A stateless firewall that attaches to the Linux kernel's XDP hook through (e)BPF for fast packet processing. This firewall is designed to read filtering rules based off of a config file on the file system and filter incoming packets. Both IPv4 and IPv6 are supported! The protocols currently supported are TCP, UDP, and ICMP. With that said, the program includes counters for passed and dropped packets which are written to stdout (may be disabled if need to be via the no_stats config option explained below).

Additionally, if the host's network configuration or network interface card (NIC) doesn't support the XDP DRV hook (AKA native; occurs before SKB creation), the program will attempt to attach to the XDP SKB hook (AKA generic; occurs after SKB creation which is where IPTables and NFTables are processed via the netfilter kernel module). You may use overrides through the command-line to force SKB or offload modes.

With that said, reasons for a host's network configuration not supporting XDP's DRV hook may be the following.

• Running an outdated kernel that doesn't support your NIC's driver.
• Your NIC's driver not yet being supported. Here's a NIC driver XDP support list. With enough Linux kernel development knowledge, you could try implementing XDP DRV support into your non-supported NIC's driver (I'd highly recommend giving this video a watch!).
• You don't have enough RX/TX queues (e.g. not enabling multi-queue) or your RX/TX queue counts aren't matching. From the information I gathered, it's recommended to have one RX and TX queue per CPU core/thread. You could try learning how to use ethtool and try altering the NIC's RX/TX queue settings (this article may be helpful!).

I hope this project helps existing network engineers/programmers interested in utilizing XDP or anybody interested in getting into those fields! (D)DoS mitigation/prevention is such an important part of Cyber Security and understanding the concept of networking and packet flow on a low-medium level would certainly help those who are pursuing a career in the field 🙂

Command Line Usage

The following command line arguments are supported:

• --config -c => Location to config file. Default => /etc/xdpfw/xdpfw.conf.
• --offload -o => Tries to load the XDP program in hardware/offload mode (please read Offload Information below).
• --skb -s => Forces the program to load in SKB mode instead of DRV.
• --time -t => How long to run the program for in seconds before exiting. 0 or not set = infinite.
• --list -l => List all filtering rules scanned from config file.
• --help -h => Print help menu for command line options.

Offload Information

Offloading your XDP/BPF program to your system's NIC allows for the fastest packet processing you can achieve due to the NIC dropping the packets with its hardware. However, for one, there are not many NIC manufacturers that do support this feature and you're limited to the NIC's memory/processing (e.g. your BPF map sizes will be extremely limited). Additionally, there are usually stricter BPF verifier limitations for offloaded BPF programs, but you may try reaching out to the NIC's manufacturer to see if they will give you a special version of their NIC driver raising these limitations (this is what I did with one manufacturer I used).

As of this time, I am not aware of any NIC manufacturers that will be able to offload this firewall completely to the NIC due to its BPF complexity. To be honest, in the current networking age, I believe it's best to leave offloaded programs to BPF map lookups and minimum packet inspection. For example, a BPF blacklist map lookup for malicious source IPs or ports. However, XDP is still very new and I would imagine we're going to see these limitations loosened or lifted in the next upcoming years. This is why I added support for offload mode on this firewall.

Configuration File Options

Data Types

The following table quickly explains the data types used within the configuration documentation below (known data types which are not used within the configuration below will not be listed).

Main

Filter Object

TCP Options

You may additionally specified TCP header options for a filter rule which start with tcp_.

UDP Options

You may additionally specified UDP header options for a filter rule which start with udp_.

ICMP Options

You may additionally specified UDP header options for a filter rule which start with icmp_.

Notes

• All settings within a filter rule other than enabled and action are not required. This means you do not have to define them within your config.
• When a filter rule's setting is set (not NULL), but doesn't match the packet, the program moves onto the next filter rule. Therefore, all of the filter rule's settings that are set must match the packet in order to perform the action specified. Think of it as something like if src_ip == "10.50.0.3" and udp_dport == 27015: action.
• As of right now, you can specify up to 60 total filter rules. You may increase this limit by raising the MAX_FILTERS constant in the src/xdpfw.h file and then recompile the firewall. If you receive a BPF program too large error, this is due to BPF's limitations with complexity and jumps. You may try increasing BPF limitations manually or with a patch. If you want to do this, please read this README from my XDP Forwarding project.

Configuration Example

Here's an example of a config:

interface = "ens18";
update_time = 15;

filters = (
    {
        enabled = true,
        action = 0,

        udp_enabled = true,
        udp_dport = 27015
    },
    {
        enabled = true,
        action = 1,

        tcp_enabled = true,
        tcp_syn = true,
        tcp_dport = 27015
    },
    {
        enabled = true,
        action = 0,

        icmp_enabled = true,
        icmp_code = 0
    },
    {
        enabled = true,
        action = 0,
        src_ip = "10.50.0.4"
    }
);

Building & Installation

Before building, ensure the following packages are installed. These packages are installed via apt (Ubuntu, Debian, etc.), but there should be similar package names in other package managers.

# Install dependencies.
sudo apt install -y libconfig-dev llvm clang libelf-dev build-essential

# Install dependencies for building LibXDP and LibBPF.
sudo apt install -y libpcap-dev m4 gcc-multilib

# You need tools for your kernel since we need BPFTool. If this doesn't work, I'd suggest building BPFTool from source (https://github.com/libbpf/bpftool).
sudo apt install -y linux-tools-$(uname -r)

You can use git and make to build this project. The following should work:

# Clone repository via Git. Use recursive flag to download LibBPF sub-module.
git clone --recursive https://github.com/gamemann/XDP-Firewall.git

# Change directory to repository.
cd XDP-Firewall

# Build XDP-Tools and install LibXDP & LibBPF to /usr/include.
# Warning - This command uses Sudo for root access! 
# Feel free to remove sudo from the Makefile and execute as root otherwise.
make libxdp

# Build main project and install as root via Sudo.
make && sudo make install

Notes

Moved To LibXDP

On June 6th, 2023, we've moved to LibXDP from XDP Tools to load the XDP/(e)BPF program. This requires additional packages and tools to install and use with this XDP firewall as noted above.

If you're having issues with LibXDP, you may go back to commit b54c466 to use an older version of LibBPF that has worked for years with this XDP firewall.

# Make sure we're in the repository's directory.
cd XDP-Firewall

# Checkout old commit.
git checkout b54c466

# Build and install using old commit & LibBPF.
make && sudo make install

Issues On Ubuntu 20.04

If you have issues on Ubuntu 20.04 or earlier, please refer to the reply on this issue.

Basically, Clang/LLVM 12 or above is required and I'd recommend running Linux kernel 5.15 or above.

BPF For/While Loop Support + Performance Notes

This project requires for/while loop support with BPF. Older kernels will not support this and output an error such as:

libbpf: load bpf program failed: Invalid argument
libbpf: -- BEGIN DUMP LOG ---
libbpf:
back-edge from insn 113 to 100

libbpf: -- END LOG --
libbpf: failed to load program 'xdp_prog'
libbpf: failed to load object '/etc/xdpfw/xdpfw_kern.o'

It looks like BPF while/for loop support was added in kernel 5.3. Therefore, you'll need kernel 5.3 or above for this program to run properly.

Performance With For Loops

Due to the usage of a for loop inside the XDP program that handles looping through all filtering rules inside of a BPF array map, performance will be impacted depending on how many filtering rules you have configured (ultimately, the firewall doesn't scale that well). This firewall was designed to be as flexible as possible regarding configuration and is most effective when configured to add malicious source IPs to the block map for a certain amount of time which are then dropped at the beginning of the XDP program for the best performance.

Unfortunately, we can't really eliminate the for loop with the current amount of flexibility we allow (especially minimum/maximum TTL, packet lengths, IDs, etc.), unless if we were to create more BPF maps and insert many more entries which would result in a lot more memory consumed and isn't ideal at all. If we were to remove flexibility, the best approach would be to store filtering rules inside a hashed BPF map using the packet's destination IP/port as the entry's key in my opinion (this would then eliminate flexibility related to being able to specify a filtering rule to match against a single destination IP without a port, unless if we implemented multiple BPF map lookups inside the XDP program which would then impact performance). However, there are currently no plans to switch to this format due to the amount of flexibility lost and also not having the time on my side (if somebody else creates a PR to implement this, I'd be willing to have a separate branch with the new functionality for others to use if the current branch isn't working out for their needs).

The firewall is still decent at filtering non-spoofed attacks, especially when a block time is specified so that malicious IPs are filtered at the beginning of the program for some time.

Error Related To Toolchain Hardening

As stated in issue #38 by g00g1, if you have toolchain hardening enabled, you may receive the following error when compiling.

error: <unknown>:0:0: in function xdp_prog_main i32 (ptr): A call to built-in function '__stack_chk_fail' is not supported.

In order to fix this, you'll need to pass the -fno-stack-protector flag to Clang when building LibBPF and the firewall itself. You'll want to modify the Makefile for each project to add this flag. Patches for this may be found here!

Will You Make This Firewall Stateful?

There is a possibility I may make this firewall stateful in the future when I have time, but this will require a complete overhaul along with implementing application-specific filters. With that said, I am still on contract from my previous employers for certain filters of game servers. If others are willing to contribute to the project and implement these features, feel free to make pull requests!

You may also be interested in this awesome project called FastNetMon!

My Other XDP Projects

I just wanted to share other open source projects I've made which also utilize XDP (or AF_XDP sockets) for those interested. I hope code from these other projects help programmers trying to utilize XDP in their own projects!

XDP Forwarding

This project performs basic layer 3/4 forwarding using source port mapping similar to IPTables/NFTables and utilizes XDP.

Kilimanjaro

This is a complex packet processing/forwarding/dropping project I made for a gaming community I was a part of that utilizes XDP, AF_XDP, and the IPIP network protocol. I no longer work on/maintain the project, but the source code may be very helpful to other XDP developers, especially when it comes to manipulating packets inside of XDP and such.

Packet Batch (AF_XDP)

An application that utilizes fast AF_XDP Linux sockets to generate and send network packets. This is used for penetration testing including Denial of Service (DoS), network monitoring, and more!

While this application doesn't utilize native XDP or (e)BPF, I do feel it should be mentioned here. AF_XDP sockets are very fast and often used with raw XDP programs via the bpf_redirect_map() function call (also see my XDP Stats project which calculates stats in raw XDP and AF_XDP programs).

Credits

• Christian Deacon - Creator.
• Phil - Contributor.