Re: [Xen-devel] vif limitation to 1500 bytes (no vlan ?)

[Yann Dupont <Yann.Dupont@xxxxxxxxxxxxxx>]

Ian Pratt a écrit :

>  
>
>>I have a small problem; Is there a known limitation on vif 
>>code - Network-device backend or frontend - which prohibits 
>>the usage of tagged
>>(802.1q) frames inside a domU ?
>>
>>In dom0 I have a bridge with eth0 (epro1000) wich is 
>>connected to a 802.1q port this same bridge has also a vif 
>>connected, corresponding to a domU.
>>
>>Inside this domU , I can use vconfig to have numerous 
>>interfaces derived from the vif, one by vlan, BUT can only 
>>safely use a MTU of 1496. If I use the default MTU I have problems.
>>
>>For the moment I have solved the problem by using the 
>>physicall interface INSIDE the domU. But I'd like to know if 
>>this problem is known ?
>>    
>>
>
>The frontend/backend itself should work for any MTU up to something shy
>of 4KB.
>  
>
ok, good to know,

>I suspect the 1496 restriction is actually being introduced in dom0 by
>the 802.1q tagging, but this information is not being propagated to the
>frontend in the domU.
>Because this is a bridged rather than routed setup, MTU discovery won't
>help.
>  
>
yes

>I've a feeling that an 'mtu=' parameter might have been added a while
>back so that you could fix the MTU of a frontend vif in the domain
>  
>
ah , that's good news. I'll go searching for that...

>config file. We could figure this out automatically by looking at the
>bridge the vif is connected to and then pick the lowest MTU of any
>  
>
>device connected to the bridge. Patches welcome.
>  
>
A manual way to do this is already a good start...

thanks for your answer.

-- 
Yann Dupont, Cri de l'université de Nantes
Tel: 02.51.12.53.91 - Fax: 02.51.12.58.60 - Yann.Dupont@xxxxxxxxxxxxxx



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel