[Bridge] 10 gigabit linux bridge problem

[Yavetskiy Yuriy <yavetskiy@kpi.ua>]

Hello.

I'm using 10 Gigabit Ethernet cards Intel 82598EB.When I use them for 
routing, everything is OK.But when I tried to add them to bridge several 
problems occured.I had already configured bridge from 2 gigabit Ethernet 
cards intel 82576. I added a single 10-Gigabit interface in there and 
got following result:
- packets are going from gigabyt to gigabyt port without any problems
- packets are coming ang going from 10 gigabyt to gigabyt port(ARP 
request is seen by tspdamp on destination), but the arp response does 
not reach the source host (ARP response is not seen by tspdamp on the 
original host).
At the same time at the bridge it can be seen by tspdamp that the 
arp-answer came from the Gigabit port, and thus can be seen that the 
package left at 10 gigabit ethernet port.
- ARP-requests are going from gigabyt port, they are seen by tcpdamp 
(tspdamp at gigabit port sees an incoming packet, tspdamp 10-Gigabit 
sees outgoing packet). But these packets are not visible at destination.
Everything is OK with the hosts at the ends of bridge , because if you 
replace the Linux bridge on FreeBSD bridge, the packets arrive.
If you are using only 10-Gigabit ports the same problem occurs, incoming 
packet is seen by tcpdamp on bridge on the incoming 10G interface and 
outgoing packet seen on an outgoing 10G port , but packets are is not 
visible at destination.
I checked MTU everywhere, everywhere the same. Tried to increase it's 
value on bridge interfaces- and got nothing.
All iptables and ebtables rules are permissive, default policy anywhere 
is ACCEPT.
I'm using Debian with vanilla kernel 2.6.32.2.
In what may be the problem?

-- 
WBR
Yavetskiy Yuriy
ULTI-RIPE