net: move rps_sock_flow_table to net_hotdata

rps_sock_flow_table and rps_cpu_mask are used in fast path.

Move them to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-19-edumazet@google.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

include/net/hotdata.h

@@ -27,6 +27,10 @@ struct net_hotdata {
 	struct kmem_cache	*skbuff_cache;
 	struct kmem_cache	*skbuff_fclone_cache;
 	struct kmem_cache	*skb_small_head_cache;
+#ifdef CONFIG_RPS
+	struct rps_sock_flow_table __rcu *rps_sock_flow_table;
+	u32			rps_cpu_mask;
+#endif
 	int			gro_normal_batch;
 	int			netdev_budget;
 	int			netdev_budget_usecs;

include/net/rps.h

@@ -5,6 +5,7 @@
 #include <linux/types.h>
 #include <linux/static_key.h>
 #include <net/sock.h>
+#include <net/hotdata.h>
 
 #ifdef CONFIG_RPS
 
@@ -64,14 +65,11 @@ struct rps_sock_flow_table {
 
 #define RPS_NO_CPU 0xffff
 
-extern u32 rps_cpu_mask;
-extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
-
 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
 					u32 hash)
 {
 	unsigned int index = hash & table->mask;
-	u32 val = hash & ~rps_cpu_mask;
+	u32 val = hash & ~net_hotdata.rps_cpu_mask;
 
 	/* We only give a hint, preemption can change CPU under us */
 	val |= raw_smp_processor_id();
@@ -93,7 +91,7 @@ static inline void sock_rps_record_flow_hash(__u32 hash)
 	if (!hash)
 		return;
 	rcu_read_lock();
-	sock_flow_table = rcu_dereference(rps_sock_flow_table);
+	sock_flow_table = rcu_dereference(net_hotdata.rps_sock_flow_table);
 	if (sock_flow_table)
 		rps_record_sock_flow(sock_flow_table, hash);
 	rcu_read_unlock();

net/core/dev.c

@@ -4450,12 +4450,6 @@ static inline void ____napi_schedule(struct softnet_data *sd,
 
 #ifdef CONFIG_RPS
 
-/* One global table that all flow-based protocols share. */
-struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
-EXPORT_SYMBOL(rps_sock_flow_table);
-u32 rps_cpu_mask __read_mostly;
-EXPORT_SYMBOL(rps_cpu_mask);
-
 struct static_key_false rps_needed __read_mostly;
 EXPORT_SYMBOL(rps_needed);
 struct static_key_false rfs_needed __read_mostly;
@@ -4547,7 +4541,7 @@ static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
 	if (!hash)
 		goto done;
 
-	sock_flow_table = rcu_dereference(rps_sock_flow_table);
+	sock_flow_table = rcu_dereference(net_hotdata.rps_sock_flow_table);
 	if (flow_table && sock_flow_table) {
 		struct rps_dev_flow *rflow;
 		u32 next_cpu;
@@ -4557,10 +4551,10 @@ static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
 		 * This READ_ONCE() pairs with WRITE_ONCE() from rps_record_sock_flow().
 		 */
 		ident = READ_ONCE(sock_flow_table->ents[hash & sock_flow_table->mask]);
-		if ((ident ^ hash) & ~rps_cpu_mask)
+		if ((ident ^ hash) & ~net_hotdata.rps_cpu_mask)
 			goto try_rps;
 
-		next_cpu = ident & rps_cpu_mask;
+		next_cpu = ident & net_hotdata.rps_cpu_mask;
 
 		/* OK, now we know there is a match,
 		 * we can look at the local (per receive queue) flow table

net/core/sysctl_net_core.c

@@ -140,7 +140,8 @@ static int rps_sock_flow_sysctl(struct ctl_table *table, int write,
 
 	mutex_lock(&sock_flow_mutex);
 
-	orig_sock_table = rcu_dereference_protected(rps_sock_flow_table,
+	orig_sock_table = rcu_dereference_protected(
+					net_hotdata.rps_sock_flow_table,
 					lockdep_is_held(&sock_flow_mutex));
 	size = orig_size = orig_sock_table ? orig_sock_table->mask + 1 : 0;
 
@@ -161,7 +162,8 @@ static int rps_sock_flow_sysctl(struct ctl_table *table, int write,
 					mutex_unlock(&sock_flow_mutex);
 					return -ENOMEM;
 				}
-				rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1;
+				net_hotdata.rps_cpu_mask =
+					roundup_pow_of_two(nr_cpu_ids) - 1;
 				sock_table->mask = size - 1;
 			} else
 				sock_table = orig_sock_table;
@@ -172,7 +174,8 @@ static int rps_sock_flow_sysctl(struct ctl_table *table, int write,
 			sock_table = NULL;
 
 		if (sock_table != orig_sock_table) {
-			rcu_assign_pointer(rps_sock_flow_table, sock_table);
+			rcu_assign_pointer(net_hotdata.rps_sock_flow_table,
+					   sock_table);
 			if (sock_table) {
 				static_branch_inc(&rps_needed);
 				static_branch_inc(&rfs_needed);