6 files changed, 38 insertions, 2 deletions

diff --git a/share/man/man4/ip.4 b/share/man/man4/ip.4
index 4e9b1fc189bb..d5da6e1e6f65 100644
--- a/share/man/man4/ip.4
+++ b/share/man/man4/ip.4
@@ -32,7 +32,7 @@
 .\"     @(#)ip.4	8.2 (Berkeley) 11/30/93
 .\" $FreeBSD$
 .\"
-.Dd August 22, 2005
+.Dd September 26, 2005
 .Dt IP 4
 .Os
 .Sh NAME
@@ -121,6 +121,20 @@ This option is only really useful when set to 255 preventing packets
 from outside the directly connected networks reaching local listeners
 on sockets.
 .Pp
+.Dv IP_DONTFRAG
+may be used to set the Don't Fragment flag on IP packets.
+Currently this option is respected only on
+.Xr udp 4
+and Raw
+.Xr ip 4
+sockets, unless the IP_HDRINCL option has been set.
+On
+.Xr tcp 4
+sockets the Don't Fragment flag is controlled by the Path
+MTU Discovery option.
+Sending a packet larger than the MTU size of the egress interface,
+determined by the destination address, returns an EMSGSIZE error.
+.Pp
 If the
 .Dv IP_RECVDSTADDR
 option is enabled on a
diff --git a/sys/netinet/in.h b/sys/netinet/in.h
index eede9115884e..e06f200f3ba9 100644
--- a/sys/netinet/in.h
+++ b/sys/netinet/in.h
@@ -416,6 +416,7 @@ __END_DECLS
 
 #define	IP_RECVTTL		65   /* bool; receive IP TTL w/dgram */
 #define	IP_MINTTL		66   /* minimum TTL for packet or drop */
+#define	IP_DONTFRAG		67   /* don't fragment packet */
 
 /*
  * Defaults and limits for options
diff --git a/sys/netinet/in_pcb.h b/sys/netinet/in_pcb.h
index 3d267ae9edcd..fc4dfe40169e 100644
--- a/sys/netinet/in_pcb.h
+++ b/sys/netinet/in_pcb.h
@@ -283,6 +283,7 @@ struct inpcbinfo {		/* XXX documentation, prefixes */
 #define	INP_MTUDISC		0x100	/* user can do MTU discovery */
 #define	INP_FAITH		0x200	/* accept FAITH'ed connections */
 #define	INP_RECVTTL		0x400	/* receive incoming IP TTL */
+#define	INP_DONTFRAG		0x800	/* don't fragment packet */
 
 #define IN6P_IPV6_V6ONLY	0x008000 /* restrict AF_INET6 socket for v6 */
 
diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
index b292d803eb41..0ef9a716dcd2 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -1197,6 +1197,7 @@ ip_ctloutput(so, sopt)
 		case IP_RECVIF:
 		case IP_FAITH:
 		case IP_ONESBCAST:
+		case IP_DONTFRAG:
 			error = sooptcopyin(sopt, &optval, sizeof optval,
 					    sizeof optval);
 			if (error)
@@ -1254,6 +1255,9 @@ ip_ctloutput(so, sopt)
 			case IP_ONESBCAST:
 				OPTSET(INP_ONESBCAST);
 				break;
+			case IP_DONTFRAG:
+				OPTSET(INP_DONTFRAG);
+				break;
 			}
 			break;
 #undef OPTSET
@@ -1351,6 +1355,7 @@ ip_ctloutput(so, sopt)
 		case IP_PORTRANGE:
 		case IP_FAITH:
 		case IP_ONESBCAST:
+		case IP_DONTFRAG:
 			switch (sopt->sopt_name) {
 
 			case IP_TOS:
@@ -1403,6 +1408,9 @@ ip_ctloutput(so, sopt)
 			case IP_ONESBCAST:
 				optval = OPTBIT(INP_ONESBCAST);
 				break;
+			case IP_DONTFRAG:
+				optval = OPTBIT(INP_DONTFRAG);
+				break;
 			}
 			error = sooptcopyout(sopt, &optval, sizeof optval);
 			break;
diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c
index 8a128eb515d0..bc250a40f0ad 100644
--- a/sys/netinet/raw_ip.c
+++ b/sys/netinet/raw_ip.c
@@ -271,7 +271,10 @@ rip_output(struct mbuf *m, struct socket *so, u_long dst)
 		INP_LOCK(inp);
 		ip = mtod(m, struct ip *);
 		ip->ip_tos = inp->inp_ip_tos;
-		ip->ip_off = 0;
+		if (inp->inp_flags & INP_DONTFRAG)
+			ip->ip_off = IP_DF;
+		else
+			ip->ip_off = 0;
 		ip->ip_p = inp->inp_ip_p;
 		ip->ip_len = m->m_pkthdr.len;
 		if (jailed(inp->inp_socket->so_cred))
diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c
index 6de2d7b77474..1f53137b8ba8 100644
--- a/sys/netinet/udp_usrreq.c
+++ b/sys/netinet/udp_usrreq.c
@@ -846,6 +846,15 @@ udp_output(inp, m, addr, control, td)
 	ui->ui_dport = fport;
 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
 
+	/*
+	 * Set the Don't Fragment bit in the IP header.
+	 */
+	if (inp->inp_flags & INP_DONTFRAG) {
+		struct ip *ip;
+		ip = (struct ip *)&ui->ui_i;
+		ip->ip_off |= IP_DF;
+	}
+
 	ipflags = 0;
 	if (inp->inp_socket->so_options & SO_DONTROUTE)
 		ipflags |= IP_ROUTETOIF;