NanoPi R4S-RK3399 is a great new OpenWrt device

eginnc · October 16, 2025, 12:07pm

Looks like NanoPi R4S OpenWrt support is back with LAN initialization fix:

PR 20406 merged in main

PR 20418 pending in 24.10

eginnc · October 19, 2025, 1:20pm

I can confirm the dead LAN fix works in current main snapshot. I just updated an R4S gateway that previously needed a cold start for the LAN to come up and all is working well. Update was performed remotely.

BoltonUK · October 20, 2025, 1:18pm

I can also confirm the Snapshot firmware has also fixed the LAN issue for me as well!

Does the Snapshot firmware also address the issue where the R4S CPU’s are on the lowest ones and not the faster cores? I haven't had time to check as of yet but because Im planning on using SQM just thought I'll ask?

mikeruni · October 22, 2025, 11:26pm

Confirmed in my NanoPi R5S.

Packet steering MUST be disabled as a service at startup to avoid override "40-net-smp-affinity" particular settings from FriendlyWRT (24.10.2 r28739-d9340319c6).

System --> Startup: Mark "packet_steering" as DISABLED then reboot.
Captura de pantalla 2025-10-22 182307

More details here: https://forum.openwrt.org/t/netgear-r6220-sqm-results-downstream-cut-in-half-and-my-optimal-settings/114301/2?u=mikeruni

Tekneek · October 23, 2025, 7:23pm

I updated to 24.10.4 using Attended Sysupgrade and it went perfectly.

mark88 · November 2, 2025, 7:45pm

@anaelorlinski What is the easiest way to update AdGuard Home from your installation (as it is newer than packages available for OpenWRT) ?

anaelorlinski · November 2, 2025, 10:28pm

I started the build of a new image with latest adguardhome. You could also replace the binary directly with the one from adguardhome builds.
Note that in my image I use the premade binaries from AdguardHome and is not built from source with the OpenWrt receipe (because of nodeJS and golang dependencies that takes forever to build).

mark88 · November 2, 2025, 11:45pm

Thx, I do understand this and know how to update image (forked your repo and change hash), but I don’t want to re-install everything just for AdGuard. Is there any easy way to update it after installation of image?

mark88 · February 6, 2026, 11:01pm

Do we need to do any “fixes“ in v24 ? I remember before this should be 3f
cat /sys/class/net/eth0/queues/rx-0/rps_cpus → 01

cat /sys/class/net/eth1/queues/rx-0/rps_cpus → 04

anon_openwrt · February 7, 2026, 12:22am

No, I don’t think OpenWRT uses RPS‑based packet steering anymore. IRQ affinity is now handled correctly by the built‑in SMP‑affinity scripts, which pin the NIC interrupts to the big A72 cores.

mark88 · February 9, 2026, 2:47pm

This is MUST patch for latest version (I’ve tested v24 and I always got ERROR, after patch → OK)

You can fix it in the build or after (/www/luci-static/resources/luci.js)

github.com/openwrt/luci

luci-static: fix race condition when probing features

committed 04:34PM - 24 Dec 25 UTC

systemcrash

+8 -10

follow-up to a2fce95248a260d9471efbff94ad0cde7a90700d hasSystemFeature() inspec…ts the cached result of probeSystemFeatures() which needs to have completed. Once all of the legacy (Lua) stuff is ready, this block runs: initDOM() { ... L.loaded = true; document.dispatchEvent(new CustomEvent('luci-loaded')); }, This commit now calls View.load() once system feature probing has completed, and LuCI itself has loaded. This ensures that the following load paradigm succeeds: return view.extend({ load() { return Promise.all([ ..., L.hasSystemFeature('x'), ]); }, ... The luci-loaded check prevents waiting when LuCI is already loaded (the common case after initial page load), but the listener handles the race condition where a View is instantiated before initDOM() completes. The flow: If L.loaded is true: initDOM() already ran --> skip waiting, proceed immediately. If L.loaded is false: initDOM() hasn't run yet --> add listener --> wait for it to complete. Signed-off-by: Paul Donald <newtwen+github@gmail.com>

more about this -> https://github.com/openwrt/luci/issues/7698

woutercoppens · March 5, 2026, 9:53am

Is this still relevant for OpenWrt-25.12?

github.com/StarWhiz/NanoPi-R6S-CPU-Optimization-for-Gigabit-SQM

R4S%20CPU%20Optimization/README.md

main

# Performance Tweak for R4S

This guide is based on my [main guide for the R6S](https://github.com/StarWhiz/NanoPi-R6S-CPU-Optimization-for-Gigabit-SQM) except this guide is for the R4S instead. This applies to FriendlyWrt firmware only. Official OpenWrt 24.10.0 released on Feb 03, 2025 is already optimized and only needs packet steering (Enabled All CPUs) checked under Network > Interfaces > Global network options tab.

Please read that for full understanding. Otherwise let's just get on to it! This won't be as detailed as my main guide, but you sure can speed run this!

Before this modification I was only able to push 630 Mbps cake SQM. After the mod I was able to push 780 Mbps on cake SQM.

This assumes you installed nano already
```
opkg update
opkg install nano
```

## Script for checking current affinity
Run these commands
```
touch checkaffinity.sh
chmod +x checkaffinity.sh
nano checkaffinity.sh

This file has been truncated. show original

marcin7 · March 7, 2026, 12:12am

Did anyone go through with Attended Sysupgrade from 24.10.5 to 25.12.0 with/without keeping settings?
Any cause for concern, maybe based on how Attended Sysupgrade was for the earlier builds?

ManuVice · March 7, 2026, 7:18pm

yes with keeping settings. no problem here.

edit: If anyone asks, cake_mq isnt available for R4S

bobbenheim · March 9, 2026, 1:56pm

Most likely yes, as far as i am aware the packet steering script hasn’t changed.

woutercoppens · March 10, 2026, 8:38am

Yes with settings kept, no issues.
I always used owut to do upgrades. Don’t forget to take a backup first.

marcin7 · March 10, 2026, 9:26am

Thank you for your replies!
I can now also confirm that Attended Sysupgrade worked like a charm with settings preserved. It's been almost 48 hours and no issues at all.

I was only caught by surprise by the change in the logic for AdBlock/DNS-forcing/firewall, but I guess I should have read the documentation.

anon_openwrt · May 20, 2026, 8:41pm

I was experiencing throttling and thermal protection shutdowns (95C) when using Wireguard+SQM, so I 3D printed a fan shroud + Noctua NF-A6x15 fan and dropped the temp by 30C. Also, no more Kentucky fried fingers when touching the R4S under load.

abysso2 · May 20, 2026, 9:07pm

can you share the 3d modell prinitng scheme?

anon_openwrt · May 20, 2026, 9:11pm

Works great, it snaps onto the R4S aluminum case and then you just need to drill some holes for the fan screws.