Any issues with packet updates?
Scott MacKay <scottmackay-/[email protected]>
| Newsgroups | gmane.linux.network.bridge.ebtables.devel |
|---|---|
| Message-ID | <[email protected]> |
Question summary: Are there any issues when mangling packets as they pass thru the bridge, changing TCP/UDP payload content and size? The symptoms are very, very slow thruput (sometimes higher volume bursts) but traffic eventually completing. Details I am using RH 8.0, with a 2.4.20 kernel containing the bridge 0.0.10 and ebtables 2.0.003 loaded. I have a pair of machines, both with 2 ether cards. Each has their 'intranet' on eth0 and their 'internet' on eth1 (tho the 2 eth1s are just in a hub connecting the 2). I have a userspace (iptables) QUEUE target on my dual ether machine, loaded in PREROUTING (marked for when incoming to the physical interface) and POSTROUTING (marked for when outgoing on the physical interface) on eth1. The jist is basically all outgoing packets (POSTROUTING) between the hosts will be marked on the outdev_name field in the userspace module and all incoming packets (PREROUTING) will be marked on the indev_name field in the userpace module. Overall, the intent of the userspace module is to mangle TCP/UDP packets on the 'internet'. Each side mangles on output, demangles on input, changing both the tcp/udp payload and it's size. My configuration, when the mangling is turned off, works perfectly fine with great thruput. As soon as I activate the mangling, it slows to a crawl, but does get there and is received correctly. In router mode, everything looks like it works perfectly fine. __________________________________________________ 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