Joel <jo...@sdf.org> writes: > Hello -- How can I get NetBSD 10 to obey the MTU size advertised by my > OpenBSD router advertisement daemon? > > On OpenBSD, /etc/rad.conf has > mtu 1480 > > On NetBSD, my interface MTU is still 1500. > > It's a rockpro64 awge interface running NetBSD 10.1 GENERIC64. > > Note: I need to set MTU because I use tunnelbroker.net, which has a > max MTU of 1480. IPv6 and MTU can be cranky, especially through a > tunnel.
About Hurricane Electric (and anybody else's) tunnels, many others have the same sorts of feelings. I think you are blurring the default behavior of IPv6 the mtu of your interface, used for communicating on the LAN for things not IP the mtu that should be used for IPv4 default route the mtu that should be used for on-link IPv6 the mtu that should be used for the IPv6 default route PMTUD > tcpdump -vvv -A from OpenBSD of the RA to NetBSD: > > fe80::8647:9ff:fe2d:eeec > fe80::2480:27ff:fedf:b6ae: icmp6: router > advertisement(chlim=0, pref=medium, router_ltime=1800, > reachable_time=0, retrans_time=0)(src lladdr: > 84:47:09:2d:ee:ec)(prefix info: LA valid_ltime=5400, > preferred_ltime=2700, prefix=2001:470:e833:1ab::/64)[ndp opt] (len > 120, hlim 255) The local Ethernet can support 1500, surely. So when your NetBSD machine autoconfigures within your prefix, it will use the interface MTU when communicating with other addresses within that prefix, and fe80:. That seems correct. You didn't show us, but there is presumably a default route advertised also, via route6d or some such. That should perhaps be advertising a reduced mtu; it's really only packets using that default route that need to be shorter. There is also Path MTU Discovery, and in theory you should get back an ICMP6 fragmentation required, and TCP should drop the mss. Bad firewall config can mess that up. I am pretty sure the v6 rules say that packets of 1280 have to be supported, or are the normal assumption. I find that outgoing tcp6 packets have 1240 of payload, even though interface MTU is 1500. I also have a tunnel but my router is NetBSD (I don't think that matters). So, you can: Explain what oversize packets happen, figure out why, and ask about them. Figure out if there is some MTU associated with the default route and RIPng. Figure out if the base RA should have an MTU field; your tcpdump output does not show that. Figure out why PMTU detection is not working. Run "ifconfig awge0 1480" and stop worrying entirely.
