bridge only issues with packet mangling

Hiyas,
   I still seem to have nothing but problems with the
firewall bridge when I mangle packets.  Under 2.4.20
with the latest ebtables, I have 2 hosts with 2
interfaces.  eth0 on each connects to its own 'lan'
segment and the eth1's are connected.  I use a
userspace QUEUE target for my work with virtually no
rules currently.  For the rule:
iptables -t mangle -A POSTROUTING -o eth1 -j QUEUE
I, in my queue, mangle TCP/UDP packet payload, thus
mangling all packets going 'out' to the other linux
box.
For the rule:
iptables -t mangle -A PREROUTING -i eth1 -j QUEUE
I, in my queue, demangle TCP/UDP packet payload, thos
demangling all packets coming 'in' from the other
linux box.

This configuration works perfectly fine in router mode
and my userspace, in router mode and in bridge mode
with my demangling turned off also works fine.  As
soon as I try to mangle in bridge mode, however,
traffic slows to tens of bytes/second.  When moving
the input rule from PREROUTING to INPUT it seems to
pick up a bit, but I have no idea why that would make
a difference (especially since I would guess packets
seen on the 'PREROUTING' but not destined to go down
'INPUT' would just get dropped.  
In addition, because I was concerned that conn track
may somehow be affecting it, I installed the 'raw'
table patch from the patch-o-matic ('raw' table is
consulted before conn tracking), shifting the
PREROUTING rule from 'mangle' to 'raw'.  Router mode
still worked fine, but bridge is still not.
To the best of my knowledge, there is no load issue
(this is the only app & packet counter does not show
an excessinve # packets being passed).  Are there any
ideas why mangling packets would cause such a horrid
slowdown?

-Scott

__________________________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo
http://search.yahoo.com


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf