icmp6: Improve validation of PMTU

Currently we accept any pmtu between IPV6_MMTU(1280B) and the link mtu.
In some network topologies could allow a bad actor to perform a DOS attack.
Contrary to IPv4 in IPv6 oversized packets are dropped, and a ICMP
PACKET_TOO_BIG message is sent back to the sender.
After receiving an ICMPv6 packet with pmtu bigger than the
current one the victim will start sending frames that will be dropped
a router with reduced MTU.
Although it will eventually receive another message with correct pmtu,
an attacker can still just inject their spoofed packets frequently
enough to overwrite the correct value.
This issue is described in detail in RFC8201, section 6.
Fix this by checking the current pmtu, and accepting the new one only
if it's smaller.

Approved by:	mw(mentor)
Reviewed by:	tuexen
MFC after:	1 week
Sponsored by:	Stormshield
Obtained from:	Semihalf
Differential Revision: https://reviews.freebsd.org/D35871

1 files changed, 6 insertions, 1 deletions

diff --git a/sys/netinet6/icmp6.c b/sys/netinet6/icmp6.c
index 3632f9396be0..832061ad9328 100644
--- a/sys/netinet6/icmp6.c
+++ b/sys/netinet6/icmp6.c
@@ -1111,6 +1111,7 @@ icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
 	u_int mtu = ntohl(icmp6->icmp6_mtu);
 	struct in_conninfo inc;
+	uint32_t max_mtu;
 
 #if 0
 	/*
@@ -1151,7 +1152,11 @@ icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
 		return;
 
-	if (mtu < tcp_maxmtu6(&inc, NULL)) {
+	max_mtu = tcp_hc_getmtu(&inc);
+	if (max_mtu == 0)
+		max_mtu = tcp_maxmtu6(&inc, NULL);
+
+	if (mtu < max_mtu) {
 		tcp_hc_updatemtu(&inc, mtu);
 		ICMP6STAT_INC(icp6s_pmtuchg);
 	}