An earlier commit effectively turned out the fast forwading path

due to its lack of support for ICMP redirects. The following commit
adds redirects to the fastforward path, again allowing for decent
forwarding performance in the kernel.

Reviewed by: ae, melifaro
Sponsored by: Rubicon Communications, LLC (d/b/a "Netgate")

sys/netinet/ip_fastfwd.c

@@ -111,6 +111,43 @@ __FBSDID("$FreeBSD$");
 
 #include <machine/in_cksum.h>
 
+#define	V_ipsendredirects	VNET(ipsendredirects)
+
+static struct mbuf *
+ip_redir_alloc(struct mbuf *m, struct nhop_object *nh,
+    struct ip *ip, in_addr_t *addr)
+{
+	struct mbuf *mcopy = m_gethdr(M_NOWAIT, m->m_type);
+	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
+		/*
+		 * It's probably ok if the pkthdr dup fails (because
+		 * the deep copy of the tag chain failed), but for now
+		 * be conservative and just discard the copy since
+		 * code below may some day want the tags.
+		 */
+		m_free(mcopy);
+		return (NULL);
+	} 
+	mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
+	mcopy->m_pkthdr.len = mcopy->m_len;
+	m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
+
+	if (nh != NULL &&
+	    ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
+		struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
+		u_long src = ntohl(ip->ip_src.s_addr);
+
+		if (nh_ia != NULL && (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
+			if (nh->nh_flags & NHF_GATEWAY)
+				*addr = nh->gw4_sa.sin_addr.s_addr;
+			else
+				*addr = ip->ip_dst.s_addr;
+		}
+	}
+	return (mcopy);
+}
+
+
 static int
 ip_findroute(struct nhop_object **pnh, struct in_addr dest, struct mbuf *m)
 {
@@ -156,13 +193,14 @@ ip_tryforward(struct mbuf *m)
 {
 	struct ip *ip;
 	struct mbuf *m0 = NULL;
-	struct nhop_object *nh;
+	struct nhop_object *nh = NULL;
 	struct sockaddr_in dst;
 	struct in_addr dest, odest, rtdest;
 	uint16_t ip_len, ip_off;
 	int error = 0;
 	struct m_tag *fwd_tag = NULL;
-
+	struct mbuf *mcopy = NULL;
+	struct in_addr redest;
 	/*
 	 * Are we active and forwarding packets?
 	 */
@@ -386,6 +424,13 @@ ip_tryforward(struct mbuf *m)
 	else
 		dst.sin_addr = dest;
 
+	/*
+	 * Handle redirect case.
+	 */
+	redest.s_addr = 0;
+	if (V_ipsendredirects && (nh->nh_ifp == m->m_pkthdr.rcvif))
+		mcopy = ip_redir_alloc(m, nh, ip, &redest.s_addr);
+
 	/*
 	 * Check if packet fits MTU or if hardware will fragment for us
 	 */
@@ -455,7 +500,16 @@ ip_tryforward(struct mbuf *m)
 		IPSTAT_INC(ips_forward);
 		IPSTAT_INC(ips_fastforward);
 	}
+
+	/* Send required redirect */
+	if (mcopy != NULL) {
+		icmp_error(mcopy, ICMP_REDIRECT, ICMP_REDIRECT_HOST, redest.s_addr, 0);
+		mcopy = NULL; /* Freed by caller */
+	}
+
 consumed:
+	if (mcopy != NULL)
+		m_freem(mcopy);
 	return NULL;
 drop:
 	if (m)

sys/netinet/ip_input.c

@@ -111,8 +111,11 @@ SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW
     &VNET_NAME(ipforwarding), 0,
     "Enable IP forwarding between interfaces");
 
-VNET_DEFINE_STATIC(int, ipsendredirects) = 1;	/* XXX */
-#define	V_ipsendredirects	VNET(ipsendredirects)
+/* 
+ * Respond with an ICMP host redirect when we forward a packet out of
+ * the same interface on which it was received.  See RFC 792.
+ */
+VNET_DEFINE(int, ipsendredirects) = 1;
 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
     &VNET_NAME(ipsendredirects), 0,
     "Enable sending IP redirects");
@@ -571,7 +574,7 @@ ip_input(struct mbuf *m)
 	 * case skip another inbound firewall processing and update
 	 * ip pointer.
 	 */
-	if (V_ipforwarding != 0 && V_ipsendredirects == 0
+	if (V_ipforwarding != 0
 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 	    && (!IPSEC_ENABLED(ipv4) ||
 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)

sys/netinet/ip_var.h

@@ -182,6 +182,7 @@ struct inpcbinfo;
 
 VNET_DECLARE(int, ip_defttl);			/* default IP ttl */
 VNET_DECLARE(int, ipforwarding);		/* ip forwarding */
+VNET_DECLARE(int, ipsendredirects);
 #ifdef IPSTEALTH
 VNET_DECLARE(int, ipstealth);			/* stealth forwarding */
 #endif
@@ -197,6 +198,7 @@ extern struct	pr_usrreqs rip_usrreqs;
 #define	V_ip_id			VNET(ip_id)
 #define	V_ip_defttl		VNET(ip_defttl)
 #define	V_ipforwarding		VNET(ipforwarding)
+#define	V_ipsendredirects	VNET(ipsendredirects)
 #ifdef IPSTEALTH
 #define	V_ipstealth		VNET(ipstealth)
 #endif