Re: Mount NFSv4.1/NFSv4.2 filesystem in barebox?

[Martin Wege <martin.l.wege@gmail.com>]

On Mon, Oct 26, 2026 at 8:57 AM Ahmad Fatoum <a.fatoum@pengutronix.de> wrote:
>
> Hello Martin,
>
> On 10/25/26 22:07, Martin Wege wrote:
> > Hello!
> >
> > Does barebox have support to mount a NFSv4.1/NFSv.2 filesystem?
>
> No, the barebox support is for NFSv3 over UDP. This does not prevent you
> from using NFSv4/TCP during kernel boot up for the rootfs however.
>
> We would like to support NFSv4 eventually, but given that the we had
> to fix a fair share of security issues with the current network stack
> and NFSv3 implementation, we should to try to avoid repeating this
> in future.
>
> Incidentally, I just created a PR for rust support in barebox:
> https://github.com/barebox/barebox/pull/96
>
> The goal is to add smoltcp as TCP/IP stack on top of it.
>
> Then we can look towards supporting TCP protocols like HTTP and NFSv4,
> but it will take a while.
>
> This may be the third time that someone asked for NFSv4 support on the
> mailing list and it would be useful to know how are you going to use it
> and where NFSv3/UDP falls short. Is this because you are using Linux
> kernel-nfs-server and NFSv3/UDP is not enabled by your distro or is it
> because of something else?

Four issues:
1. NIST2 requirements basically EXCLUDE NFSv2/NFSv3 over UDP
2. NFSv2 and to some extent NFSv3+NFSv4.0 are declared as "obsolete"
by IETF, which again is an issue for NIST2. The whole thing has been
"red flagged" by certification authorities now several times.
3. NFSv2/NFSv3 over UDP is FRAGILE like frozen glass in a volcano
4. NFSv4.1 has much better performance and features which would make
starting through a bigger initramfs faster. Example: You would prefer
that your glass cockpit in a plan reboots a bit faster if something
went wrong (power outage) mid-flight.

Thanks,
Martin