'INPUT/OUTPUT' vs 'FORWARD' (was 'any issues with packet updates')

Hiyas,  
A slightly new focus given some results.  As before, I
basically have 2 linux boxes, both dual homed, both
with 2.4.20 and the latest ebtables stuff.  'eth0' is
considered to have local hosts behind it and the
'eth1's are connected together.
I write rules in iptables, mainly to send traffic thru
QUEUE to a custom userspace module.  This module
mangles TCP & UDP packet payloads, changing size and
contents.  The intent is to mangle packet data across
eth1, each machine mangling as it goes out eth1 and
demangling as it comes back in on eth1.  The userspace
module, when mangling is off (but the rest of the
userspace module is traversed) works fine. 
When I set up rules such as:
iptables -t mangle -A PREROUTING -j QUEUE -i eth1
iptables -t mangle -A POSTROUTING -j QUEUE -o eth1
Traffic thru is horribly slow, tens or hundreds of
bytes/second.  (note I set the rules up the above so I
can determine incoming (-i eth1) packets to demangle
and outgoing (-o eth1) to mangle).
From suggestions, I changed the rules to ket off of
INPUT and OUTPUT, just testing the mangling of packets
from/to the 2 linux boxes.  That, as it turns out,
works.  When I add in a FORWARD line, however, the
forwarded traffic still has the same slowdowns.  Any
ideas on why the FORWARD chain would act differently?

-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