polka-eBPF

Building environment for eBPF:

It is necessary to change the '.sh' shell script files to executable mode.

cd polka-eBPF/
$ chmod +x *.sh

1. To run, simply:

cd polka-eBPF/
sudo python3 topo_2h-eBPF.py

mininet> h1 ping h2
mininet> h2 ping h1

2. POLKA implementation on the host using eBPF:

Source codes encap_map-h x .c and decap.c codes, respectively using tc anchored in the egress, and xdp.

To compile encap using maps (encap_map-h x .c), a targeted compilation is required, which becomes:

sudo clang -O2 -g -target bpf -D__TARGET_ARCH_x86 -c encap_map.c -o encap_map.o

And, to compile encap using maps (decap.c), a targeted compilation is required, which becomes:

sudo clang -O2 -target bpf -c decap.c -o decap.o

3. Ex.: Loading values ​​into eBPF/TC maps encap_maps-h1.o:

0A 00 02 02 -> IP Address in hex 48 82 03 80 00 00 00 00 00 00 -> Polka key in hex

h1 bpftool map update name ip_key_map key hex 0A 00 02 02 value hex 48 82 03 80 00 00 00 00 00 00

4. Listing the contents of the maps:

h1 bpftool map dump name ip_key_map

5. Deleting map content:

h1 bpftool map delete name ip_key_map_h1 key hex 0A 00 02 02

_ Note: In a Mininet emulation, it is necessary to create, for each host, a separate eBPF object file (e.g., encap_map_h1.o for h1-eth0, encap_map_h2.o for h2-eth0). For each object, a distinct map must also be created — for example, ip_key_map_h1 for h1-eth0 (encap_map_h1.o) and ip_key_map_h2 for h2-eth0 (encap_map_h2.o). This is required because the namespaces of h1 and h2 share the same maps; therefore, by using different names, each host/interface maintains its own independent map table. _