[IPV6]: Add IPV6_V6ONLY socket option support.

Documentation/networking/ip-sysctl.txt

@@ -462,6 +462,15 @@ ak@muc.de
 IPv6 has no global variables such as tcp_*.  tcp_* settings under ipv4/ also
 apply to IPv6 [XXX?].
 
+bindv6only - BOOLEAN
+	Default value for IPV6_V6ONLY socket option,
+	which restricts use of the IPv6 socket to IPv6 communication 
+	only.
+		TRUE: disable IPv4-mapped address feature
+		FALSE: enable IPv4-mapped address feature
+
+	Default: FALSE (as specified in RFC2553bis)
+
 conf/default/*:
 	Change the interface-specific default settings.
 

include/linux/in6.h

@@ -156,6 +156,7 @@ 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

include/linux/ipv6.h

 #ifndef _IPV6_H
 #define _IPV6_H
 
+#include <linux/config.h>
 #include <linux/in6.h>
 #include <asm/byteorder.h>
 
@@ -152,7 +153,8 @@ 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;
@@ -195,6 +197,15 @@ struct tcp6_sock {
 
 #define inet6_sk(__sk) ((struct raw6_sock *)__sk)->pinet6
 #define raw6_sk(__sk) (&((struct raw6_sock *)__sk)->raw6)
+
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+#define __ipv6_only_sock(sk)	(inet6_sk(sk)->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

include/linux/sysctl.h

@@ -357,7 +357,8 @@ enum
 enum {
 	NET_IPV6_CONF=16,
 	NET_IPV6_NEIGH=17,
-	NET_IPV6_ROUTE=18
+	NET_IPV6_ROUTE=18,
+	NET_IPV6_BINDV6ONLY=20,
 };
 
 enum {

include/net/ipv6.h

@@ -102,6 +102,9 @@ struct frag_hdr {
 
 #include <net/sock.h>
 
+/* sysctls */
+extern int sysctl_ipv6_bindv6only;
+
 extern struct ipv6_mib		ipv6_statistics[NR_CPUS*2];
 #define IP6_INC_STATS(field)		SNMP_INC_STATS(ipv6_statistics, field)
 #define IP6_INC_STATS_BH(field)		SNMP_INC_STATS_BH(ipv6_statistics, field)

net/ipv4/tcp_ipv4.c

@@ -47,9 +47,13 @@
  *					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>
@@ -62,6 +66,7 @@
 #include <net/inet_common.h>
 
 #include <linux/inet.h>
+#include <linux/ipv6.h>
 #include <linux/stddef.h>
 #include <linux/ipsec.h>
 
@@ -176,7 +181,9 @@ static inline int tcp_bind_conflict(struct sock *sk, struct tcp_bind_bucket *tb)
 	int sk_reuse = sk->reuse;
 
 	for ( ; sk2; sk2 = sk2->bind_next) {
-		if (sk != sk2 && sk->bound_dev_if == sk2->bound_dev_if) {
+		if (sk != sk2 &&
+		    !ipv6_only_sock(sk2) &&
+		    sk->bound_dev_if == sk2->bound_dev_if) {
 			if (!sk_reuse || !sk2->reuse ||
 			    sk2->state == TCP_LISTEN) {
 				struct inet_opt *inet2 = inet_sk(sk2);
@@ -412,25 +419,25 @@ static struct sock *__tcp_v4_lookup_listener(struct sock *sk, u32 daddr,
 	struct sock *result = NULL;
 	int score, hiscore;
 
-	hiscore=0;
+	hiscore=-1;
 	for (; sk; sk = sk->next) {
 		struct inet_opt *inet = inet_sk(sk);
 
-		if (inet->num == hnum) {
+		if (inet->num == hnum && !ipv6_only_sock(sk)) {
 			__u32 rcv_saddr = inet->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;
@@ -454,6 +461,7 @@ __inline__ struct sock *tcp_v4_lookup_listener(u32 daddr, unsigned short hnum,
 
 		if (inet->num == hnum && !sk->next &&
 		    (!inet->rcv_saddr || inet->rcv_saddr == daddr) &&
+		    (sk->family == PF_INET || !ipv6_only_sock(sk)) &&
 		    !sk->bound_dev_if)
 			goto sherry_cache;
 		sk = __tcp_v4_lookup_listener(sk, daddr, hnum, dif);

net/ipv4/udp.c

@@ -66,6 +66,9 @@
  *					datagrams.
  *	Hirokazu Takahashi	:	sendfile() on UDP works now.
  *		Arnaldo C. Melo :	convert /proc/net/udp to seq_file
+ *	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
@@ -87,6 +90,7 @@
 #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/tcp.h>
@@ -170,6 +174,7 @@ static int udp_v4_get_port(struct sock *sk, unsigned short snum)
 
 			if (inet2->num == snum &&
 			    sk2 != sk &&
+			    !ipv6_only_sock(sk2) &&
 			    sk2->bound_dev_if == sk->bound_dev_if &&
 			    (!inet2->rcv_saddr ||
 			     !inet->rcv_saddr ||
@@ -228,29 +233,29 @@ struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport, u32 daddr, u16 dport, i
 	for(sk = udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]; sk != NULL; sk = sk->next) {
 		struct inet_opt *inet = inet_sk(sk);
 
-		if (inet->num == hnum) {
-			int score = 0;
+		if (inet->num == hnum && !ipv6_only_sock(sk)) {
+			int score = (sk->family == PF_INET ? 1 : 0);
 			if (inet->rcv_saddr) {
 				if (inet->rcv_saddr != daddr)
 					continue;
-				score++;
+				score+=2;
 			}
 			if (inet->daddr) {
 				if (inet->daddr != saddr)
 					continue;
-				score++;
+				score+=2;
 			}
 			if (inet->dport) {
 				if (inet->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) {
@@ -288,6 +293,7 @@ static inline struct sock *udp_v4_mcast_next(struct sock *sk,
 		    (inet->daddr && inet->daddr != rmt_addr)		||
 		    (inet->dport != rmt_port && inet->dport)		||
 		    (inet->rcv_saddr && inet->rcv_saddr != loc_addr)	||
+		    ipv6_only_sock(s)					||
 		    (s->bound_dev_if && s->bound_dev_if != dif))
 			continue;
 		break;

net/ipv6/af_inet6.c

@@ -88,6 +88,8 @@ extern void ipv6_sysctl_register(void);
 extern void ipv6_sysctl_unregister(void);
 #endif
 
+int sysctl_ipv6_bindv6only;
+
 #ifdef INET_REFCNT_DEBUG
 atomic_t inet6_sock_nr;
 #endif
@@ -218,7 +220,8 @@ static int inet6_create(struct socket *sock, int protocol)
 	np->mcast_hops	= -1;
 	np->mc_loop	= 1;
 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
-
+	np->ipv6only	= sysctl_ipv6_bindv6only;
+	
 	/* Init the ipv4 part of the socket since we can have sockets
 	 * using v6 API for ipv4.
 	 */

net/ipv6/ipv6_sockglue.c

@@ -157,7 +157,8 @@ int ipv6_setsockopt(struct sock *sk, int level, int optname, char *optval,
 				break;
 			}
 
-			if (!(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
+			if (ipv6_only_sock(sk) ||
+			    !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
 				retv = -EADDRNOTAVAIL;
 				break;
 			}
@@ -203,6 +204,13 @@ int ipv6_setsockopt(struct sock *sk, int level, int optname, char *optval,
 		}
 		goto e_inval;
 
+	case IPV6_V6ONLY:
+		if (inet_sk(sk)->num)
+			goto e_inval;
+		np->ipv6only = valbool;
+		retv = 0;
+		break;
+
 	case IPV6_PKTINFO:
 		np->rxopt.bits.rxinfo = valbool;
 		retv = 0;
@@ -475,6 +483,10 @@ int ipv6_getsockopt(struct sock *sk, int level, int optname, char *optval,
 		break;
 	}
 
+	case IPV6_V6ONLY:
+		val = np->ipv6only;
+		break;
+
 	case IPV6_PKTINFO:
 		val = np->rxopt.bits.rxinfo;
 		break;

net/ipv6/sysctl_net_ipv6.c

@@ -17,6 +17,8 @@ extern ctl_table ipv6_route_table[];
 
 ctl_table ipv6_table[] = {
 	{NET_IPV6_ROUTE, "route", NULL, 0, 0555, ipv6_route_table},
+	{NET_IPV6_BINDV6ONLY, "bindv6only",
+	 &sysctl_ipv6_bindv6only, sizeof(int), 0644, NULL, &proc_dointvec},
 	{0}
 };
 

net/ipv6/tcp_ipv6.c

@@ -14,6 +14,9 @@
  *
  *	Fixes:
  *	Hideaki YOSHIFUJI	:	sin6_scope_id support
+ *	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
  *      modify it under the terms of the GNU General Public License
@@ -90,14 +93,35 @@ static inline int ipv6_rcv_saddr_equal(struct sock *sk, struct sock *sk2)
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	int addr_type = ipv6_addr_type(&np->rcv_saddr);
 
-	return !inet_sk(sk2)->rcv_saddr || addr_type == IPV6_ADDR_ANY ||
-	       (sk2->family == AF_INET6 &&
-	        !ipv6_addr_cmp(&np->rcv_saddr,
-			       sk2->state != TCP_TIME_WAIT ?
-			       &inet6_sk(sk2)->rcv_saddr :
-			       &((struct tcp_tw_bucket *)sk)->v6_rcv_saddr)) ||
-	       (addr_type == IPV6_ADDR_MAPPED && sk2->family == AF_INET &&
-		inet_sk(sk)->rcv_saddr == inet_sk(sk2)->rcv_saddr);
+	if (!inet_sk(sk2)->rcv_saddr && !ipv6_only_sock(sk))
+		return 1;
+
+	if (sk2->family == AF_INET6 &&
+	    ipv6_addr_any(&inet6_sk(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(&inet6_sk(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 ?
+			    &inet6_sk(sk2)->rcv_saddr :
+			    &((struct tcp_tw_bucket *)sk)->v6_rcv_saddr)))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_MAPPED &&
+	    !ipv6_only_sock(sk2) &&
+	    (!inet_sk(sk2)->rcv_saddr ||
+	     !inet_sk(sk)->rcv_saddr ||
+	     inet_sk(sk)->rcv_saddr == inet_sk(sk2)->rcv_saddr))
+		return 1;
+
+	return 0;
 }
 
 static inline int tcp_v6_bind_conflict(struct sock *sk,
@@ -612,6 +636,9 @@ 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];

net/ipv6/udp.c

@@ -11,6 +11,9 @@
  *
  *	Fixes:
  *	Hideaki YOSHIFUJI	:	sin6_scope_id support
+ *	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
  *      modify it under the terms of the GNU General Public License
@@ -50,16 +53,41 @@
 
 struct udp_mib udp_stats_in6[NR_CPUS*2];
 
+/* XXX This is identical to tcp_ipv6.c:ipv6_rcv_saddr_equal, put
+ * XXX it somewhere common. -DaveM
+ */
 static __inline__ int udv6_rcv_saddr_equal(struct sock *sk, struct sock *sk2)
 {
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	int addr_type = ipv6_addr_type(&np->rcv_saddr);
 
-	return !inet_sk(sk2)->rcv_saddr || addr_type == IPV6_ADDR_ANY ||
-	       (sk2->family == AF_INET6 &&
-	        !ipv6_addr_cmp(&np->rcv_saddr, &inet6_sk(sk2)->rcv_saddr)) ||
-	       (addr_type == IPV6_ADDR_MAPPED && sk2->family == AF_INET &&
-		inet_sk(sk)->rcv_saddr == inet_sk(sk2)->rcv_saddr);
+	if (!inet_sk(sk2)->rcv_saddr && !ipv6_only_sock(sk))
+		return 1;
+
+	if (sk2->family == AF_INET6 && 
+	    ipv6_addr_any(&inet6_sk(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(&inet6_sk(sk2)->rcv_saddr) == IPV6_ADDR_MAPPED) : 1)))
+		return 1;
+
+	if (sk2->family == AF_INET6 && 
+	    !ipv6_addr_cmp(&inet6_sk(sk)->rcv_saddr,
+			   &inet6_sk(sk2)->rcv_saddr))
+		return 1;
+
+	if (addr_type == IPV6_ADDR_MAPPED &&
+	    !ipv6_only_sock(sk2) &&
+	    (!inet_sk(sk2)->rcv_saddr || 
+	     !inet_sk(sk)->rcv_saddr ||
+	     inet_sk(sk)->rcv_saddr == inet_sk(sk2)->rcv_saddr))
+		return 1;
+
+	return 0;
 }
 
 /* Grrr, addr_type already calculated by caller, but I don't want
@@ -228,6 +256,8 @@ 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;
 	}
@@ -263,6 +293,9 @@ 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;
@@ -793,8 +826,11 @@ static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg
 	fl.oif = 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(iocb, sk, msg, ulen);
+		}
 
 		if (addr_len < SIN6_LEN_RFC2133)
 			return -EINVAL;
@@ -841,6 +877,9 @@ static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg
 	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;