11445,11 +11449,11 @@ static int dev_cpu_dead(unsigned int oldcpu)
 	/* Process offline CPU's input_pkt_queue */
 	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
 		netif_rx(skb);
-		input_queue_head_incr(oldsd);
+		rps_input_queue_head_incr(oldsd);
 	}
 	while ((skb = skb_dequeue(&oldsd->input_pkt_queue))) {
 		netif_rx(skb);
-		input_queue_head_incr(oldsd);
+		rps_input_queue_head_incr(oldsd);
 	}
 
 	return 0;
-- 
2.44.0.478.gd926399ef9-goog

