Implement IP_DONTFRAG IP socket option enabling the Don't Fragment

flag on IP packets.  Currently this option is only repected on udp
and raw ip sockets.  On tcp sockets the DF flag is controlled by the
path MTU discovery option.

Sending a packet larger than the MTU size of the egress interface
returns an EMSGSIZE error.

Discussed with:	rwatson
Sponsored by:	TCP/IP Optimization Fundraise 2005

5 files changed, 24 insertions, 2 deletions

diff --git a/sys/netinet/in.h b/sys/netinet/in.h
index eede911..e06f200 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 3d267ae..fc4dfe4 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 900cd1f..eac1955 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -142,7 +142,7 @@ ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro,
 #endif /* FAST_IPSEC */
 
 	M_ASSERTPKTHDR(m);
-	
+
 	if (ro == NULL) {
 		ro = &iproute;
 		bzero(ro, sizeof (*ro));
@@ -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 8a128eb..bc250a4 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 6de2d7b..1f53137 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;