Re: [PATCH net-next 1/3] udp_tunnel: allow to turn off path mtu discovery on encap sockets

[David Ahern <dsahern@gmail.com>]

On 7/18/20 12:56 AM, Stefano Brivio wrote:
> On Fri, 17 Jul 2020 09:04:51 -0600
> David Ahern <dsahern@gmail.com> wrote:
> 
>> On 7/17/20 6:27 AM, Stefano Brivio wrote:
>>>>  
>>>>> Note that this doesn't work as it is because of a number of reasons
>>>>> (skb doesn't have a dst, pkt_type is not PACKET_HOST), and perhaps we
>>>>> shouldn't be using icmp_send(), but at a glance that looks simpler.    
>>>>
>>>> Yes, it also requires that the bridge has IP connectivity
>>>> to reach the inner ip, which might not be the case.  
>>>
>>> If the VXLAN endpoint is a port of the bridge, that needs to be the
>>> case, right? Otherwise the VXLAN endpoint can't be reached.
>>>   
>>>>> Another slight preference I have towards this idea is that the only
>>>>> known way we can break PMTU discovery right now is by using a bridge,
>>>>> so fixing the problem there looks more future-proof than addressing any
>>>>> kind of tunnel with this problem. I think FoU and GUE would hit the
>>>>> same problem, I don't know about IP tunnels, sticking that selftest
>>>>> snippet to whatever other test in pmtu.sh should tell.    
>>>>
>>>> Every type of bridge port that needs to add additional header on egress
>>>> has this problem in the bridge scenario once the peer of the IP tunnel
>>>> signals a PMTU event.  
>>>
>>> Yes :(
>>
>> The vxlan/tunnel device knows it is a bridge port, and it knows it is
>> going to push a udp and ip{v6} header. So why not use that information
>> in setting / updating the MTU? That's what I was getting at on Monday
>> with my comment about lwtunnel_headroom equivalent.
> 
> If I understand correctly, you're proposing something similar to my
> earlier draft from:
> 
> 	<20200713003813.01f2d5d3@elisabeth>
> 	https://lore.kernel.org/netdev/20200713003813.01f2d5d3@elisabeth/
> 
> the problem with it is that it wouldn't help: the MTU is already set to
> the right value for both port and bridge in the case Florian originally
> reported.

I am definitely hand waving; I have not had time to create a setup
showing the problem. Is there a reproducer using only namespaces?

> 
> Also, given the implications on overriding configured MTUs, and
> introducing (further) IP logic into the bridge, if Florian's idea of
> injecting ICMP messages could be implemented in a generic function:
> 
> On Wed, 15 Jul 2020 16:33:56 +0200
> Florian Westphal <fw@strlen.de> wrote:
> 
>> Yes, it might be possible to move the proposed icmp inject into
>> skb_tunnel_check_pmtu() -- it gets the needed headroom passed as arg,
>> it could detect when device driver is in a bridge and it already knows
>> when skb has no dst entry that it a pmtu change could be propagated to.
> 
> I think that would be preferable: then it's fixed for all tunnels in a
> generic, probably simpler way, without those two issues.
> 
> But then again, we're talking about Linux bridge. Unfortunately this
> doesn't fix the problem with Open vSwitch either.
> 

Of course. (insert sarcastic jab at ovs here)