OpenVZ Forum


Home » Mailing lists » Users » MDNS service discovery with VETH interface
MDNS service discovery with VETH interface [message #13740] Mon, 04 June 2007 14:09 Go to next message
Daniel Pittman is currently offline  Daniel Pittman
Messages: 26
Registered: January 2007
Junior Member
G'day. I have a VE that I set up using a VETH interface rather than a
VENET interface so I could test ZeroConf service discovery and other
broadcast/multicast protocols.

I have the VETH device bridged with the physical Ethernet device on the
hardware node and can see standard broadcast packets without any
problem.

I don't see the multicast packets cross the interface boundary though;
it is my belief that the VETH device doesn't correctly handle the
packets being sent to the Ethernet MAC '01:00:5e:00:00:fb'

This is running kernel '2.6.18-028stab033.1-ovz'


Have I missed some critical step in getting multicast services working
with the VETH device or so?


I think, reading the veth.c source, that the problem is in the veth_xmit
function on line 305 in the current HEAD in get -- the code tests:

if (!is_broadcast_ether_addr(((struct ethhdr *)skb->data)->h_dest)) {

This should probably also pass a multicast packet through to the VE, not
just the broadcast packets. That way multicast traffic is seen inside
the system.

The appropriate test would be 'is_multicast_ether_addr(...)' in addition
to the broadcast.

I have not supplied a patch here because I don't know what, if any,
extra implications permitting multicast packets in have -- or if there
is some other mechanism that I have missed that would cause them to pass
through as expected.

Regards,
Daniel
--
Digital Infrastructure Solutions -- making IT simple, stable and secure
Phone: 0401 155 707 email: contact@digital-infrastructure.com.au
http://digital-infrastructure.com.au/
Re: MDNS service discovery with VETH interface [message #13741 is a reply to message #13740] Mon, 04 June 2007 14:31 Go to previous message
dev is currently offline  dev
Messages: 1693
Registered: September 2005
Location: Moscow
Senior Member

Daniel Pittman wrote:
> G'day. I have a VE that I set up using a VETH interface rather than a
> VENET interface so I could test ZeroConf service discovery and other
> broadcast/multicast protocols.
>
> I have the VETH device bridged with the physical Ethernet device on the
> hardware node and can see standard broadcast packets without any
> problem.
>
> I don't see the multicast packets cross the interface boundary though;
> it is my belief that the VETH device doesn't correctly handle the
> packets being sent to the Ethernet MAC '01:00:5e:00:00:fb'
>
> This is running kernel '2.6.18-028stab033.1-ovz'
>
>
> Have I missed some critical step in getting multicast services working
> with the VETH device or so?
>
>
> I think, reading the veth.c source, that the problem is in the veth_xmit
> function on line 305 in the current HEAD in get -- the code tests:
>
> if (!is_broadcast_ether_addr(((struct ethhdr *)skb->data)->h_dest)) {
>
> This should probably also pass a multicast packet through to the VE, not
> just the broadcast packets. That way multicast traffic is seen inside
> the system.
>
> The appropriate test would be 'is_multicast_ether_addr(...)' in addition
> to the broadcast.
>
> I have not supplied a patch here because I don't know what, if any,
> extra implications permitting multicast packets in have -- or if there
> is some other mechanism that I have missed that would cause them to pass
> through as expected.

It should be ok to add this check. Thanks for noticing.
patch attached and commited.

Thanks,
Kirill


--- ./drivers/net/veth.c.ve2378 2007-06-04 18:26:11.000000000 +0400
+++ ./drivers/net/veth.c 2007-06-04 18:26:16.000000000 +0400
@@ -302,7 +302,7 @@ static int veth_xmit(struct sk_buff *skb
if (unlikely(rcv->owner_env->disable_net))
goto outf;
/* Filtering */
- if (!is_broadcast_ether_addr(((struct ethhdr *)skb->data)->h_dest)) {
+ if (!is_multicast_ether_addr(((struct ethhdr *)skb->data)->h_dest)) {
if (ve_is_super(dev->owner_env) &&
!ve_is_super(rcv->owner_env) &&
!veth_from_netdev(rcv)->allow_mac_change) {
Previous Topic: Live CD fails in live migration.
Next Topic: Debian etch + 2.6.18 + openvz
Goto Forum:
  


Current Time: Tue Dec 06 11:47:32 GMT 2022

Total time taken to generate the page: 0.02136 seconds