Skip to contentSkip to conversation

conntrack - UDP = good

DISCUSSION Tobias DiPasquale 4 participants 4 messages Jun 23, 2004 – Jun 24, 2004

lore
Share & tools
lore
#0 lore raw

Hi all,

Is there a way to get conntrack to _not_ track UDP connections (or also ICMP)? It seems rather pointless to me and its certainly taking up way too much memory on my box just for some DNS queries. Any ideas?

Signature · 2 lines
[ Tobias DiPasquale ]
0x636f6465736c696e67657240676d61696c2e636f6d
#1 lore raw
Quoted 5 lines · Tobias DiPasquale

Hi all,

Is there a way to get conntrack to _not_ track UDP connections (or also ICMP)? It seems rather pointless to me and its certainly taking up way too much memory on my box just for some DNS queries. Any ideas?

This may help:

http://lists.netfilter.org/pipermail/netfilter/2003-October/047892.html

Regards,

Antony.

Signature · 7 lines
Perfection in design is achieved not when there is nothing left to add, but 
rather when there is nothing left to take away.

 - Antoine de Saint-Exupery

                                                     Please reply to the list;
                                                           please don't CC me.
#2 lore raw
Quoted 5 lines · Antony Stone
Quoted 2 lines · Tobias DiPasquale

Hi all,

Quoted 8 lines
Quoted 3 lines

Is there a way to get conntrack to _not_ track UDP connections (or also ICMP)? It seems rather pointless to me and its certainly taking up way too much memory on my box just for some DNS queries. Any ideas?

This may help:

http://lists.netfilter.org/pipermail/netfilter/2003-October/047892.html

The raw table support is now an option in 2.6.6 and above.

Not sure about the iptables support that comes with your distro though.

#3 lore raw

Fecora kernel + conntrack + QUEUE mangle good?

Hello,
I have a userspace app which does some mangling on
packets in the userspace QUEUE target under 2.4.x. It relies on conntrack being loaded to defragment packets (it is in the POSTROUTING,mangle table). Works dandy under 2.4.x.
I loaded up a Fedora RC2 machine (2.6.5 maybe?) to
see how well it would work and for most packets it was fine. I was losing large packets, however (ie: 24K UDP packets). After some investigation and putting some checksum code in place, it looks like the packet's contents were getting corrupted by comparing checksums on the sender QUEUE and the receiver's QUEUE userspace apps. I am still checking on when this occurs, but when using 24K, the last bits of the defragmented packet were all '0's instead of the original value. I believe that if I do not alter the packet, that there are no issues (need to confirm today). I will be continuing tests today, but are there any known issues when altering large packets in userspace under the 2.6 kernel? Thanks in advance!!

-Scott

__________________________________
Do you Yahoo!?
Yahoo! Mail is new and improved - Check it out!
http://promotions.yahoo.com/new_mail