Merge branch 'udp_tunnel-speed-up-udp-tunnel-device-destruction-part-i'

Kuniyuki Iwashima says:

====================
udp_tunnel: Speed up UDP tunnel device destruction (Part I)

Most of the UDP tunnel devices call synchronize_rcu() twice
during destruction, for example, vxlan has

  1) synchronize_rcu() in udp_tunnel_sock_release()

  2) synchronize_net() in vxlan_sock_release()

The goal of this series is to remove the former, and another
followup series removes the latter.

synchronize_rcu() was added in udp_tunnel_sock_release() by
commit 3cf7203ca620 ("net/tunnel: wait until all sk_user_data
reader finish before releasing the sock").

This was intended to protect the fast path of a dying vxlan
from dereferencing vxlan_sock->sock->sk after sock_orphan()
has set sock->sk to NULL.

Most of the UDP tunnel devices store struct socket to its
private struct, but it is NOT needed in the fast paths;
struct sock is used there, but struct socket is only used
for tunnel setup / teardown.

This is probably because UDP tunnel functions accept struct
socket, but even such functions do not need it, except for
udp_tunnel_sock_release(), which can safely access sk->sk_socket.

The overview of the series:

  Patch 1 -  5 : Convert UDP tunnel helper to take struct sock
  Patch 6      : Small fix for 10-years-old bug
  Patch 7 - 14 : Store struct sock in tunnel devices
  Patch 15     : Remove synchronize_rcu() in udp_tunnel_sock_release()

With this change, a script creating/upping vxlan in 4000 netns
runs 10x faster.
====================

[See Link for benchmark results.]

Link: https://patch.msgid.link/20260502031401.3557229-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

3 files changed, 11 insertions, 10 deletions

diff --git a/include/net/amt.h b/include/net/amt.h
index c881bc8b673b..a0255491f5b0 100644
--- a/include/net/amt.h
+++ b/include/net/amt.h
@@ -331,7 +331,7 @@ struct amt_dev {
 	enum amt_status		status;
 	/* Generated key */
 	siphash_key_t		key;
-	struct socket	  __rcu *sock;
+	struct sock	  __rcu *sk;
 	u32			max_groups;
 	u32			max_sources;
 	u32			hash_buckets;
diff --git a/include/net/udp_tunnel.h b/include/net/udp_tunnel.h
index 47c23d4a1740..498b7b262fa9 100644
--- a/include/net/udp_tunnel.h
+++ b/include/net/udp_tunnel.h
@@ -94,7 +94,7 @@ struct udp_tunnel_sock_cfg {
 };
 
 /* Setup the given (UDP) sock to receive UDP encapsulated packets */
-void setup_udp_tunnel_sock(struct net *net, struct socket *sock,
+void setup_udp_tunnel_sock(struct net *net, struct sock *sk,
 			   struct udp_tunnel_sock_cfg *sock_cfg);
 
 /* -- List of parsable UDP tunnel types --
@@ -127,12 +127,12 @@ struct udp_tunnel_info {
 };
 
 /* Notify network devices of offloadable types */
-void udp_tunnel_push_rx_port(struct net_device *dev, struct socket *sock,
+void udp_tunnel_push_rx_port(struct net_device *dev, struct sock *sk,
 			     unsigned short type);
-void udp_tunnel_drop_rx_port(struct net_device *dev, struct socket *sock,
+void udp_tunnel_drop_rx_port(struct net_device *dev, struct sock *sk,
 			     unsigned short type);
-void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type);
-void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type);
+void udp_tunnel_notify_add_rx_port(struct sock *sk, unsigned short type);
+void udp_tunnel_notify_del_rx_port(struct sock *sk, unsigned short type);
 
 /* Transmit the skb using UDP encapsulation. */
 void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
@@ -176,7 +176,7 @@ static inline void udp_tunnel_set_inner_protocol(struct sk_buff *skb,
 		skb_set_inner_protocol(skb, inner_proto);
 }
 
-void udp_tunnel_sock_release(struct socket *sock);
+void udp_tunnel_sock_release(struct sock *sk);
 
 struct rtable *udp_tunnel_dst_lookup(struct sk_buff *skb,
 				     struct net_device *dev,
@@ -188,7 +188,7 @@ struct rtable *udp_tunnel_dst_lookup(struct sk_buff *skb,
 struct dst_entry *udp_tunnel6_dst_lookup(struct sk_buff *skb,
 					 struct net_device *dev,
 					 struct net *net,
-					 struct socket *sock, int oif,
+					 struct sock *sk, int oif,
 					 struct in6_addr *saddr,
 					 const struct ip_tunnel_key *key,
 					 __be16 sport, __be16 dport, u8 dsfield,
diff --git a/include/net/vxlan.h b/include/net/vxlan.h
index 0ee50785f4f1..dfba89695efc 100644
--- a/include/net/vxlan.h
+++ b/include/net/vxlan.h
@@ -185,7 +185,8 @@ struct vxlan_metadata {
 /* per UDP socket information */
 struct vxlan_sock {
 	struct hlist_node hlist;
-	struct socket	 *sock;
+	struct sock	  *sk;
+	struct rcu_head	  rcu;
 	struct hlist_head vni_list[VNI_HASH_SIZE];
 	refcount_t	  refcnt;
 	u32		  flags;
@@ -448,7 +449,7 @@ static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
 
 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
 {
-	return vs->sock->sk->sk_family;
+	return vs->sk->sk_family;
 }
 
 #if IS_ENABLED(CONFIG_IPV6)