net: ipv4: Add ability to have GRE ignore DF bit in IPv4 payloads

    In the presence of firewalls which improperly block ICMP Unreachable
    (including Fragmentation Required) messages, Path MTU Discovery is
    prevented from working.

    A workaround is to handle IPv4 payloads opaquely, ignoring the DF bit--as
    is done for other payloads like AppleTalk--and doing transparent
    fragmentation and reassembly.

    Redux includes the enforcement of mutual exclusion between this feature
    and Path MTU Discovery as suggested by Alexander Duyck.

    Cc: Alexander Duyck <alexander.duyck@gmail.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/ip_tunnels.h

@@ -132,6 +132,7 @@ struct ip_tunnel {
 	int			ip_tnl_net_id;
 	struct gro_cells	gro_cells;
 	bool			collect_md;
+	bool			ignore_df;
 };
 
 #define TUNNEL_CSUM		__cpu_to_be16(0x01)

include/uapi/linux/if_tunnel.h

@@ -113,6 +113,7 @@ enum {
 	IFLA_GRE_ENCAP_SPORT,
 	IFLA_GRE_ENCAP_DPORT,
 	IFLA_GRE_COLLECT_METADATA,
+	IFLA_GRE_IGNORE_DF,
 	__IFLA_GRE_MAX,
 };
 

net/ipv4/ip_gre.c

@@ -841,17 +841,19 @@ static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
 	return ipgre_tunnel_validate(tb, data);
 }
 
-static void ipgre_netlink_parms(struct net_device *dev,
+static int ipgre_netlink_parms(struct net_device *dev,
 				struct nlattr *data[],
 				struct nlattr *tb[],
 				struct ip_tunnel_parm *parms)
 {
+	struct ip_tunnel *t = netdev_priv(dev);
+
 	memset(parms, 0, sizeof(*parms));
 
 	parms->iph.protocol = IPPROTO_GRE;
 
 	if (!data)
-		return;
+		return 0;
 
 	if (data[IFLA_GRE_LINK])
 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
@@ -880,16 +882,26 @@ static void ipgre_netlink_parms(struct net_device *dev,
 	if (data[IFLA_GRE_TOS])
 		parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
 
-	if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
+	if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
+		if (t->ignore_df)
+			return -EINVAL;
 		parms->iph.frag_off = htons(IP_DF);
+	}
 
 	if (data[IFLA_GRE_COLLECT_METADATA]) {
-		struct ip_tunnel *t = netdev_priv(dev);
-
 		t->collect_md = true;
 		if (dev->type == ARPHRD_IPGRE)
 			dev->type = ARPHRD_NONE;
 	}
+
+	if (data[IFLA_GRE_IGNORE_DF]) {
+		if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
+		  && (parms->iph.frag_off & htons(IP_DF)))
+			return -EINVAL;
+		t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
+	}
+
+	return 0;
 }
 
 /* This function returns true when ENCAP attributes are present in the nl msg */
@@ -960,16 +972,19 @@ static int ipgre_newlink(struct net *src_net, struct net_device *dev,
 {
 	struct ip_tunnel_parm p;
 	struct ip_tunnel_encap ipencap;
+	int err;
 
 	if (ipgre_netlink_encap_parms(data, &ipencap)) {
 		struct ip_tunnel *t = netdev_priv(dev);
-		int err = ip_tunnel_encap_setup(t, &ipencap);
+		err = ip_tunnel_encap_setup(t, &ipencap);
 
 		if (err < 0)
 			return err;
 	}
 
-	ipgre_netlink_parms(dev, data, tb, &p);
+	err = ipgre_netlink_parms(dev, data, tb, &p);
+	if (err < 0)
+		return err;
 	return ip_tunnel_newlink(dev, tb, &p);
 }
 
@@ -978,16 +993,19 @@ static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
 {
 	struct ip_tunnel_parm p;
 	struct ip_tunnel_encap ipencap;
+	int err;
 
 	if (ipgre_netlink_encap_parms(data, &ipencap)) {
 		struct ip_tunnel *t = netdev_priv(dev);
-		int err = ip_tunnel_encap_setup(t, &ipencap);
+		err = ip_tunnel_encap_setup(t, &ipencap);
 
 		if (err < 0)
 			return err;
 	}
 
-	ipgre_netlink_parms(dev, data, tb, &p);
+	err = ipgre_netlink_parms(dev, data, tb, &p);
+	if (err < 0)
+		return err;
 	return ip_tunnel_changelink(dev, tb, &p);
 }
 
@@ -1024,6 +1042,8 @@ static size_t ipgre_get_size(const struct net_device *dev)
 		nla_total_size(2) +
 		/* IFLA_GRE_COLLECT_METADATA */
 		nla_total_size(0) +
+		/* IFLA_GRE_IGNORE_DF */
+		nla_total_size(1) +
 		0;
 }
 
@@ -1057,6 +1077,9 @@ static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
 			t->encap.flags))
 		goto nla_put_failure;
 
+	if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
+		goto nla_put_failure;
+
 	if (t->collect_md) {
 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
 			goto nla_put_failure;
@@ -1084,6 +1107,7 @@ static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
 	[IFLA_GRE_ENCAP_SPORT]	= { .type = NLA_U16 },
 	[IFLA_GRE_ENCAP_DPORT]	= { .type = NLA_U16 },
 	[IFLA_GRE_COLLECT_METADATA]	= { .type = NLA_FLAG },
+	[IFLA_GRE_IGNORE_DF]	= { .type = NLA_U8 },
 };
 
 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {

net/ipv4/ip_tunnel.c

@@ -682,7 +682,7 @@ void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 	}
 
 	df = tnl_params->frag_off;
-	if (skb->protocol == htons(ETH_P_IP))
+	if (skb->protocol == htons(ETH_P_IP) && !tunnel->ignore_df)
 		df |= (inner_iph->frag_off&htons(IP_DF));
 
 	max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)