[Xen-users] Bridges stop forwarding

[Gavin Hamill <gdh@xxxxxxxxxxxxx>]

Hello :) I'm using 3.0.2 on Ubuntu Dapper, set up using the packages
from http://ubuntu.cs.mtsu.edu/packages/xen/ along with the
Dom0 kernel and modules from the 3.0.2 32-bit distribution tarball.

I am trying to model a network using multiple DomUs and bridge
instances, running Quagga and OSPF amongst them, but I'm experiencing a
very much more basic connectivity problem.

Very simply, after a while, a bridge stops forwarding, yet it still
knows about the MAC address on each port:

root@scribe:~# brctl showmacs tower-sw
port no mac addr                is local?       ageing timer
  2     00:16:3e:63:e2:8c       no                 1.99
  1     00:16:3e:73:4f:20       no                 0.44
  2     fe:ff:ff:ff:ff:ff       yes                0.00
root@scribe:~# brctl  show
...
u10-tower   8000.feffffffffff  no vif14.0
                                  vif10.2
...

I have 6 bridges in total, including xenbr0, and all have a bridge ID
of 8000.feffffffffff - is this bad? Hence, the MAC address of each
bridge is FE:FF:FF:FF:FF:FF (as well as that of each vifX.Y)

The DomU's at either end of the u10-tower bridge are 10.0.1.1 and
10.0.1.254, and when I try to ping each from its counterpart, I see this
on Dom0:

root@scribe:~# tcpdump -i tower-sw -n
tcpdump: WARNING: tower-sw: no IPv4 address assigned
tcpdump: verbose output suppressed, use -v or -vv for full protocol
decode listening on tower-sw, link-type EN10MB (Ethernet), capture size
96 bytes 
11:42:00.850064 arp who-has 10.0.1.1 tell 10.0.1.254
11:42:01.312926 arp who-has 10.0.1.254 tell 10.0.1.1
11:42:01.849986 arp who-has 10.0.1.1 tell 10.0.1.254
11:42:02.312928 arp who-has 10.0.1.254 tell 10.0.1.1
11:42:02.850053 arp who-has 10.0.1.1 tell 10.0.1.254
11:42:03.312926 arp who-has 10.0.1.254 tell 10.0.1.1

Even if I assign an IP address to u10-tower on Dom0, I can not ping
either host...

Here's the really interesting part.. watch what happens as 10.0.1.254
shuts down starting with a few 1-second pings from 10.0.1.1... 

12:12:02.532932 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:03.532925 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:04.532930 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:05.532932 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:06.532932 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:07.532936 arp who-has 10.0.1.254 tell 10.0.1.1
12:12:07.852683 IP 10.0.1.1 > 10.0.1.254: ICMP echo request, id 12039,
seq 15, length 64 
12:12:07.853398 IP 10.0.1.1 > 10.0.1.254: ICMP echo request, id 12039,
seq 16, length 64 
12:12:07.853814 arp reply 10.0.1.1 is-at 00:16:3e:63:e2:8c
12:12:07.854081 arp reply 10.0.1.1 is-at 00:16:3e:63:e2:8c
12:12:07.854474 arp reply 10.0.1.1 is-at 00:16:3e:63:e2:8c
12:12:07.854707 arp reply 10.0.1.1 is-at 00:16:3e:63:e2:8c
12:12:07.855049 arp reply 10.0.1.1 is-at 00:16:3e:63:e2:8c
(there were dozens of these replies)

The echo request + subsequent flood of ARP replies happens the instant
the DomU says "System halted".

It appears to be a repeatable problem - I just let the system sit for
~15 mins and the bridge stops forwarding :(

Does this sound at all familiar to anyone? :)

Cheers,
Gavin.



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