nfqueue-rs is a wrapper library for libnetfilter-queue. The goal is to provide a library to gain access to packets queued by the kernel packet filter.
It is important to note that these bindings will not follow blindly libnetfilter_queue API. For ex., some higher-level wrappers will be provided for the open/bind/create mechanism (using one function call instead of three).
The API is not yet stable.
To use the library, a program must
- open a queue
- bind to a network family (
AF_PACKETfor IPv4) - provide a callback function, which will be automatically called when a packet is received. The callback must return a verdict
- create the queue, providing the queue number (which must match the
--queue-numfrom the iptables rules, see below - run a loop, waiting for events. The program should also provide a clean way to exit the loop (for ex on
SIGINT)
See nfq-example for a minimal example.
You must add rules in netfilter to send packets to the userspace queue. The number of the queue (--queue-num option in netfilter) must match the number provided to create_queue().
Example of iptables rules:
iptables -A OUTPUT --destination 1.2.3.4 -j NFQUEUE --queue-num 0
Of course, you should be more restrictive, depending on your needs.
nfqueue-rs does not require root privileges, but needs to open a netlink socket and send/receive packets to the kernel.
You have several options:
- Use the CAP_NET_ADMIN capability in order to allow your application to receive from and to send packets to kernel-space:
setcap 'cap_net_admin=+ep' /path/to/program - Run your program as
rootand drop privileges