On this page
Editorial correction, 1 November 2026: Earlier hardware and performance claims were not supported by traceable test records. They have been withdrawn. Documentary verification is not physical testing; no replacement measurements are claimed.
Why WireGuard on a Banana Pi
This guide concerns a full-Linux host and site-to-site planning, not OpenWrt UCI configuration. Earlier relative-speed, board-compatibility and expected-throughput claims lacked traceable tests and are withdrawn. A board name alone does not establish kernel, userspace or package support.
Prerequisites
Confirm the exact maintained Linux image, userspace architecture, kernel/module support and WireGuard package for both peers. Plan non-overlapping LAN/tunnel addressing, endpoint reachability, return routes and explicit firewall policy. Keep local console access and an independent configuration backup before changing routing. No current VPS price or provider availability is asserted.
Install WireGuard
Use the chosen distribution’s current WireGuard package instructions and verify kernel support. A kernel version alone does not prove the module exists; distributions can backport support. Official WireGuard installation guidance is the starting point. Do not install an OpenWrt package command on Debian or treat a vendor image as a supported Debian release.
Generate Key Pairs
The official WireGuard quick start describes restrictive key creation and public-key derivation. Generate keys locally with restrictive permissions before writing them; do not create a world-readable private key and tighten it afterward. Exchange public keys only and never paste private keys into a selector or diagnostic report.
Configure Both Peers
Record each peer’s public key, endpoint where needed, allowed tunnel/subnet routes and chosen listen policy. The private key stays on its owner. AllowedIPs is part of routing/peer selection, not an automatic firewall policy. Use the distribution’s documented configuration path; examples from a different OS/network manager may conflict.
Site-to-Site Routing
Forwarding and route changes can remove remote access or expose unintended networks. Preserve configuration and local recovery first. Define both forward and return routes for the real LANs; decide explicitly whether routing or NAT is appropriate. Do not enable blanket masquerading merely because an SBC is a gateway. Validate on an isolated network before production use.
MTU Tuning for Allwinner A20
Measure the actual path and account for encapsulation, IP version and uplink constraints. The previous A20-specific MTU/throughput table was not supported by raw records and is withdrawn. No universal optimal MTU is asserted. Keep each trial’s MTU, symptoms, raw results and rollback setting.
Persistent Configuration with wg-quick
On a distribution that provides wg-quick, follow its unit and configuration documentation. Enable startup only after local validation and a recovery test. OpenWrt uses a different network integration. A running service is not proof that the tunnel carries the intended application traffic.
Firewall Integration
Review the actual host firewall manager, interface names, allowed ingress and routed policy. Back up rules and maintain local recovery before applying changes. The earlier blanket UFW forwarding/NAT recipe is withheld because it did not establish safe routing for both sites. Permit only the planned traffic, then test both allowed and denied paths.
Throughput Testing with iperf3
Use separate client/server systems and distinguish endpoint from forwarding load. Record firmware/kernel, link negotiation, direction, protocol, stream count, MTU and raw repeated runs. No A20 speed or CPU-utilisation expectation is published without evidence, and a faster storage root is not a demonstrated VPN performance fix.
Troubleshooting
Investigate endpoint reachability, key pairing, allowed routes, return routing, firewall policy and path MTU separately. Handshake timestamps do not prove end-to-end application reachability. Protect any local diagnostic output: service logs and packet captures can contain private information. A basic host report does not inspect the tunnel.
Related Reading
Use the exact distribution’s WireGuard documentation, the official quick start and local recovery instructions. Router hardware selection and full-Linux host health are separate tasks. No remote repair or tunnel verification is offered by FleetCare.