net: speedup udp receive path

Since commit 95766fff ([UDP]: Add memory accounting.), 
each received packet needs one extra sock_lock()/sock_release() pair.

This added latency because of possible backlog handling. Then later,
ticket spinlocks added yet another latency source in case of DDOS.

This patch introduces lock_sock_bh() and unlock_sock_bh()
synchronization primitives, avoiding one atomic operation and backlog
processing.

skb_free_datagram_locked() uses them instead of full blown
lock_sock()/release_sock(). skb is orphaned inside locked section for
proper socket memory reclaim, and finally freed outside of it.

UDP receive path now take the socket spinlock only once.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 10 insertions, 0 deletions

diff --git a/include/net/sock.h b/include/net/sock.h
index cf12b1e..d361c77 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1021,6 +1021,16 @@ extern void release_sock(struct sock *sk);
 				SINGLE_DEPTH_NESTING)
 #define bh_unlock_sock(__sk)	spin_unlock(&((__sk)->sk_lock.slock))
 
+static inline void lock_sock_bh(struct sock *sk)
+{
+	spin_lock_bh(&sk->sk_lock.slock);
+}
+
+static inline void unlock_sock_bh(struct sock *sk)
+{
+	spin_unlock_bh(&sk->sk_lock.slock);
+}
+
 extern struct sock		*sk_alloc(struct net *net, int family,
 					  gfp_t priority,
 					  struct proto *prot);