Can this be upstreamed to openwrt/packages?

[jonirrings]

Hi! This is a fantastic project — I've tested the aarch64 binary from v1.5.0 on my OpenWrt router (BPI-R4), and it works great.

I've also put together a package Makefile and successfully built it from source in the OpenWrt build environment — it runs well on the device too.

Would you be open to having this submitted to openwrt/packages as an official package? The package would be maintained under your name as the upstream author. If you're okay with it, I'm happy to submit the PR — or if you'd prefer, I can send it to ImmortalWrt's packages feed first (which often serves as a staging area before upstreaming to openwrt/packages).

Let me know what you think!

[domcyrus]

Hi @jonirrings, thanks for putting this together and testing it on actual hardware!

I’d be very happy to see RustNet submitted directly to openwrt/packages, so please feel free to open the PR. Since you created and tested the package, I’d also be happy for you to list yourself as its package maintainer. I can, of course, continue supporting it from the upstream side.

I’d also be interested in your feedback on how well RustNet fits the OpenWrt use case. It has primarily been designed with general-purpose desktop systems in mind rather than resource-constrained routers and embedded devices. Have you noticed anything we could improve to make RustNet better suited to OpenWrt?

Thanks again for taking this on!

[domcyrus]

@jonirrings Please go ahead and submit pinned to v1.5.0, and feel free to bump to v1.6.0 once it's out (roughly 4-5 weeks away).

One heads-up for the package: v1.5.0 has a bug where --json-log silently writes nothing after the privilege drop (#416, fixed in #486 for v1.6.0). The workaround is to run with --no-uid-drop, which is a reasonable default on OpenWrt anyway. Might be worth a note in the package description or an init script default until 1.6.0.

[jonirrings]

Thanks for the quick response and the heads-up on the JSON log bug!

I'll go ahead and prepare the PR to openwrt/packages, pinning to v1.5.0 for now and bumping to v1.6.0 once it lands. And thanks for the offer on maintainership — I'm happy to take that on.

Regarding the privilege drop issue: on OpenWrt, binaries typically run as root since there's no multi-user environment in the traditional sense. Does the --json-log bug still surface in that scenario, or does root-only execution effectively sidestep the privilege drop path? If --no-uid-drop is the safe default regardless, I can bake it into the package (e.g., via an init script or a wrapper) until 1.6.0 ships.

As for how RustNet fits OpenWrt — honestly the resource footprint seems fine on the BPI-R4 (aarch64, 4GB RAM). The main consideration for router use is probably:

• Binary size: the stripped binary is around ~3.8MB on aarch64, which is quite reasonable even for smaller devices
• Feature flags: eBPF is not enabled by default on in OpenWrt build.

I'll share more detailed feedback once I've run it in production for a bit. Thanks again!

[domcyrus]

Yes, the bug still surfaces when running as plain root: rustnet drops privileges to nobody after startup whenever it starts with euid 0, sudo or not. That's how the original report on OpenWrt hit it. So baking --no-uid-drop into the init script is the right call. It also stays useful after v1.6.0, since keeping root lets process attribution work where eBPF isn't available.

[jonirrings]

PR created, waiting for review ~~