[IPV6]: Add IPV6_V6ONLY socket option support.

Cherry-picked from Linux 524354b4d086a4f013343d727eaccb7b4c39eb25

* pfinet/glue-include/linux/ipv6.h: Include linux/config.h>
(__ipv6_only_sock, ipv6_only_sock): New macros
* pfinet/linux-src/include/linux/ipv6.h: Likewise.
* pfinet/linux-src/include/linux/in6.h (IPV6_V6ONLY): New macro.
* pfinet/linux-src/include/linux/sysctl.h (NET_IPV6_BINDV6ONLY): New macro.
* pfinet/linux-src/include/net/ipv6.h (sysctl_ipv6_bindv6only): Declare
variable.
* pfinet/linux-src/include/net/sock.h (ipv6_pinfo): Add ipv6only field.
* pfinet/linux-src/net/ipv4/tcp_ipv4.c: Include linux/ipv6.h.
(tcp_v4_get_port, tcp_v4_lookup_listener): Test for ipv6_only_sock.
* pfinet/linux-src/net/ipv4/udp.c: Include linux/ipv6.h.
(udp_v4_get_port, udp_v4_lookup_longway, udp_v4_mcast_next): Test for
ipv6_only_sock.
* pfinet/linux-src/net/ipv6/af_inet6.c (sysctl_ipv6_bindv6only): New
variable.
(inet6_create): Initialize ipv6only field to sysctl_ipv6_bindv6only.
* pfinet/linux-src/net/ipv6/ipv6_sockglue.c (ipv6_setsockopt): Test for
ipv6_only_sock.
(ipv6_setsockopt, ipv6_getsockopt): Support IPV6_V6ONLY case.
* pfinet/linux-src/net/ipv6/tcp_ipv6.c (ipv6_rcv_saddr_equal): New inline
function.
(tcp_v6_get_port): Replace old tests with ipv6_rcv_saddr_equal.
(tcp_v6_connect): Test for __ipv6_only_sock.
* pfinet/linux-src/net/ipv6/udp_ipv6.c (udv6_rcv_saddr_equal): New inline
function.
(udp_v6_get_port): Replace old tests with udv6_rcv_saddr_equal.
(udpv6_connect, udpv6_sendmsg): Test for __ipv6_only_sock.

11 files changed, 162 insertions, 34 deletions

diff --git a/pfinet/linux-src/include/linux/in6.h b/pfinet/linux-src/include/linux/in6.h
index ca5e768b..35e018c7 100644
--- a/pfinet/linux-src/include/linux/in6.h
+++ b/pfinet/linux-src/include/linux/in6.h
@@ -179,6 +179,7 @@ struct in6_flowlabel_req
 #define IPV6_MTU_DISCOVER	23
 #define IPV6_MTU		24
 #define IPV6_RECVERR		25
+#define IPV6_V6ONLY		26
 
 /* IPV6_MTU_DISCOVER values */
 #define IPV6_PMTUDISC_DONT		0
diff --git a/pfinet/linux-src/include/linux/ipv6.h b/pfinet/linux-src/include/linux/ipv6.h
index 84564bae..299db1a5 100644
--- a/pfinet/linux-src/include/linux/ipv6.h
+++ b/pfinet/linux-src/include/linux/ipv6.h
@@ -1,6 +1,7 @@
 #ifndef _IPV6_H
 #define _IPV6_H
 
+#include <linux/config.h>
 #include <linux/in6.h>
 #include <asm/byteorder.h>
 
@@ -118,6 +119,14 @@ struct inet6_skb_parm
 	__u16			dst1;
 };
 
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+#define __ipv6_only_sock(sk)   (sk->net_pinfo.af_inet6.ipv6only)
+#define ipv6_only_sock(sk)     ((sk)->family == PF_INET6 && __ipv6_only_sock(sk))
+#else
+#define __ipv6_only_sock(sk)   0
+#define ipv6_only_sock(sk)     0
+#endif
+
 #endif
 
 #endif
diff --git a/pfinet/linux-src/include/linux/sysctl.h b/pfinet/linux-src/include/linux/sysctl.h
index 00bebabe..1067bfbb 100644
--- a/pfinet/linux-src/include/linux/sysctl.h
+++ b/pfinet/linux-src/include/linux/sysctl.h
@@ -276,7 +276,8 @@ enum
 enum {
 	NET_IPV6_CONF=16,
 	NET_IPV6_NEIGH=17,
-	NET_IPV6_ROUTE=18
+	NET_IPV6_ROUTE=18,
+	NET_IPV6_BINDV6ONLY=20,
 };
 
 enum {
diff --git a/pfinet/linux-src/include/net/ipv6.h b/pfinet/linux-src/include/net/ipv6.h
index 50e2299f..77520eeb 100644
--- a/pfinet/linux-src/include/net/ipv6.h
+++ b/pfinet/linux-src/include/net/ipv6.h
@@ -86,6 +86,9 @@ struct frag_hdr {
 
 #include <net/sock.h>
 
+/* sysctls */
+extern int sysctl_ipv6_bindv6only;
+
 extern struct ipv6_mib		ipv6_statistics;
 extern struct icmpv6_mib	icmpv6_statistics;
 extern struct udp_mib		udp_stats_in6;
diff --git a/pfinet/linux-src/include/net/sock.h b/pfinet/linux-src/include/net/sock.h
index 7789ec40..96fdfe8b 100644
--- a/pfinet/linux-src/include/net/sock.h
+++ b/pfinet/linux-src/include/net/sock.h
@@ -170,7 +170,8 @@ struct ipv6_pinfo {
 	__u8			mc_loop:1,
 	                        recverr:1,
 	                        sndflow:1,
-	                        pmtudisc:2;
+	                        pmtudisc:2,
+	                        ipv6only:1;
 
 	struct ipv6_mc_socklist	*ipv6_mc_list;
 	struct ipv6_fl_socklist *ipv6_fl_list;
diff --git a/pfinet/linux-src/net/ipv4/tcp_ipv4.c b/pfinet/linux-src/net/ipv4/tcp_ipv4.c
index 60d2bdb4..df2c8b7c 100644
--- a/pfinet/linux-src/net/ipv4/tcp_ipv4.c
+++ b/pfinet/linux-src/net/ipv4/tcp_ipv4.c
@@ -45,9 +45,13 @@
  *	Vitaly E. Lavrov	:	Transparent proxy revived after year coma.
  *	Andi Kleen		:	Fix new listen.
  *	Andi Kleen		:	Fix accept error reporting.
+ *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
+ *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
+ *					a single port at the same time.
  */
 
 #include <linux/config.h>
+
 #include <linux/types.h>
 #include <linux/fcntl.h>
 #include <linux/random.h>
@@ -61,6 +65,7 @@
 #include <asm/segment.h>
 
 #include <linux/inet.h>
+#include <linux/ipv6.h>
 #include <linux/stddef.h>
 
 extern int sysctl_tcp_timestamps;
@@ -258,7 +263,8 @@ static int tcp_v4_get_port(struct sock *sk, unsigned short snum)
 			int sk_reuse = sk->reuse;
 
 			for( ; sk2 != NULL; sk2 = sk2->bind_next) {
-				if (sk->bound_dev_if == sk2->bound_dev_if) {
+				if (!ipv6_only_sock(sk2) &&
+				    sk->bound_dev_if == sk2->bound_dev_if) {
 					if (!sk_reuse	||
 					    !sk2->reuse	||
 					    sk2->state == TCP_LISTEN) {
@@ -378,23 +384,23 @@ static struct sock *tcp_v4_lookup_listener(u32 daddr, unsigned short hnum, int d
 	struct sock *result = NULL;
 	int score, hiscore;
 
-	hiscore=0;
+	hiscore=-1;
 	for(sk = tcp_listening_hash[tcp_lhashfn(hnum)]; sk; sk = sk->next) {
-		if(sk->num == hnum) {
+		if(sk->num == hnum && !ipv6_only_sock(sk)) {
 			__u32 rcv_saddr = sk->rcv_saddr;
 
-			score = 1;
+			score = (sk->family == PF_INET ? 1 : 0);
 			if(rcv_saddr) {
 				if (rcv_saddr != daddr)
 					continue;
-				score++;
+				score+=2;
 			}
 			if (sk->bound_dev_if) {
 				if (sk->bound_dev_if != dif)
 					continue;
-				score++;
+				score+=2;
 			}
-			if (score == 3)
+			if (score == 5)
 				return sk;
 			if (score > hiscore) {
 				hiscore = score;
diff --git a/pfinet/linux-src/net/ipv4/udp.c b/pfinet/linux-src/net/ipv4/udp.c
index 1ceb43e4..f9be2e04 100644
--- a/pfinet/linux-src/net/ipv4/udp.c
+++ b/pfinet/linux-src/net/ipv4/udp.c
@@ -61,6 +61,9 @@
  *					return ENOTCONN for unconnected sockets (POSIX)
  *		Janos Farkas	:	don't deliver multi/broadcasts to a different
  *					bound-to-device socket
+ *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
+ *	Alexey Kuznetsov:		allow both IPv4 and IPv6 sockets to bind
+ *					a single port at the same time.
  *
  *
  *		This program is free software; you can redistribute it and/or
@@ -104,6 +107,7 @@
 #include <linux/mm.h>
 #include <linux/config.h>
 #include <linux/inet.h>
+#include <linux/ipv6.h>
 #include <linux/netdevice.h>
 #include <net/snmp.h>
 #include <net/ip.h>
@@ -178,6 +182,7 @@ gotit:
 		     sk2 = sk2->next) {
 			if (sk2->num == snum &&
 			    sk2 != sk &&
+			    !ipv6_only_sock(sk2) &&
 			    sk2->bound_dev_if == sk->bound_dev_if &&
 			    (!sk2->rcv_saddr ||
 			     !sk->rcv_saddr ||
@@ -236,29 +241,30 @@ struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport, u32 daddr, u16 dport, i
 	int badness = -1;
 
 	for(sk = udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]; sk != NULL; sk = sk->next) {
-		if((sk->num == hnum) && !(sk->dead && (sk->state == TCP_CLOSE))) {
-			int score = 0;
+	  if((sk->num == hnum) && !ipv6_only_sock(sk)
+	      && !(sk->dead && (sk->state == TCP_CLOSE))) {
+			int score = (sk->family == PF_INET ? 1 : 0);
 			if(sk->rcv_saddr) {
 				if(sk->rcv_saddr != daddr)
 					continue;
-				score++;
+				score+=2;
 			}
 			if(sk->daddr) {
 				if(sk->daddr != saddr)
 					continue;
-				score++;
+				score+=2;
 			}
 			if(sk->dport) {
 				if(sk->dport != sport)
 					continue;
-				score++;
+				score+=2;
 			}
 			if(sk->bound_dev_if) {
 				if(sk->bound_dev_if != dif)
 					continue;
-				score++;
+				score+=2;
 			}
-			if(score == 4) {
+			if(score == 9) {
 				result = sk;
 				break;
 			} else if(score > badness) {
@@ -389,6 +395,7 @@ static inline struct sock *udp_v4_mcast_next(struct sock *sk,
 		    (s->daddr && s->daddr!=raddr)			||
 		    (s->dport != rnum && s->dport != 0)			||
 		    (s->rcv_saddr  && s->rcv_saddr != laddr)		||
+		    ipv6_only_sock(s)					||
 		    (s->bound_dev_if && s->bound_dev_if != dif))
 			continue;
 		break;
diff --git a/pfinet/linux-src/net/ipv6/af_inet6.c b/pfinet/linux-src/net/ipv6/af_inet6.c
index ca428188..f10ff8d7 100644
--- a/pfinet/linux-src/net/ipv6/af_inet6.c
+++ b/pfinet/linux-src/net/ipv6/af_inet6.c
@@ -84,6 +84,8 @@ extern void ipv6_sysctl_register(void);
 extern void ipv6_sysctl_unregister(void);
 #endif
 
+int sysctl_ipv6_bindv6only;
+
 static int inet6_create(struct socket *sock, int protocol)
 {
 	struct sock *sk;
@@ -137,6 +139,8 @@ static int inet6_create(struct socket *sock, int protocol)
 	sk->net_pinfo.af_inet6.mc_loop	  = 1;
 	sk->net_pinfo.af_inet6.pmtudisc	  = IPV6_PMTUDISC_WANT;
 
+	sk->net_pinfo.af_inet6.ipv6only	  = sysctl_ipv6_bindv6only;
+
 	/* Init the ipv4 part of the socket since we can have sockets
 	 * using v6 API for ipv4.
 	 */
diff --git a/pfinet/linux-src/net/ipv6/ipv6_sockglue.c b/pfinet/linux-src/net/ipv6/ipv6_sockglue.c
index 6a48d1be..f1f67811 100644
--- a/pfinet/linux-src/net/ipv6/ipv6_sockglue.c
+++ b/pfinet/linux-src/net/ipv6/ipv6_sockglue.c
@@ -146,7 +146,8 @@ int ipv6_setsockopt(struct sock *sk, int level, int optname, char *optval,
 				goto addrform_done;
 			}
 
-			if (!(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
+			if (ipv6_only_sock(sk) ||
+			    !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
 				retv = -EADDRNOTAVAIL;
 				goto addrform_done;
 			}
@@ -181,6 +182,15 @@ addrform_done:
 		}
 		break;
 
+	case IPV6_V6ONLY:
+		if (sk->num) {
+			retv = -EINVAL;
+			goto out;
+		}
+		np->ipv6only = valbool;
+		retv = 0;
+		break;
+
 	case IPV6_PKTINFO:
 		np->rxopt.bits.rxinfo = valbool;
 		retv = 0;
@@ -395,6 +405,9 @@ int ipv6_getsockopt(struct sock *sk, int level, int optname, char *optval,
 		if (!val)
 			return -ENOTCONN;
 		break;
+	case IPV6_V6ONLY:
+		val = np->ipv6only;
+		break;
 	default:
 		return -EINVAL;
 	}
diff --git a/pfinet/linux-src/net/ipv6/tcp_ipv6.c b/pfinet/linux-src/net/ipv6/tcp_ipv6.c
index c761e76c..3fba9af6 100644
--- a/pfinet/linux-src/net/ipv6/tcp_ipv6.c
+++ b/pfinet/linux-src/net/ipv6/tcp_ipv6.c
@@ -4,6 +4,9 @@
  *
  *	Authors:
  *	Pedro Roque		<roque@di.fc.ul.pt>	
+ *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
+ *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
+ *					a single port at the same time.
  *
  *	$Id: tcp_ipv6.c,v 1.3 2007/10/13 01:43:00 stesie Exp $
  *
@@ -79,6 +82,43 @@ static __inline__ int tcp_v6_sk_hashfn(struct sock *sk)
 	return tcp_v6_hashfn(laddr, lport, faddr, fport);
 }
 
+static inline int ipv6_rcv_saddr_equal(struct sock *sk, struct sock *sk2)
+{
+	struct ipv6_pinfo *np = &sk->net_pinfo.af_inet6;
+	int addr_type = ipv6_addr_type(&np->rcv_saddr);
+
+	if (!sk2->rcv_saddr && !ipv6_only_sock(sk))
+		return 1;
+
+	if (sk2->family == AF_INET6 &&
+	    ipv6_addr_any(&sk2->rcv_saddr) &&
+	    !(ipv6_only_sock(sk2) && addr_type == IPV6_ADDR_MAPPED))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_ANY &&
+	    (!ipv6_only_sock(sk) ||
+	     !(sk2->family == AF_INET6 ?
+	       ipv6_addr_type(&sk2->rcv_saddr) == IPV6_ADDR_MAPPED : 1)))
+		return 1;
+
+	if (sk2->family == AF_INET6 &&
+	    !ipv6_addr_cmp(&np->rcv_saddr,
+			   (sk2->state != TCP_TIME_WAIT ?
+			    &sk2->rcv_saddr :
+			    &((struct tcp_tw_bucket *)sk)->v6_rcv_saddr)))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_MAPPED &&
+	    !ipv6_only_sock(sk2) &&
+	    (!sk2->rcv_saddr ||
+	     !sk->rcv_saddr ||
+	     sk->rcv_saddr == sk2->rcv_saddr))
+		return 1;
+
+	return 0;
+}
+
+
 /* Grrr, addr_type already calculated by caller, but I don't want
  * to add some silly "cookie" argument to this method just for that.
  * But it doesn't matter, the recalculation is in the rarest path
@@ -134,18 +174,11 @@ static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
 
 			for( ; sk2 != NULL; sk2 = sk2->bind_next) {
 				if (sk->bound_dev_if == sk2->bound_dev_if) {
-					if (!sk_reuse	||
+					if ((!sk_reuse	||
 					    !sk2->reuse	||
-					    sk2->state == TCP_LISTEN) {
-						/* NOTE: IPv6 tw bucket have different format */
-						if (!sk2->rcv_saddr	||
-						    addr_type == IPV6_ADDR_ANY ||
-						    !ipv6_addr_cmp(&sk->net_pinfo.af_inet6.rcv_saddr,
-								   sk2->state != TCP_TIME_WAIT ?
-								   &sk2->net_pinfo.af_inet6.rcv_saddr :
-								   &((struct tcp_tw_bucket*)sk)->v6_rcv_saddr))
-							break;
-					}
+					    sk2->state == TCP_LISTEN) &&
+					  ipv6_rcv_saddr_equal(sk, sk2))
+						break;
 				}
 			}
 			/* If we found a conflict, fail. */
@@ -460,6 +493,9 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 
 		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
 
+		if (__ipv6_only_sock(sk))
+			return -ENETUNREACH;
+
 		sin.sin_family = AF_INET;
 		sin.sin_port = usin->sin6_port;
 		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
diff --git a/pfinet/linux-src/net/ipv6/udp_ipv6.c b/pfinet/linux-src/net/ipv6/udp_ipv6.c
index f6968ae4..bbc4f027 100644
--- a/pfinet/linux-src/net/ipv6/udp_ipv6.c
+++ b/pfinet/linux-src/net/ipv6/udp_ipv6.c
@@ -5,6 +5,10 @@
  *	Authors:
  *	Pedro Roque		<roque@di.fc.ul.pt>	
  *
+ *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
+ *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
+ *					a single port at the same time.
+ *
  *	Based on linux/ipv4/udp.c
  *
  *	$Id: udp_ipv6.c,v 1.3 2007/10/13 01:43:00 stesie Exp $
@@ -46,6 +50,41 @@
 
 struct udp_mib udp_stats_in6;
 
+static __inline__ int udv6_rcv_saddr_equal(struct sock *sk, struct sock *sk2)
+{
+	struct ipv6_pinfo *np = &sk->net_pinfo.af_inet6;
+	int addr_type = ipv6_addr_type(&np->rcv_saddr);
+
+	if (!sk2->rcv_saddr && !ipv6_only_sock(sk))
+		return 1;
+
+	if (sk2->family == AF_INET6 &&
+	    ipv6_addr_any(&sk2->rcv_saddr) &&
+	    !(ipv6_only_sock(sk2) && addr_type == IPV6_ADDR_MAPPED))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_ANY &&
+	    (!ipv6_only_sock(sk) ||
+	     !(sk2->family == AF_INET6 ?
+	       ipv6_addr_type(&sk2->rcv_saddr) == IPV6_ADDR_MAPPED : 1)))
+		return 1;
+
+	if (sk2->family == AF_INET6 &&
+	    !ipv6_addr_cmp(&sk->rcv_saddr,
+			   &sk2->rcv_saddr))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_MAPPED &&
+	    !ipv6_only_sock(sk2) &&
+	    (!sk2->rcv_saddr ||
+	     !sk->rcv_saddr ||
+	     sk->rcv_saddr == sk2->rcv_saddr))
+		return 1;
+
+	return 0;
+}
+
+
 /* Grrr, addr_type already calculated by caller, but I don't want
  * to add some silly "cookie" argument to this method just for that.
  */
@@ -104,11 +143,8 @@ gotit:
 			if (sk2->num == snum &&
 			    sk2 != sk &&
 			    sk2->bound_dev_if == sk->bound_dev_if &&
-			    (!sk2->rcv_saddr ||
-			     addr_type == IPV6_ADDR_ANY ||
-			     !ipv6_addr_cmp(&sk->net_pinfo.af_inet6.rcv_saddr,
-					    &sk2->net_pinfo.af_inet6.rcv_saddr)) &&
-			    (!sk2->reuse || !sk->reuse))
+			    (!sk2->reuse || !sk->reuse) &&
+			    udv6_rcv_saddr_equal(sk, sk2))
 				goto fail;
 		}
 	}
@@ -208,6 +244,8 @@ int udpv6_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 	int			err;
 
 	if (usin->sin6_family == AF_INET) {
+		if (__ipv6_only_sock(sk))
+			return -EAFNOSUPPORT;
 		err = udp_connect(sk, uaddr, addr_len);
 		goto ipv4_connected;
 	}
@@ -251,6 +289,9 @@ int udpv6_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 	if (addr_type == IPV6_ADDR_MAPPED) {
 		struct sockaddr_in sin;
 
+		if (__ipv6_only_sock(sk))
+			return -ENETUNREACH;
+
 		sin.sin_family = AF_INET;
 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
 		sin.sin_port = usin->sin6_port;
@@ -806,8 +847,11 @@ static int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, int ulen)
 	fl.fl6_flowlabel = 0;
 
 	if (sin6) {
-		if (sin6->sin6_family == AF_INET)
+		if (sin6->sin6_family == AF_INET) {
+			if (__ipv6_only_sock(sk))
+				return -ENETUNREACH;
 			return udp_sendmsg(sk, msg, ulen);
+		}
 
 		if (addr_len < sizeof(*sin6))
 			return(-EINVAL);
@@ -854,6 +898,9 @@ static int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, int ulen)
 	if (addr_type == IPV6_ADDR_MAPPED) {
 		struct sockaddr_in sin;
 
+		if (__ipv6_only_sock(sk))
+			return -ENETUNREACH;
+
 		sin.sin_family = AF_INET;
 		sin.sin_addr.s_addr = daddr->s6_addr32[3];
 		sin.sin_port = udh.uh.dest;