CARP vs. if_bridge

[Boris Kochergin]

Max Laier wrote:
> On Thursday 18 February 2010 18:02:55 Boris Kochergin wrote:
>   
>> Ahoy. I'm seeing what appears to be erroneous interaction between CARP
>> and if_bridge on multiple machines with a variety of Ethernet
>> controllers and architectures. I've observed it on 7.2-R and 8.0-R. The
>> test setup is simple enough:
>>
>> CARP master:
>>
>> FreeBSD t30 8.0-RELEASE-p1 FreeBSD 8.0-RELEASE-p1 #5: Sun Feb 14
>> 20:22:41 EST 2010     root at t30:/usr/obj/usr/src/sys/T30  i386
>>
>> lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> metric 0 mtu 16384
>>         options=3<RXCSUM,TXCSUM>
>>         inet6 fe80::1%lo0 prefixlen 64 scopeid 0x1
>>         inet6 ::1 prefixlen 128
>>         inet 127.0.0.1 netmask 0xff000000
>> dc0: flags=8943<UP,BROADCAST,RUNNING,PROMISC,SIMPLEX,MULTICAST> metric 0
>> mtu 1500
>>         options=8<VLAN_MTU>
>>         ether 00:04:5a:a8:e0:bf
>>         inet 192.168.0.2 netmask 0xffffff00 broadcast 192.168.0.255
>>         media: Ethernet autoselect (100baseTX <full-duplex>)
>>         status: active
>> carp0: flags=49<UP,LOOPBACK,RUNNING> metric 0 mtu 1500
>>         inet 192.168.0.1 netmask 0xffffff00
>>         carp: MASTER vhid 1 advbase 1 advskew 0
>>
>> CARP backup:
>>
>> FreeBSD ultra5 8.0-RELEASE FreeBSD 8.0-RELEASE #0: Thu Feb 18 15:19:39
>> UTC 2010     boris at ultra5:/usr/obj/usr/src/sys/GENERIC.carp  sparc64
>>
>> hme0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 1500
>>         options=b<RXCSUM,TXCSUM,VLAN_MTU>
>>         ether 08:00:20:f5:65:d4
>>         media: Ethernet autoselect
>> xl0: flags=8943<UP,BROADCAST,RUNNING,PROMISC,SIMPLEX,MULTICAST> metric 0
>> mtu 1500
>>         options=9<RXCSUM,VLAN_MTU>
>>         ether 00:01:03:2c:06:6d
>>         inet 192.168.0.3 netmask 0xffffff00 broadcast 192.168.0.255
>>         media: Ethernet autoselect (100baseTX <full-duplex>)
>>         status: active
>> lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> metric 0 mtu 16384
>>         options=3<RXCSUM,TXCSUM>
>>         inet6 fe80::1%lo0 prefixlen 64 scopeid 0x4
>>         inet6 ::1 prefixlen 128
>>         inet 127.0.0.1 netmask 0xff000000
>> carp0: flags=49<UP,LOOPBACK,RUNNING> metric 0 mtu 1500
>>         inet 192.168.0.1 netmask 0xffffff00
>>         carp: MASTER vhid 1 advbase 1 advskew 100
>> bridge0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu
>> 1500
>>         ether 3a:e6:09:2d:da:bc
>>         id 00:00:00:00:00:00 priority 32768 hellotime 2 fwddelay 15
>>         maxage 20 holdcnt 6 proto rstp maxaddr 100 timeout 1200
>>         root id 00:00:00:00:00:00 priority 32768 ifcost 0 port 0
>>         member: xl0 flags=143<LEARNING,DISCOVER,AUTOEDGE,AUTOPTP>
>>                 ifmaxaddr 0 port 2 priority 128 path cost 200000
>>         member: hme0 flags=8<SPAN>
>>                 ifmaxaddr 0 port 1 priority 128 path cost 200000
>>
>> In summary, I have a basic CARP configuration and, on the backup CARP
>> machine, a bridge with the CARP device's physical interface in it. The
>> purpose of this setup is the ability to monitor traffic passing through
>> that interface using another machine. If the master CARP machine is
>> disconnected from the network, the CARP interface on the backup machine
>> correctly changes to the MASTER state, but does not act on traffic bound
>> for the shared IP address--192.168.0.1. tcpdump shows the traffic coming
>> in on the correct physical interface, but it is never replied to, or, in
>> the case of routing, forwarded. Removing xl0 from the bridge on the
>> backup machine instantly fixes this, and the shared IP address behaves
>> as expected. Adding xl0 back to the bridge while the backup CARP
>> interface is in the MASTER state keeps things running correctly, so the
>> problem is only observed when xl0 is part of the bridge during the CARP
>> transition from BACKUP to MASTER. Thoughts?
>>     
>
> I assume the bridge filters out the traffic as it thinks the destination is 
> elsewhere (it has previously seen ARPs from the other MASTER entering via 
> xl0).  It shouldn't do that, but that's a different story.  You can try to 
> force edge or ptp status on xl0, not sure if this does the trick, but it's 
> worth a try.
>
> Regards,
>   Max
>   
Sure. No go, though, I'm afraid. It's not an operational show-stopper 
for me, at least. In the worst case, I can always hack up a PCAP program 
to copy the frames around in user space.

-Boris