Is there anyway for Netfilter_queue to provide only control packets.

314 views

I am trying to provide ability to intercept network connections originating from, and coming into Linux machines at various stages during the network connections life-cycle.

These stages include the following:
1. Just before an outbound network connection is made - i.e. when the first SYN packet is sent out.
2. Just after an outbound network connection is established.
3. Just after the connection is terminated.
4. When an inbound connection is established - i.e. when the first SYN packet is received from outside.

For above requirements, I was evaluating netfilter_queue to get the packets in the user-space and then decide the verdict whether to allow or drop the packet.

My main concern here is, that once I am done with netfilter_queue registration, I'll start getting all the packets. But I am only interested in control packets and don't want data packets to be sent to user-space.

So, my question is - Is there any existing way to tell the netfilter_queue kernel module to send only the control packets and not the data packets ? Also, would it be possible to get notified about the connection establishment and termination ?

If there is no ready way to achive above, then would it make sense to modify libnfnetfilter_queue and netfilter_queue kernel module to provide only control packets depending upon the config mode set,
i.e. introducing NFQNL_COPY_CONTROL_PACKET for copying only control packets to user-space ?

posted Oct 1, 2013 by Naveena Garg

Looking for an answer? Promote on:

Facebook Share Button

Twitter Share Button

LinkedIn Share Button

Similar Questions

+2 votes

Can not see VM packets using Netfilter

I’m recently developing a simple program using netfilter and I’m having a tricky problem. My program is mainly to log the src and dst ip address of all the packets. This program is run in the host machine. I have several virtual machines on that host machine.

The problem is, I can not capture the packets generated or destined at the VMs. All the VMs use bridging network to connect.

Can anyone help me?

0 votes

How to Filtering and Queue the Packets of a Process using Iptable?

I am trying to queue the packets of a process so I can use libnetfilter_queue to modify them.

I read in the documentation that I should use --pid-owner processid to filter packets of a process and iptables -I

  -j NFQUEUE --queue-num  to add them to the queue.

I read the documentation but I am still confused how to do this. If any one can help me to understand this command I would appreciate it a lot.

+1 vote

How to check whether any netfilter rules are set on a host

Is there a way to find out if there any iptables rules set on a machine ?

There are some indirect ways which will not always work; for example, I know that on most hosts, iptables -S will return the following output (when no iptable rules are set)
-P INPUT ACCEPT
-P FORWARD ACCEPT
-P OUTPUT ACCEPT

So you can check whether or not the number of output lines is greater than 3 (as an indication of whether or not iptables rules are set). But there are hosts on which there are more chains then these 3; these chains are set by application/services, even without any iptable rules which are set. And after running iptables -F on these machines, iptables -S will still show more than 3 chains, even that there are no iptables rules set in these chains.

So the question is - is there a way to know whether or not netfilter rules are set on a host, regardless of the number of chains ?

0 votes

How to bind NF_ARP family in netfilter queue

Is it possible to bind multiple address families in netfilter queue? I see IPv4 show up in my queue, but not ARP. With error code removed, here is how I'm calling nfq_bind:

netfilterqueue_handle = nfq_open();
netfilterqueue_queue = nfq_create_queue( netfilterqueue_handle, 0,

nfq_bind_pf( netfilterqueue_handle, AF_INET );
nfq_bind_pf( netfilterqueue_handle, NF_ARP );

I'm thinking the more likely possibility is the iptable rules I'm using to send traffic to the queue are too restrictive. Here are the rules I have:

# Generated by iptables-save v1.4.21 on Sat Feb 14 10:40:46 2015
*nat
:PREROUTING ACCEPT [161:14105]
:INPUT ACCEPT [56:4995]
:OUTPUT ACCEPT [56:4496]
:POSTROUTING ACCEPT [56:4496]
-A POSTROUTING -s 10.0.1.0/24 -o eth0 -j MASQUERADE
COMMIT
# Completed on Sat Feb 14 10:40:46 2015
# Generated by iptables-save v1.4.21 on Sat Feb 14 10:40:46 2015
*filter
:INPUT ACCEPT [1017:217421]
:FORWARD DROP [53:2307]
:OUTPUT ACCEPT [934:211104]
:MYRA - [0:0]
-A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j MYRA
-A FORWARD -s 10.0.1.0/24 -o eth0 -m conntrack --ctstate NEW -j MYRA
-A MYRA -j NFQUEUE --queue-num 0 --queue-bypass
COMMIT
# Completed on Sat Feb 14 10:40:46 2015

Do I have to add another FORWARD line to get ARP to jump to MYRA? What would it look like?

+5 votes

IPTable: How to forward http traffic to two IPs

I want to forward all http traffic coming in from a perticular IP at local port 8080, to 2 particular IP Addresses (port 80). Is it enough to prepend the following in my IPtable

-A PREROUTING -s xx.xx.xx.xx/24 -p tcp --dport 8080 -j DNAT --to-destination xxx.xxx.xxx.xxx:80
-A PREROUTING -s xx.xx.xx.xx/24 -p tcp --dport 8080 -j DNAT --to-destination yyy.yyy.yyy.yyy:80

...