Clamp MSS to the MTU of the source only.

Currently, tinc clamps MSS to the minimum of the MTU of both the source
and the destination of the packet containing the MSS option. This is
slightly suboptimal, because both the MTU and the MSS can differ based
on the direction of travel.

For example, if the MTU from A to B is 1500, and the MTU from B to A is
1000, then tinc will clamp MSS in *both* directions to 960, despite
the fact that A to B could use a 1460 MSS.

RFC 793 defines the MSS option as follows:

  If this option is present, then it communicates the maximum
  receive segment size at the TCP which sends this segment.

It follows that the MSS should be clamped to the MTU *to* the *sender*
of the MSS segment. It should not be clamped to the MTU to the
destination of the MSS segment.

src/route.c

@@ -109,10 +109,6 @@ static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *pac
 	if(!source || !via || !(via->options & OPTION_CLAMP_MSS))
 		return;
 
-	uint16_t mtu = source->mtu;
-	if(via != myself && via->mtu < mtu)
-		mtu = via->mtu;
-
 	/* Find TCP header */
 	int start = ether_size;
 	uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
@@ -163,7 +159,7 @@ static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *pac
 
 		/* Found it */
 		uint16_t oldmss = DATA(packet)[start + 22 + i] << 8 | DATA(packet)[start + 23 + i];
-		uint16_t newmss = mtu - start - 20;
+		uint16_t newmss = source->mtu - start - 20;
 		uint16_t csum = DATA(packet)[start + 16] << 8 | DATA(packet)[start + 17];
 
 		if(oldmss <= newmss)