Xen users - Sep 2005 - RE: veth0 is from netback and vifu.0 is fromthebridge-utils?

> That''s great, except when you need *two* bridges, for two 
> independent subnets, and there is no, as far as I can see, peth1.
You can have multiple vethX interfaces on recent builds -- see
netback/loopback.c.
I think the current default is a rather miserly 1 (and should be increased), but
can be set on the command line.

Ian  
 
> Em Wednesday 14 September 2005 16:30, John Wilson escreveu:
> > Yo,
> >
> > I had big problems with this myself, but now as far as I can tell, 
> > veth0 (renamed to eth0 in the v recent releases) is the virtual nic 
> > for dom0 which corresponds to the virtual eth0 nics in the 
> domU''s (the 
> > physical nic itself has been renamed to peth0).
> >
> > The vifU.N''s are the dom0 links to the virtual nics ethN in 
> domU, and 
> > aren''t to be treated as proper network interfaces in thier 
> own right. 
> > If you assign an IP address to, say, vif1.0 and attempt to 
> communicate 
> > with eth0 in dom1, you can ping between them and such, but 
> ssh nor vnc 
> > wont work, complaining for the most part about bad checksums.
> >
> > The vif0.0 interface in dom0 is linked to dom0''s virtual nic,
veth0
> > (or eth0 in the recent builds).
> >
> > Now... The correct way to interface the cards and provide 
> connectivity 
> > between the domains and the outside world is to attach all 
> the vifU.N 
> > interfaces to a bridge in dom0 (normally xen-br0) together with the 
> > physical nic, eth0 (or peth0 recently). IP addresses are 
> then assigned 
> > to the eth0 links and inter-domain sshing, vncing and communication 
> > with the outside world all result.
> >
> > Here''s a badly drawn diagram to illustrate...
> >
> >   ____________________________         _____________________
> >
> > |dom0                        |       |dom1                 |
> > |  _______                   |       |       ________      |
> > |
> > | |eth0   |                  |       |      |eth0    |     |
> > | |IP_____|   +--------------+-------+------|IP______|     |
> > |
> > | ____|_______|_____________ |       |                     |
> > |
> > || ___|___ ___|___ _______  ||       |                     |
> > ||
> > |||vif0.0 |vif1.0 |peth0  | ||       |                     |
> > |||_______|_______|_______| ||       |                     |
> > ||
> > ||xen-br0_____________|_____||       |                     |
> > |
> > |_____________________|______|       |_____________________|
> >
> > ______________________|_____________________>>LAN>>>
> >
> >
> > I''ve used eth0 as the virtual nics and peth0 as the physical
nic.
> > IP has been used to denote devices with an IP address. All the 
> > interfaces there are automatically generated by Xen, and I used 
> > ifconfig to statically set IP addresses (because I was 
> having trouble 
> > with our lab DHCP server, although it was assigning 
> addresses to the 
> > correct cards), and brctl addif xen-br0 vif1.0 etc. to add 
> the vifs to 
> > the bridge, because for some reason Xen wasn''t...
> >
> > Well thats my contribution for this evening.
> >
> >
> >
> > John
> >
> >
> >
> > P.S. Official documentation/Howto/wiki is a little on the 
> sparse side 
> > for this Ian.
> >
> > Quoting Anthony.Golia@MorganStanley.com:
> > > On Thu, 8 Sep 2005, Ian Pratt wrote:
> > >> > Hi.  what creates veth0 and vifu.0, when? 
what''s the
> underlying 
> > >> > architecture behind them?  Just trying to understand the
> > >> > internals.
> > >>
> > >> netback/loopback.c creates them.
> > >>
> > >> It''s effectively a point to point link, allowing
domain 0 to
> > >> connect on to the bridge in the same manner that other 
> domains do. 
> > >> (before, packets to domain 0 were short-cutted and came off
the
> > >> bridge directly, which could lead to traffic from other
domains
> > >> being stalled if a lot of traffic was destined to user 
> space in dom0.
> > >
> > > thx.  what''s the diff betwn the two (veth0 and vifu.0) ?
> > >
> > >> Best,
> > >> Ian
> > >>
> > >> )
> > >
> > > Cheers,
> > > Anthony
> > >
> > > _______________________________________________
> > > Xen-users mailing list
> > > Xen-users@lists.xensource.com
> > > http://lists.xensource.com/xen-users
> >
> > _______________________________________________
> > Xen-users mailing list
> > Xen-users@lists.xensource.com
> > http://lists.xensource.com/xen-users
> 
> --
> Ramiro Brito Willmersdorf  Dep. Engenharia Mecânica/UFPE
> ramiro@willmersdorf.net        tel: +55 81 2126-8231e239          
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
> 
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

That''s excellent. I''ll wait for Rik to pick this one up for a
testing .rpm, though, because I never *ever* managed to build a 
RedHat kernel with external patches :)

Em Thursday 15 September 2005 16:05, Ian Pratt escreveu:
> > That''s great, except when you need *two* bridges, for two
> > independent subnets, and there is no, as far as I can see, peth1.
>
> You can have multiple vethX interfaces on recent builds -- see
> netback/loopback.c. I think the current default is a rather miserly 1 (and
> should be increased), but can be set on the command line.
>
> Ian
-- 
Ramiro Brito Willmersdorf  Dep. Engenharia Mecânica/UFPE
ramiro@willmersdorf.net        tel: +55 81 2126-8231e239          

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