net: introduce SO_MAX_PACING_RATE

As mentioned in commit afe4fd0624 ("pkt_sched: fq: Fair Queue packet
scheduler"), this patch adds a new socket option.

SO_MAX_PACING_RATE offers the application the ability to cap the
rate computed by transport layer. Value is in bytes per second.

u32 val = 1000000;
setsockopt(sockfd, SOL_SOCKET, SO_MAX_PACING_RATE, &val, sizeof(val));

To be effectively paced, a flow must use FQ packet scheduler.

Note that a packet scheduler takes into account the headers for its
computations. The effective payload rate depends on MSS and retransmits
if any.

I chose to make this pacing rate a SOL_SOCKET option instead of a
TCP one because this can be used by other protocols.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

@@ -81,6 +81,8 @@
 
 #define SO_SELECT_ERR_QUEUE	45
 
-#define SO_BUSY_POLL			46
+#define SO_BUSY_POLL		46
+
+#define SO_MAX_PACING_RATE	47
 
 #endif /* _UAPI_ASM_SOCKET_H */

arch/avr32/include/uapi/asm/socket.h

@@ -76,4 +76,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* __ASM_AVR32_SOCKET_H */

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

@@ -78,6 +78,8 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_SOCKET_H */
 
 

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

@@ -76,5 +76,7 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_SOCKET_H */
 

arch/h8300/include/uapi/asm/socket.h

@@ -76,4 +76,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_SOCKET_H */

arch/ia64/include/uapi/asm/socket.h

@@ -85,4 +85,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_IA64_SOCKET_H */

arch/m32r/include/uapi/asm/socket.h

@@ -76,4 +76,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_M32R_SOCKET_H */

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

@@ -94,4 +94,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _UAPI_ASM_SOCKET_H */

arch/mn10300/include/uapi/asm/socket.h

@@ -76,4 +76,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_SOCKET_H */

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

@@ -75,6 +75,8 @@
 
 #define SO_BUSY_POLL		0x4027
 
+#define SO_MAX_PACING_RATE	0x4048
+
 /* O_NONBLOCK clashes with the bits used for socket types.  Therefore we
  * have to define SOCK_NONBLOCK to a different value here.
  */

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

@@ -83,4 +83,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif	/* _ASM_POWERPC_SOCKET_H */

arch/s390/include/uapi/asm/socket.h

@@ -82,4 +82,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* _ASM_SOCKET_H */

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

@@ -72,6 +72,8 @@
 
 #define SO_BUSY_POLL		0x0030
 
+#define SO_MAX_PACING_RATE	0x0031
+
 /* Security levels - as per NRL IPv6 - don't actually do anything */
 #define SO_SECURITY_AUTHENTICATION		0x5001
 #define SO_SECURITY_ENCRYPTION_TRANSPORT	0x5002

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

@@ -87,4 +87,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif	/* _XTENSA_SOCKET_H */

include/net/sock.h

@@ -363,6 +363,7 @@ struct sock {
 	int			sk_wmem_queued;
 	gfp_t			sk_allocation;
 	u32			sk_pacing_rate; /* bytes per second */
+	u32			sk_max_pacing_rate;
 	netdev_features_t	sk_route_caps;
 	netdev_features_t	sk_route_nocaps;
 	int			sk_gso_type;

include/uapi/asm-generic/socket.h

@@ -78,4 +78,6 @@
 
 #define SO_BUSY_POLL		46
 
+#define SO_MAX_PACING_RATE	47
+
 #endif /* __ASM_GENERIC_SOCKET_H */

net/core/sock.c

@@ -914,6 +914,13 @@ set_rcvbuf:
 		}
 		break;
 #endif
+
+	case SO_MAX_PACING_RATE:
+		sk->sk_max_pacing_rate = val;
+		sk->sk_pacing_rate = min(sk->sk_pacing_rate,
+					 sk->sk_max_pacing_rate);
+		break;
+
 	default:
 		ret = -ENOPROTOOPT;
 		break;
@@ -1177,6 +1184,10 @@ int sock_getsockopt(struct socket *sock, int level, int optname,
 		break;
 #endif
 
+	case SO_MAX_PACING_RATE:
+		v.val = sk->sk_max_pacing_rate;
+		break;
+
 	default:
 		return -ENOPROTOOPT;
 	}
@@ -2319,6 +2330,7 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 	sk->sk_ll_usec		=	sysctl_net_busy_read;
 #endif
 
+	sk->sk_max_pacing_rate = ~0U;
 	/*
 	 * Before updating sk_refcnt, we must commit prior changes to memory
 	 * (Documentation/RCU/rculist_nulls.txt for details)

net/ipv4/tcp_input.c

@@ -735,7 +735,7 @@ static void tcp_update_pacing_rate(struct sock *sk)
 	if (tp->srtt > 8 + 2)
 		do_div(rate, tp->srtt);
 
-	sk->sk_pacing_rate = min_t(u64, rate, ~0U);
+	sk->sk_pacing_rate = min_t(u64, rate, sk->sk_max_pacing_rate);
 }
 
 /* Calculate rto without backoff.  This is the second half of Van Jacobson's