txhash: Add socket option to control TX hash rethink behavior

Add the SO_TXREHASH socket option to control hash rethink behavior per socket.
When default mode is set, sockets disable rehash at initialization and use
sysctl option when entering listen state. setsockopt() overrides default
behavior.
Signed-off-by: Akhmat Karakotov <hmukos@yandex-team.ru>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

arch/alpha/include/uapi/asm/socket.h

@@ -133,6 +133,8 @@
 
 #define SO_RESERVE_MEM		73
 
+#define SO_TXREHASH		74
+
 #if !defined(__KERNEL__)
 
 #if __BITS_PER_LONG == 64

arch/mips/include/uapi/asm/socket.h

@@ -144,6 +144,8 @@
 
 #define SO_RESERVE_MEM		73
 
+#define SO_TXREHASH		74
+
 #if !defined(__KERNEL__)
 
 #if __BITS_PER_LONG == 64

arch/parisc/include/uapi/asm/socket.h

@@ -125,6 +125,8 @@
 
 #define SO_RESERVE_MEM		0x4047
 
+#define SO_TXREHASH		0x4048
+
 #if !defined(__KERNEL__)
 
 #if __BITS_PER_LONG == 64

arch/sparc/include/uapi/asm/socket.h

@@ -126,6 +126,8 @@
 
 #define SO_RESERVE_MEM           0x0052
 
+#define SO_TXREHASH              0x0053
+
 
 #if !defined(__KERNEL__)
 

include/net/sock.h

@@ -316,6 +316,7 @@ struct sk_filter;
   *	@sk_rcvtimeo: %SO_RCVTIMEO setting
   *	@sk_sndtimeo: %SO_SNDTIMEO setting
   *	@sk_txhash: computed flow hash for use on transmit
+  *	@sk_txrehash: enable TX hash rethink
   *	@sk_filter: socket filtering instructions
   *	@sk_timer: sock cleanup timer
   *	@sk_stamp: time stamp of last packet received
@@ -491,6 +492,7 @@ struct sock {
 	u32			sk_ack_backlog;
 	u32			sk_max_ack_backlog;
 	kuid_t			sk_uid;
+	u8			sk_txrehash;
 #ifdef CONFIG_NET_RX_BUSY_POLL
 	u8			sk_prefer_busy_poll;
 	u16			sk_busy_poll_budget;
@@ -2066,18 +2068,10 @@ static inline void sk_set_txhash(struct sock *sk)
 
 static inline bool sk_rethink_txhash(struct sock *sk)
 {
-	u8 rehash;
-
-	if (!sk->sk_txhash)
-		return false;
-
-	rehash = READ_ONCE(sock_net(sk)->core.sysctl_txrehash);
-
-	if (rehash) {
+	if (sk->sk_txhash && sk->sk_txrehash == SOCK_TXREHASH_ENABLED) {
 		sk_set_txhash(sk);
 		return true;
 	}
-
 	return false;
 }
 

include/uapi/asm-generic/socket.h

@@ -128,6 +128,8 @@
 
 #define SO_RESERVE_MEM		73
 
+#define SO_TXREHASH		74
+
 #if !defined(__KERNEL__)
 
 #if __BITS_PER_LONG == 64 || (defined(__x86_64__) && defined(__ILP32__))

include/uapi/linux/socket.h

@@ -31,6 +31,7 @@ struct __kernel_sockaddr_storage {
 
 #define SOCK_BUF_LOCK_MASK (SOCK_SNDBUF_LOCK | SOCK_RCVBUF_LOCK)
 
+#define SOCK_TXREHASH_DEFAULT	((u8)-1)
 #define SOCK_TXREHASH_DISABLED	0
 #define SOCK_TXREHASH_ENABLED	1
 

net/core/sock.c

@@ -1447,6 +1447,14 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 	}
 
+	case SO_TXREHASH:
+		if (val < -1 || val > 1) {
+			ret = -EINVAL;
+			break;
+		}
+		sk->sk_txrehash = (u8)val;
+		break;
+
 	default:
 		ret = -ENOPROTOOPT;
 		break;
@@ -1834,6 +1842,10 @@ int sock_getsockopt(struct socket *sock, int level, int optname,
 		v.val = sk->sk_reserved_mem;
 		break;
 
+	case SO_TXREHASH:
+		v.val = sk->sk_txrehash;
+		break;
+
 	default:
 		/* We implement the SO_SNDLOWAT etc to not be settable
 		 * (1003.1g 7).
@@ -3279,6 +3291,7 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 	sk->sk_pacing_rate = ~0UL;
 	WRITE_ONCE(sk->sk_pacing_shift, 10);
 	sk->sk_incoming_cpu = -1;
+	sk->sk_txrehash = SOCK_TXREHASH_DEFAULT;
 
 	sk_rx_queue_clear(sk);
 	/*

net/ipv4/inet_connection_sock.c

@@ -1046,6 +1046,9 @@ int inet_csk_listen_start(struct sock *sk)
 	sk->sk_ack_backlog = 0;
 	inet_csk_delack_init(sk);
 
+	if (sk->sk_txrehash == SOCK_TXREHASH_DEFAULT)
+		sk->sk_txrehash = READ_ONCE(sock_net(sk)->core.sysctl_txrehash);
+
 	/* There is race window here: we announce ourselves listening,
 	 * but this transition is still not validated by get_port().
 	 * It is OK, because this socket enters to hash table only