MTU questions
Shadow Hawkins on Thursday, 28 July 2011 18:17:56
I understand MTU is important for IPv6 since there is no fragmentation (but apparently there is a super-datagram protocol, and there could be a stream protocol built on that in the future). Our IP4 ISP connection has MTU 1500 (ethernet).
On Redhat(Centos) 5.6:
A 6to4 tunnel on that gets MTU 1480. This makes sense, subtract 20 for the IP header.
However, the tunnel set up by aiccu has MTU 1280. What are the additional 200 bytes per packet used for? Or is aiccu assuming a PPP connection or something?
A point-to-point 6in4 tunnel used to route IP6 packets over an IP4 IPSec connection has default MTU 1480. Do I need to subtract the IPSec overhead? Or does the full path MTU get auto computed somehow? Is there a way to query what the effective MTU between two IPv6 nodes is?
MTU questions
Jeroen Massar on Thursday, 28 July 2011 18:22:57
You most likely mean a 6in4 (proto-41) tunnel instead of '6to4' (though 6to4 just uses 6in4)
However, the tunnel set up by aiccu has MTU 1280.
That is the default, see the FAQ on ways to change it.
Or is aiccu assuming a PPP connection or something?
It is a default for all tunnels in SixXS and it is the default as it is the minimum packet size that is supposed to be supported for IPv6 and it is also the packet size that works almost anywhere.
Do I need to subtract the IPSec overhead?
Depends where the IPSEC happens. If it happens on the IPv4 layer, then yes as it will be between IPv4 - IPSEC - IPv6, if it happens on the IPv6 layer it won't as it will be included in the IPv6 packet size.
Is there a way to query what the effective MTU between two IPv6 nodes is?
The protocol is called "Path MTU discovery", use traceroute6 on linux for instance to figure this out.
MTU questions
Shadow Hawkins on Thursday, 28 July 2011 23:10:24
Used tracepath6 to experimentally find MTU of 1384 for 6in4 over IPv4 IPSec. This says IPSec uses 76 bytes.
1500(Ether) - 20(IPV4) - 76(IPSec) - 20(6in4) = 1384
Posting is only allowed when you are logged in. |