Experimental support for opportunitic use of jumbograms where only some hosts

on the local network support them.

This adds a new socket option, SO_JUMBO, and a new route flag,
RTF_JUMBO. If _both_ the socket option is set and the route for the host
has RTF_JUMBO set, ip_output will fragment the packet to the largest
possible size for the link, ignoring the card's MTU.

The semantics of this feature will be evolving rapidly; talk to us
if you intend to use it.

ok deraadt@ marius@

9 files changed, 34 insertions, 13 deletions

diff --git a/sys/kern/uipc_socket.c b/sys/kern/uipc_socket.c
index cddf0d3452f..14fa544fd84 100644
--- a/sys/kern/uipc_socket.c
+++ b/sys/kern/uipc_socket.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: uipc_socket.c,v 1.56 2004/11/18 15:09:07 markus Exp $	*/
+/*	$OpenBSD: uipc_socket.c,v 1.57 2005/05/27 04:55:27 mcbride Exp $	*/
 /*	$NetBSD: uipc_socket.c,v 1.21 1996/02/04 02:17:52 christos Exp $	*/
 
 /*
@@ -1026,6 +1026,7 @@ sosetopt(so, level, optname, m0)
 		case SO_REUSEADDR:
 		case SO_REUSEPORT:
 		case SO_OOBINLINE:
+		case SO_JUMBO:
 			if (m == NULL || m->m_len < sizeof (int)) {
 				error = EINVAL;
 				goto bad;
@@ -1164,6 +1165,7 @@ sogetopt(so, level, optname, mp)
 		case SO_REUSEPORT:
 		case SO_BROADCAST:
 		case SO_OOBINLINE:
+		case SO_JUMBO:
 			*mtod(m, int *) = so->so_options & optname;
 			break;
 
diff --git a/sys/net/route.h b/sys/net/route.h
index 8f4260dd23b..db6058e186d 100644
--- a/sys/net/route.h
+++ b/sys/net/route.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: route.h,v 1.27 2005/05/26 22:37:34 henning Exp $	*/
+/*	$OpenBSD: route.h,v 1.28 2005/05/27 04:55:27 mcbride Exp $	*/
 /*	$NetBSD: route.h,v 1.9 1996/02/13 22:00:49 christos Exp $	*/
 
 /*
@@ -139,6 +139,7 @@ struct rtentry {
 #define RTF_CLONED	0x10000		/* this is a cloned route */
 #define RTF_SOURCE	0x20000		/* this route has a source selector */
 #define RTF_MPATH	0x40000		/* multipath route or operation */
+#define RTF_JUMBO	0x80000		/* try to use jumbo frames */
 
 #ifndef _KERNEL
 /* obsoleted */
diff --git a/sys/net/rtsock.c b/sys/net/rtsock.c
index f19aabbf2cd..2085dd99568 100644
--- a/sys/net/rtsock.c
+++ b/sys/net/rtsock.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: rtsock.c,v 1.43 2004/09/16 22:31:29 henning Exp $	*/
+/*	$OpenBSD: rtsock.c,v 1.44 2005/05/27 04:55:27 mcbride Exp $	*/
 /*	$NetBSD: rtsock.c,v 1.18 1996/03/29 00:32:10 cgd Exp $	*/
 
 /*
@@ -396,6 +396,12 @@ report:
 				    rt->rt_ifp = ifp;
 				}
 			}
+
+			/* XXX Hack to allow the jumbo flag to be toggled */
+			if (rtm->rtm_flags & RTF_JUMBO)
+				rt->rt_flags = (rt->rt_flags & ~rtm->rtm_use) |
+				    (rtm->rtm_flags & rtm->rtm_use);
+
 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
 			    &rt->rt_rmx);
 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
diff --git a/sys/netinet/in.h b/sys/netinet/in.h
index a0d14d5878f..84985ed987a 100644
--- a/sys/netinet/in.h
+++ b/sys/netinet/in.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: in.h,v 1.65 2005/05/24 04:20:25 markus Exp $	*/
+/*	$OpenBSD: in.h,v 1.66 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: in.h,v 1.20 1996/02/13 23:41:47 christos Exp $	*/
 
 /*
@@ -320,6 +320,11 @@ struct ip_mreq {
 #define INET_ADDRSTRLEN		16
 
 /*
+ * JUMBO MTU
+ */
+#define IP_JUMBO_MTU	9000
+
+/*
  * Definitions for inet sysctl operations.
  *
  * Third level is protocol number.
diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c
index cdb0fa48949..724d481f022 100644
--- a/sys/netinet/ip_output.c
+++ b/sys/netinet/ip_output.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ip_output.c,v 1.170 2005/04/25 17:55:52 brad Exp $	*/
+/*	$OpenBSD: ip_output.c,v 1.171 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: ip_output.c,v 1.28 1996/02/13 23:43:07 christos Exp $	*/
 
 /*
@@ -693,6 +693,10 @@ sendit:
 	}
 #endif
 
+	/* Try to use jumbograms? */
+	if (flags & IP_JUMBO && ro->ro_rt && ro->ro_rt->rt_flags & RTF_JUMBO)
+		mtu = IP_JUMBO_MTU;
+
 	/*
 	 * If small enough for interface, can just send directly.
 	 */
diff --git a/sys/netinet/ip_var.h b/sys/netinet/ip_var.h
index 4697fe53824..33b34eb54f8 100644
--- a/sys/netinet/ip_var.h
+++ b/sys/netinet/ip_var.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ip_var.h,v 1.32 2004/06/22 07:35:20 cedric Exp $	*/
+/*	$OpenBSD: ip_var.h,v 1.33 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: ip_var.h,v 1.16 1996/02/13 23:43:20 christos Exp $	*/
 
 /*
@@ -152,8 +152,9 @@ struct	ipstat {
 #define	IP_RAWOUTPUT		0x2		/* raw ip header exists */
 #define	IP_ROUTETOIF		SO_DONTROUTE	/* bypass routing tables */
 #define	IP_ALLOWBROADCAST	SO_BROADCAST	/* can send broadcast packets */
-#define	IP_MTUDISC		0x0400		/* pmtu discovery, set DF */
-#define IP_ROUTETOETHER		0x0800		/* ether addresses given */
+#define IP_JUMBO		SO_JUMBO	/* try to use the jumbo mtu */
+#define	IP_MTUDISC		0x0800		/* pmtu discovery, set DF */
+#define IP_ROUTETOETHER		0x1000		/* ether addresses given */
 
 extern struct ipstat ipstat;
 extern LIST_HEAD(ipqhead, ipq)	ipq;	/* ip reass. queue */
diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c
index 1aa90898f3b..5791bc7f7ec 100644
--- a/sys/netinet/raw_ip.c
+++ b/sys/netinet/raw_ip.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: raw_ip.c,v 1.36 2005/01/14 14:51:28 mcbride Exp $	*/
+/*	$OpenBSD: raw_ip.c,v 1.37 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: raw_ip.c,v 1.25 1996/02/18 18:58:33 christos Exp $	*/
 
 /*
@@ -187,7 +187,7 @@ rip_output(struct mbuf *m, ...)
 	va_end(ap);
 
 	inp = sotoinpcb(so);
-	flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
+	flags = (so->so_options & (SO_DONTROUTE|SO_JUMBO)) | IP_ALLOWBROADCAST;
 
 	/*
 	 * If the user handed us a complete IP packet, use it.
diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c
index 3c449120b65..587ae62f5c3 100644
--- a/sys/netinet/udp_usrreq.c
+++ b/sys/netinet/udp_usrreq.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: udp_usrreq.c,v 1.103 2005/04/25 17:55:52 brad Exp $	*/
+/*	$OpenBSD: udp_usrreq.c,v 1.104 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: udp_usrreq.c,v 1.28 1996/03/16 23:54:03 christos Exp $	*/
 
 /*
@@ -1002,7 +1002,8 @@ udp_output(struct mbuf *m, ...)
 
 	udpstat.udps_opackets++;
 	error = ip_output(m, inp->inp_options, &inp->inp_route,
-	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
+	    inp->inp_socket->so_options &
+	    (SO_DONTROUTE | SO_BROADCAST | SO_JUMBO),
 	    inp->inp_moptions, inp, (void *)NULL);
 
 bail:
diff --git a/sys/sys/socket.h b/sys/sys/socket.h
index a429a775b41..2ef396b5947 100644
--- a/sys/sys/socket.h
+++ b/sys/sys/socket.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: socket.h,v 1.51 2005/04/04 22:18:47 hshoexer Exp $	*/
+/*	$OpenBSD: socket.h,v 1.52 2005/05/27 04:55:28 mcbride Exp $	*/
 /*	$NetBSD: socket.h,v 1.14 1996/02/09 18:25:36 christos Exp $	*/
 
 /*
@@ -66,6 +66,7 @@
 #define	SO_LINGER	0x0080		/* linger on close if data present */
 #define	SO_OOBINLINE	0x0100		/* leave received OOB data in line */
 #define	SO_REUSEPORT	0x0200		/* allow local address & port reuse */
+#define SO_JUMBO	0x0400		/* try to use jumbograms */
 
 /*
  * Additional options, not kept in so_options.