| Question regarding ipv6 MTU I have a HE tunnel terminated at a router running Tomato-RAF. On my LAN I have a mixture of clients running Windows 7, XP and Open Suse.
Without making any changes to the ipv6 configuration on the router, aside from configuring the tunnel, I can ping ipv6 sites, however, with the exception of ipv6.google.com, the browser times-out when I try to connect. I also fail the large packet size test at test-ipv6.
Following some advice I found at Sixxs (»www.sixxs.net/wiki/Windows_configuration) I changed the ipv6 mtu on my windows box to 1280 and everything worked correctly. I was also able to pass all tests at test-ipv6.
As an experiment, I removed the router, created a fresh install of Windows 7 and terminated the tunnel on the PC. without making any additional adjustments, I was able to connect to all ipv6 sites and pass all tests.
I carried out some additional tests using tracepath6 and and by sending various sized ICMP packets and it would seem the maximum ipv6 mtu I can use is 1440.
So my question is, why do I have to adjust the MTU size on the PC when the router is the terminus for the tunnel but not when the PC is the terminus?
On the router the MTU is set to 0 which is default.
Thanks |
|
|
|
 timcuthBraves FanPremium
join:2000-09-18
Pelham, AL
 ·AT&T Southeast
| I don't know. I have not had any of those problems. My tunnel terminates at my router and is distributed to an Ubuntu PC, a Win 7 PC, and a Win XP PC. All are working well and I have not touched MTU settings, anywhere.
Tim
--
"Life is like this long line, except at the end there ain't no merry-go-round." - Arthur on The King of Queens
~ Project Hope ~ |
|
 leiboldPremium,MVM
join:2002-07-09
Sunnyvale, CA
kudos:2
·SONIC.NET
·Pacific Bell - SBC
| reply to Dapper
By default an IPv6 tunnel assumes that the IPv4 connectivity between the two tunnel endpoints can transmit a normal payload of an Ethernet frame (1500 bytes) without fragmentation.
If this is not the case (for example because your ISP uses PPPoE) you may have to specify a lower MTU.
--
Got some spare cpu cycles ? Join Team Helix or Team Starfire! |
|
 Logan_APremium
join:2010-12-14
Methuen, MA | reply to Dapper
I believe Tomato-RAF uses the router advertisement daemon radvd to inform the IPv6 hosts the settings for the network
You may need to edit/add an AdvLinkMTU entry in the radvd.conf file under the interface section to set the IPv6 MTU for your network.
I don't use Tomato, but I quickly read over the code at »repo.or.cz/w/tomato.git/blob/ref···s.c#l534
it seem if you set the nvram var "ipv6_tun_mtu" to 1440, it will set the AdvLinkMTU entry to 1440 in the radvd config file. |
|
 rchandraStargate Universe fanPremium
join:2000-11-09
14225-2105 | reply to Dapper
Unlike IPv4, IPv6 routers will not fragment packets. As you're likely connected via Ethernet, you have a maximum of 1500 octets of payload. Some (I don't remember how much) is taken up by your 6in4 tunnel. Some more may be taken up, as another poster noted, by PPPoE if your ISP uses it.
An MTU of 0 sounds wrong. Some software may be doing PMTU discovery, and 0 may be being returned and tripping it up. An MTU of 1280 is the minimum which MUST be supported under IPv6. I would personally be setting my MTU on "inside" systems to about 1450, either through setting it explicitly or, as another poster suggested, with the advertised MTU. If possible, I would also be sure a similar MTU option is sent with your DHCPOFFERs.
--
English is a difficult enough language to interpret correctly when its rules are followed, let alone when a writer chooses not to follow those rules.
Jeopardy! replies and randomcaps REALLY suck! |
|
1 edit | reply to Dapper
Thank you for the replies, this is looking like it's something to do with pppoe, which I do use. I checked the mtu size for ipv4 and the max allowed was 1460.
@Logan_A thanks for pointing out the settings in radvd,I'll take a look at those.
@rchandra Apologies for my poor explanation regarding the mtu settings in the firmware. In the case a 0 (zero) in the mtu field in the router simply indicates the default value. From what I can gather, the value it uses is whatever the underlying mtu happens to be. running ifconfig shows it set to 1440 for the six0 interface, which is the same value I discovered yesterday.
I guess this starting to make sense now. Because the router doesn't fragment ipv6 and it's set itself to 1440 the original configuration for windows, which had an mtu of 1500 would have failed. So, reducing the value on the PC to a value less than 1440 allowed the packets to be sent.
I've now set the value on the Windows PC, at leats, to 1440 and all seems well. I'll take a look at DHCP options and radvd.
|
|
