Xen users - Feb 2008 - Arp problem with network bridging and VLANs

I''m attempting to put together a Xen box, which has multiple VLANs 
running into it, on one bridged interface.  I mostly copied the way it 
was done here: http://renial.net/weblog/2007/02/27/xen-vlan/

I basically have:
bond0 is the bond device (composed of eth0 and eth1), vlan757 is the 
vlan device for vlan 757 on bond0, and "xen757" is a bridge created by
the above scripts, which has the vlan757 device in it, and which I am 
using for the Xen VM''s bridge interface.

from ip addr show:
--------
2: eth0: <BROADCAST,MULTICAST,SLAVE,UP,10000> mtu 1500 qdisc pfifo_fast 
master bond0 qlen 1000
    link/ether 00:14:4f:7d:8a:46 brd ff:ff:ff:ff:ff:ff
3: eth1: <BROADCAST,MULTICAST,SLAVE,UP,10000> mtu 1500 qdisc pfifo_fast 
master bond0 qlen 1000
    link/ether 00:14:4f:7d:8a:46 brd ff:ff:ff:ff:ff:ff
6: bond0: <BROADCAST,MULTICAST,PROMISC,MASTER,UP,10000> mtu 1500 qdisc 
noqueue
    link/ether 00:14:4f:7d:8a:46 brd ff:ff:ff:ff:ff:ff
    inet 172.18.6.25/24 brd 172.18.6.255 scope global bond0
10: vlan757@bond0: <BROADCAST,MULTICAST,NOARP,MASTER,UP,10000> mtu 1500 
qdisc noqueue
    link/ether fe:ff:ff:ff:ff:ff brd ff:ff:ff:ff:ff:ff
24: xen757: <BROADCAST,MULTICAST,NOARP,UP,10000> mtu 1500 qdisc noqueue
    link/ether fe:ff:ff:ff:ff:ff brd ff:ff:ff:ff:ff:ff
---
(There''s a few more vlan and xen bridge devices than I show, but they 
all look the same as these, and aren''t currently being used.)

from brctl show:
----
xen757          8000.feffffffffff       no              vlan757
                                                        vif10.0
----

and on the domU guest:
----
eth0 Link encap:Ethernet  HWaddr 00:16:3E:45:AD:38
     inet addr:128.211.157.78  Bcast:128.211.157.255  Mask:255.255.255.0
     UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
     RX packets:11588 errors:0 dropped:0 overruns:0 frame:0
     TX packets:373 errors:0 dropped:0 overruns:0 carrier:0
     collisions:0 txqueuelen:1000
     RX bytes:863098 (842.8 KiB)  TX bytes:24161 (23.5 KiB)
---

The problem seems to be that arp replies don''t get passed through the 
bridge device (xen757) from vlan757 to vif10.0.  I''ve messed around 
running tcpdump on the various interfaces, and I can see arp queries 
going out ok (and coming in ok, as other machines on the VLAN can send 
arp requests through to the domU guest, and replies go back ok).

However, I can see the who-has arp replies on vlan757 and xen757, but 
they don''t show up when I have tcpdump listening to vif10.0 on the dom0
host.  I''ve also noticed that (1) manually setting the arp entries on 
domU makes the domU host be able to talk to the other hosts ok, and (2) 
occasionally (but infrequently), arp replies do get all the way through 
to the domU guest.  It seems that replies from the router (which is on 
a Cisco Cat 6509 FWIW) get through just fine, but replies from other 
hosts on the same subnet do not.

Does anyone have any ideas?  I''m running Debian/etch on the dom0 and 
domU hosts.

Thanks,

Pat
-- 
Purdue University Research Computing --  http://www.itap.purdue.edu/rcac

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