Hello,
I hope it it just a configuration problem, but I am
experiencing thruput problems with bridging and packet
mangling.

I have a simple comfiguration:

HostA-------HostB--------RestOfNet

HostA and HostB both have 2.4.20 with ebtables/bridge
patches so I can specify physical devices in iptables
when I have HostB configured as a bridge.  I am not
using (currently) iptables for security, but rather so
I can inject a userspace module on traffic going from
HostA (eth1) to HostB (eth0).
When my userspace module just allows traffic thru all
is happy.  My problem comes when I mangle packets.  I
mangle the TCP/UDP payloads (recomputing the IP and
TCP/UDP checksums) between the 2 hosts (so the packets
are valid packets between the 2 but garbage to
sniffers).  It works fine when the HostB is not in
bridge mode (just a configured eth0).
The QUEUE is set for PREROUTING when coming into the
device (-A PREROUTING -i eth0 -j QUEUE) and
POSTROUTING when going out of the device (-A
POSTROUTING -o eth0 -j QUEUE) [both examples are for
HostB].  This allows me to get the packets as the come
in from across the wire (origin from the wire into
eth0) and before packets head out (origin from eth1 or
internal).

My main test is with FTP.  When both machines are
hooked w/out bridging, it works perfectly.  When I
turn the mangling off (but still have packets run thru
the queue) it works fine.  The headers are properly
checksummed headers (tho the payload is different and
the checksums are different).  Is there anything in
the beidge code which would be causing this type of
problem?

-Scott


__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com
_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge

Reply via email to