Add IPv6 support.

Submitted by: Jean-Luc Richier <Jean-Luc.Richier@imag.fr>

11 files changed, 123 insertions, 47 deletions

diff --git a/sys/modules/nfsclient/Makefile b/sys/modules/nfsclient/Makefile
index 55fcd88..0d2505a 100644
--- a/sys/modules/nfsclient/Makefile
+++ b/sys/modules/nfsclient/Makefile
@@ -6,8 +6,10 @@ SRCS=	vnode_if.h \
 	nfs_bio.c nfs_lock.c nfs_node.c nfs_socket.c nfs_subs.c nfs_nfsiod.c \
 	nfs_vfsops.c nfs_vnops.c nfs_common.c \
 	opt_inet.h opt_nfs.h opt_bootp.h
+SRCS+=	opt_inet6.h
 
 NFS_INET?=	1	# 0/1 - requires INET to be configured in kernel
+NFS_INET6?=	1	# 0/1 - requires INET6 to be configured in kernel
 
 opt_inet.h:
 	touch ${.TARGET}
@@ -15,4 +17,10 @@ opt_inet.h:
 	echo "#define INET 1" > ${.TARGET}
 .endif
 
+.if ${NFS_INET6} > 0
+opt_inet6.h:
+	echo "#define INET6 1" > ${.TARGET}
+.endif
+
 .include <bsd.kmod.mk>
+
diff --git a/sys/modules/nfsserver/Makefile b/sys/modules/nfsserver/Makefile
index 89d9664..e57ac6e 100644
--- a/sys/modules/nfsserver/Makefile
+++ b/sys/modules/nfsserver/Makefile
@@ -6,5 +6,12 @@ SRCS=	vnode_if.h \
 	nfs_serv.c nfs_srvsock.c nfs_srvcache.c nfs_srvsubs.c nfs_syscalls.c \
 	nfs_common.c \
 	opt_nfs.h
+SRCS+=	opt_inet6.h
+NFS_INET6?=	1	# 0/1 - requires INET6 to be configured in kernel
+
+.if ${NFS_INET6} > 0
+opt_inet6.h:
+	echo "#define INET6 1" > ${.TARGET}
+.endif
 
 .include <bsd.kmod.mk>
diff --git a/sys/nfsclient/nfs_lock.c b/sys/nfsclient/nfs_lock.c
index 0c9d6ac..dfd5ed0 100644
--- a/sys/nfsclient/nfs_lock.c
+++ b/sys/nfsclient/nfs_lock.c
@@ -130,13 +130,12 @@ nfs_dolock(struct vop_advlock_args *ap)
 	msg.lm_wait = ap->a_flags & F_WAIT;
 	msg.lm_getlk = ap->a_op == F_GETLK;
 	/*
-	 * XXX  -- I think this is wrong for anything other AF_INET.
-	 *
 	 * XXX: the lm_cred assignment below directly exports a ucred
 	 * structure to userland.  This is probably wrong, and should at
 	 * least be xucred.
 	 */
-	msg.lm_addr = *(VFSTONFS(vp->v_mount)->nm_nam);
+	bcopy(VFSTONFS(vp->v_mount)->nm_nam, &msg.lm_addr,
+		min(sizeof msg.lm_addr, VFSTONFS(vp->v_mount)->nm_nam->sa_len));
 	msg.lm_fh_len = NFS_ISV3(vp) ? VTONFS(vp)->n_fhsize : NFSX_V2FH;
 	bcopy(VTONFS(vp)->n_fhp, msg.lm_fh, msg.lm_fh_len);
 	msg.lm_nfsv3 = NFS_ISV3(vp);
diff --git a/sys/nfsclient/nfs_lock.h b/sys/nfsclient/nfs_lock.h
index 9de041d..af99bc5 100644
--- a/sys/nfsclient/nfs_lock.h
+++ b/sys/nfsclient/nfs_lock.h
@@ -58,7 +58,7 @@ struct lockd_msg_ident {
 	int		msg_seq;	/* Sequence number of message */
 };
 
-#define LOCKD_MSG_VERSION	1
+#define LOCKD_MSG_VERSION	2
 
 /*
  * The structure that the kernel hands us for each lock request.
@@ -69,7 +69,7 @@ typedef struct __lock_msg {
 	struct flock		lm_fl;             /* The lock request. */
 	int			lm_wait;           /* The F_WAIT flag. */
 	int			lm_getlk;		/* is this a F_GETLK request */
-	struct sockaddr 	lm_addr;		/* The address. */
+	struct sockaddr_storage lm_addr;		/* The address. */
 	int			lm_nfsv3;		/* If NFS version 3. */
 	size_t			lm_fh_len;		/* The file handle length. */
 	struct ucred		lm_cred;		/* user cred for lock req */
diff --git a/sys/nfsclient/nfs_socket.c b/sys/nfsclient/nfs_socket.c
index 5966f38..496cbe4 100644
--- a/sys/nfsclient/nfs_socket.c
+++ b/sys/nfsclient/nfs_socket.c
@@ -43,6 +43,8 @@ __FBSDID("$FreeBSD$");
  * Socket operations for use by nfs
  */
 
+#include "opt_inet6.h"
+
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>
@@ -157,7 +159,6 @@ nfs_connect(struct nfsmount *nmp, struct nfsreq *rep)
 	int s, error, rcvreserve, sndreserve;
 	int pktscale;
 	struct sockaddr *saddr;
-	struct sockaddr_in *sin;
 	struct thread *td = &thread0; /* only used for socreate and sobind */
 
 	nmp->nm_so = NULL;
@@ -172,42 +173,51 @@ nfs_connect(struct nfsmount *nmp, struct nfsreq *rep)
 	/*
 	 * Some servers require that the client port be a reserved port number.
 	 */
-	if (saddr->sa_family == AF_INET && (nmp->nm_flag & NFSMNT_RESVPORT)) {
+	if (nmp->nm_flag & NFSMNT_RESVPORT) {
 		struct sockopt sopt;
-		int ip;
-		struct sockaddr_in ssin;
+		int ip, ip2, len;
+		struct sockaddr_in6 ssin;
+		struct sockaddr *sa;
 
 		bzero(&sopt, sizeof sopt);
-		ip = IP_PORTRANGE_LOW;
+		switch(saddr->sa_family) {
+		case AF_INET:
+			sopt.sopt_level = IPPROTO_IP;
+			sopt.sopt_name = IP_PORTRANGE;
+			ip = IP_PORTRANGE_LOW;
+			ip2 = IP_PORTRANGE_DEFAULT;
+			len = sizeof (struct sockaddr_in);
+			break;
+#ifdef INET6
+		case AF_INET6:
+			sopt.sopt_level = IPPROTO_IPV6;
+			sopt.sopt_name = IPV6_PORTRANGE;
+			ip = IPV6_PORTRANGE_LOW;
+			ip2 = IPV6_PORTRANGE_DEFAULT;
+			len = sizeof (struct sockaddr_in6);
+			break;
+#endif
+		default:
+			goto noresvport;
+		}
+		sa = (struct sockaddr *)&ssin;
+		bzero(sa, len);
+		sa->sa_len = len;
+		sa->sa_family = saddr->sa_family;
 		sopt.sopt_dir = SOPT_SET;
-		sopt.sopt_level = IPPROTO_IP;
-		sopt.sopt_name = IP_PORTRANGE;
 		sopt.sopt_val = (void *)&ip;
 		sopt.sopt_valsize = sizeof(ip);
-		sopt.sopt_td = NULL;
 		error = sosetopt(so, &sopt);
 		if (error)
 			goto bad;
-		bzero(&ssin, sizeof ssin);
-		sin = &ssin;
-		sin->sin_len = sizeof (struct sockaddr_in);
-		sin->sin_family = AF_INET;
-		sin->sin_addr.s_addr = INADDR_ANY;
-		sin->sin_port = htons(0);
-		error = sobind(so, (struct sockaddr *)sin, td);
+		error = sobind(so, sa, td);
 		if (error)
 			goto bad;
-		bzero(&sopt, sizeof sopt);
-		ip = IP_PORTRANGE_DEFAULT;
-		sopt.sopt_dir = SOPT_SET;
-		sopt.sopt_level = IPPROTO_IP;
-		sopt.sopt_name = IP_PORTRANGE;
-		sopt.sopt_val = (void *)&ip;
-		sopt.sopt_valsize = sizeof(ip);
-		sopt.sopt_td = NULL;
+		ip = ip2;
 		error = sosetopt(so, &sopt);
 		if (error)
 			goto bad;
+	noresvport: ;
 	}
 
 	/*
diff --git a/sys/nfsserver/nfs.h b/sys/nfsserver/nfs.h
index 18768af..cbdb54f 100644
--- a/sys/nfsserver/nfs.h
+++ b/sys/nfsserver/nfs.h
@@ -195,13 +195,6 @@ union nethostaddr {
 	struct sockaddr *had_nam;
 };
 
-#define	nu_inetaddr	nu_haddr.had_inetaddr
-#define	nu_nam		nu_haddr.had_nam
-/* Bits for nu_flag */
-#define	NU_INETADDR	0x1
-#define NU_NAM		0x2
-#define NU_NETFAM(u)	(((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO)
-
 struct nfsrv_rec {
 	STAILQ_ENTRY(nfsrv_rec) nr_link;
 	struct sockaddr	*nr_address;
diff --git a/sys/nfsserver/nfs_srvcache.c b/sys/nfsserver/nfs_srvcache.c
index 2222ffa..1986c0b 100644
--- a/sys/nfsserver/nfs_srvcache.c
+++ b/sys/nfsserver/nfs_srvcache.c
@@ -71,7 +71,7 @@ static u_long nfsrvhash;
 #define	FALSE	0
 
 #define	NETFAMILY(rp) \
-		(((rp)->rc_flag & RC_INETADDR) ? AF_INET : AF_ISO)
+		(((rp)->rc_flag & RC_NAM) ? (rp)->rc_nam->sa_family : AF_INET)
 
 /*
  * Static array that defines which nfs rpc's are nonidempotent
@@ -240,7 +240,8 @@ loop:
 		rp->rc_flag |= RC_INETADDR;
 		rp->rc_inetaddr = saddr->sin_addr.s_addr;
 		break;
-	case AF_ISO:
+/*	case AF_INET6:	*/
+/*	case AF_ISO:	*/
 	default:
 		rp->rc_flag |= RC_NAM;
 		rp->rc_nam = dup_sockaddr(nd->nd_nam, 1);
diff --git a/sys/nfsserver/nfs_srvsock.c b/sys/nfsserver/nfs_srvsock.c
index 2143544..1feffa9 100644
--- a/sys/nfsserver/nfs_srvsock.c
+++ b/sys/nfsserver/nfs_srvsock.c
@@ -154,13 +154,13 @@ nfs_rephead(int siz, struct nfsrv_descript *nd, int err,
 	 * If this is a big reply, use a cluster else
 	 * try and leave leading space for the lower level headers.
 	 */
+	mreq->m_len = 6 * NFSX_UNSIGNED;
 	siz += RPC_REPLYSIZ;
 	if ((max_hdr + siz) >= MINCLSIZE) {
 		MCLGET(mreq, M_TRYWAIT);
 	} else
-		mreq->m_data += max_hdr;
+		mreq->m_data += min(max_hdr, M_TRAILINGSPACE(mreq));
 	tl = mtod(mreq, u_int32_t *);
-	mreq->m_len = 6 * NFSX_UNSIGNED;
 	bpos = ((caddr_t)tl) + mreq->m_len;
 	*tl++ = txdr_unsigned(nd->nd_retxid);
 	*tl++ = nfsrv_rpc_reply;
diff --git a/sys/nfsserver/nfs_srvsubs.c b/sys/nfsserver/nfs_srvsubs.c
index 5f37b1f..19a74d7 100644
--- a/sys/nfsserver/nfs_srvsubs.c
+++ b/sys/nfsserver/nfs_srvsubs.c
@@ -46,6 +46,8 @@ __FBSDID("$FreeBSD$");
  * copy data between mbuf chains and uio lists.
  */
 
+#include "opt_inet6.h"
+
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>
@@ -1051,7 +1053,9 @@ nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
 #ifdef MNT_EXNORESPORT
 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
 		saddr = (struct sockaddr_in *)nam;
-		if (saddr->sin_family == AF_INET &&
+		if ((saddr->sin_family == AF_INET ||
+		     saddr->sin_family == AF_INET6) &&
+	/* same code for INET and INET6: sin*_port at same offet */
 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
 			vput(*vpp);
 			*vpp = NULL;
@@ -1117,6 +1121,21 @@ netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
 			return (1);
 		break;
+#ifdef INET6
+	case AF_INET6:
+	{
+		register struct sockaddr_in6 *inet6addr1, *inet6addr2;
+
+		inet6addr1 = (struct sockaddr_in6 *)nam;
+		inet6addr2 = (struct sockaddr_in6 *)haddr->had_nam;
+	/* XXX - should test sin6_scope_id ? */
+		if (inet6addr1->sin6_family == AF_INET6 &&
+		    IN6_ARE_ADDR_EQUAL(&inet6addr1->sin6_addr,
+				       &inet6addr2->sin6_addr))
+			return (1);
+		break;
+	}
+#endif
 	default:
 		break;
 	};
diff --git a/sys/nfsserver/nfs_syscalls.c b/sys/nfsserver/nfs_syscalls.c
index fb60d36..f123dcc 100644
--- a/sys/nfsserver/nfs_syscalls.c
+++ b/sys/nfsserver/nfs_syscalls.c
@@ -40,6 +40,8 @@
 #include <sys/cdefs.h>
 __FBSDID("$FreeBSD$");
 
+#include "opt_inet6.h"
+
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/sysproto.h>
@@ -64,6 +66,10 @@ __FBSDID("$FreeBSD$");
 
 #include <netinet/in.h>
 #include <netinet/tcp.h>
+#ifdef INET6
+#include <net/if.h>
+#include <netinet6/in6_var.h>
+#endif
 #include <nfs/xdr_subs.h>
 #include <nfs/rpcv2.h>
 #include <nfs/nfsproto.h>
@@ -109,6 +115,10 @@ static int	nfssvc_nfsd(struct nfsd_srvargs *, caddr_t, struct thread *);
  * - adds a socket to the selection list
  * - remains in the kernel as an nfsd
  * - remains in the kernel as an nfsiod
+ * For INET6 we suppose that nfsd provides only IN6P_BINDV6ONLY sockets
+ * and that mountd provides
+ *  - sockaddr with no IPv4-mapped addresses
+ *  - mask for both INET and INET6 families if there is IPv4-mapped overlap
  */
 #ifndef _SYS_SYSPROTO_H_
 struct nfssvc_args {
@@ -230,8 +240,7 @@ nfssvc_addsock(struct file *fp, struct sockaddr *mynam, struct thread *td)
 		val = 1;
 		sosetopt(so, &sopt);
 	}
-	if (so->so_proto->pr_domain->dom_family == AF_INET &&
-	    so->so_proto->pr_protocol == IPPROTO_TCP) {
+	if (so->so_proto->pr_protocol == IPPROTO_TCP) {
 		struct sockopt sopt;
 		int val;
 
@@ -385,13 +394,33 @@ nfssvc_nfsd(struct nfsd_srvargs *nsd, caddr_t argp, struct thread *td)
 			struct sockaddr_in *sin;
 
 			sin = (struct sockaddr_in *)nam;
+			/*
+			 * INET/INET6 - same code:
+			 *    sin_port and sin6_port are at same offset
+			 */
 			port = ntohs(sin->sin_port);
 			if (port >= IPPORT_RESERVED &&
 			    nd->nd_procnum != NFSPROC_NULL) {
+#if defined(INET6) && defined(KLD_MODULE)
+	/* do not use ip6_sprintf: the nfs module should work without INET6 */
+	char b6[INET6_ADDRSTRLEN];
+#define ip6_sprintf(a) \
+	 (sprintf(b6, "%x:%x:%x:%x:%x:%x:%x:%x", \
+		  (a)->s6_addr16[0], (a)->s6_addr16[1], \
+		  (a)->s6_addr16[2], (a)->s6_addr16[3], \
+		  (a)->s6_addr16[4], (a)->s6_addr16[5], \
+		  (a)->s6_addr16[6], (a)->s6_addr16[7]), \
+	  b6)
+#endif
 			    nd->nd_procnum = NFSPROC_NOOP;
 			    nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
 			    cacherep = RC_DOIT;
 			    printf("NFS request from unprivileged port (%s:%d)\n",
+#ifdef INET6
+				   sin->sin_family == AF_INET6 ?
+					ip6_sprintf(&satosin6(sin)->sin6_addr) :
+#undef ip6_sprintf
+#endif
 				   inet_ntoa(sin->sin_addr), port);
 			}
 		    }
diff --git a/usr.sbin/rpc.lockd/kern.c b/usr.sbin/rpc.lockd/kern.c
index e79cfb0..5eb9413 100644
--- a/usr.sbin/rpc.lockd/kern.c
+++ b/usr.sbin/rpc.lockd/kern.c
@@ -47,6 +47,7 @@
 #include <string.h>
 #include <syslog.h>
 #include <unistd.h>
+#include <netdb.h>
 
 #include "nlm_prot.h"
 #include <nfs/rpcv2.h>
@@ -87,8 +88,17 @@ int	unlock_request(LOCKD_MSG *);
 #define d_calls (debug_level > 1)
 #define d_args (debug_level > 2)
 
-#define from_addr(sockaddr) \
-	(inet_ntoa((sockaddr)->sin_addr))
+static const char *
+from_addr(saddr)
+	struct sockaddr *saddr;
+{
+	static char inet_buf[INET6_ADDRSTRLEN];
+
+	if (getnameinfo(saddr, saddr->sa_len, inet_buf, sizeof(inet_buf),
+			NULL, 0, NI_NUMERICHOST) == 0)
+		return inet_buf;
+	return "???";
+}
 
 void
 client_cleanup(void)
@@ -257,7 +267,7 @@ test_request(LOCKD_MSG *msg)
 		syslog(LOG_DEBUG, "test request: %s: %s to %s",
 		    msg->lm_nfsv3 ? "V4" : "V1/3",
 		    msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
-		    from_addr((struct sockaddr_in *)&msg->lm_addr));
+		    from_addr((struct sockaddr *)&msg->lm_addr));
 
 	if (msg->lm_nfsv3) {
 		struct nlm4_testargs arg4;
@@ -326,7 +336,7 @@ lock_request(LOCKD_MSG *msg)
 		syslog(LOG_DEBUG, "lock request: %s: %s to %s",
 		    msg->lm_nfsv3 ? "V4" : "V1/3",
 		    msg->lm_fl.l_type == F_WRLCK ? "write" : "read",
-		    from_addr((struct sockaddr_in *)&msg->lm_addr));
+		    from_addr((struct sockaddr *)&msg->lm_addr));
 
 	if (msg->lm_nfsv3) {
 		arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;
@@ -396,7 +406,7 @@ unlock_request(LOCKD_MSG *msg)
 	if (d_calls)
 		syslog(LOG_DEBUG, "unlock request: %s: to %s",
 		    msg->lm_nfsv3 ? "V4" : "V1/3",
-		    from_addr((struct sockaddr_in *)&msg->lm_addr));
+		    from_addr((struct sockaddr *)&msg->lm_addr));
 
 	if (msg->lm_nfsv3) {
 		arg4.cookie.n_bytes = (char *)&msg->lm_msg_ident;