https://github.com/dorkamotorka/sockmap-ebpf

https://github.com/dorkamotorka/sockmap-ebpf

Last synced: 16 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/dorkamotorka/sockmap-ebpf
• Owner: dorkamotorka
• Created: 2024-06-06T09:30:33.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-08-29T08:58:23.000Z (4 months ago)
• Last Synced: 2024-08-29T10:14:22.587Z (4 months ago)
• Language: C
• Size: 2.32 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# eBPF SockMap

First we need to build and run the eBPF program:

```

go generate

go build

sudo ./sockmap

```

Then we spawn a server and a client process in different network namespaces as well as setting up the cgroup(v2) and the veth pairs using:

```

mkdir -p /sys/fs/cgroup/unified/test.slice

ip netns add A

ip netns add B

ip -n A link add name veth0 type veth peer name veth0 netns B

ip -n A link set dev lo up

ip -n B link set dev lo up

ip -n A link set dev veth0 up

ip -n B link set dev veth0 up

ip -n A addr add 10.0.0.1/24 dev veth0

ip -n B addr add 10.0.0.2/24 dev veth0

```

Network namespace A is where we run our server using:

```

echo $$ > /sys/fs/cgroup/unified/test.slice/cgroup.procs # Put current shell process into the cgroup to which our BPF program is attached to 

ip netns exec A sockperf server -i 10.0.0.1 --tcp --daemonize # Run the server inside the network namespace A

```

And then run the client (new shell) in the network namespace B using:

```

echo $$ > /sys/fs/cgroup/unified/test.slice/cgroup.procs # Put current shell process into the cgroup to which our BPF program is attached to

ip netns exec B sockperf ping-pong -i 10.0.0.1 --tcp --time 30 # Run the client inside the network namespace B

```

The latency results I get on my machine is `18.282 usec` without the Sockmap, and with the latency reduces to `12.997 usec`, which is a aprox. `30%` improvement!