[WireGuard] Can WireGuard be adopted for older kernels?

[WireGuard] Can WireGuard be adopted for older kernels?

[Artyom Aleksyuk]

Hi!
I'm currently using IPsec to make a virtual network between two private 
computers, and I'm searching for a simpler and faster alternative. 
Looks like WireGuard is a tool which perfectly suits my needs. However 
I'm forced to use an old kernel version (3.10) because it's the only 
version available for my device. Do you have any plans to adopt 
WireGuard for older kernels? I think that there's a lot of Linux 
machines which use an old kernel and which won't be updated in the 
foreseeable future.

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

Hey Artyom,

I face the same problem as you. I have an Edge Router Lite, and I'd
really love to run WireGuard on it, but their proprietary kernel is
stuck at 3.10. Major bummer.

The current plan is to write a userspace version in Rust, that will be
cross platform and work well on old kernels. No timeline on that
though.

However, I did at one point run wireguard on really old kernels during
testing. That was before I started using the very nice udp_tunnel_*
library inside newer kernels, which then forced the dependency. But
when I did do this, it worked. Maybe I'll dust off the ugly old code
and hack something horrible together to work on old kernels. On the
other hand, this might not really be a great use of time... I don't
know.

What system are you trying to run it on? Maybe if it's the same as me,
we can both poke Ubnt together to upgrade their kernel...

Jason

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

I'm trying to run WireGuard on a single board computer (ODROID C1).=20
It's not a problem to modify a board's kernel code, however the chip=20
manufacturer (Amlogic) is focused on Android and will very unlikely=20
release a newer kernel for this board.

A userspace version is not an option because of a huge performance=20
penalty of userspace tunnels. It will certainly be slower than IPsec,=20
and performance is especially important for embedded devices.

A small offtopic: why have you chosen to use Rust for a userspace=20
version and not C or C++? In the latter case you have an option to=20
reuse some code from the kernel version.

=D0=92 =D0=92=D1=82=D0=BE=D1=80=D0=BD=D0=B8=D0=BA, 5 =D0=B8=D1=8E=D0=BB. 20=
16 =D0=B2 1:10 , Jason A. Donenfeld=20
<Jason@zx2c4.com> =D0=BD=D0=B0=D0=BF=D0=B8=D1=81=D0=B0=D0=BB:
> Hey Artyom,
>=20
> I face the same problem as you. I have an Edge Router Lite, and I'd
> really love to run WireGuard on it, but their proprietary kernel is
> stuck at 3.10. Major bummer.
>=20
> The current plan is to write a userspace version in Rust, that will be
> cross platform and work well on old kernels. No timeline on that
> though.
>=20
> However, I did at one point run wireguard on really old kernels during
> testing. That was before I started using the very nice udp_tunnel_*
> library inside newer kernels, which then forced the dependency. But
> when I did do this, it worked. Maybe I'll dust off the ugly old code
> and hack something horrible together to work on old kernels. On the
> other hand, this might not really be a great use of time... I don't
> know.
>=20
> What system are you trying to run it on? Maybe if it's the same as me,
> we can both poke Ubnt together to upgrade their kernel...
>=20
> Jason
=

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

P.S. Sorry for the incorrect From address. It was me, not Jason.

=D0=92 =D0=92=D1=82=D0=BE=D1=80=D0=BD=D0=B8=D0=BA, 5 =D0=B8=D1=8E=D0=BB. 20=
16 =D0=B2 1:46 , Jason A. Donenfeld=20
<artyom.h31@gmail.com> =D0=BD=D0=B0=D0=BF=D0=B8=D1=81=D0=B0=D0=BB:
> I'm trying to run WireGuard on a single board computer (ODROID C1).=20
> It's not a problem to modify a board's kernel code, however the chip=20
> manufacturer (Amlogic) is focused on Android and will very unlikely=20
> release a newer kernel for this board.
>=20
> A userspace version is not an option because of a huge performance=20
> penalty of userspace tunnels. It will certainly be slower than IPsec,=20
> and performance is especially important for embedded devices.
>=20
> A small offtopic: why have you chosen to use Rust for a userspace=20
> version and not C or C++? In the latter case you have an option to=20
> reuse some code from the kernel version.
>=20
> =D0=92 =D0=92=D1=82=D0=BE=D1=80=D0=BD=D0=B8=D0=BA, 5 =D0=B8=D1=8E=D0=BB. =
2016 =D0=B2 1:10 , Jason A. Donenfeld=20
> <Jason@zx2c4.com> =D0=BD=D0=B0=D0=BF=D0=B8=D1=81=D0=B0=D0=BB:
>> Hey Artyom,
>>=20
>> I face the same problem as you. I have an Edge Router Lite, and I'd
>> really love to run WireGuard on it, but their proprietary kernel is
>> stuck at 3.10. Major bummer.
>>=20
>> The current plan is to write a userspace version in Rust, that will=20
>> be
>> cross platform and work well on old kernels. No timeline on that
>> though.
>>=20
>> However, I did at one point run wireguard on really old kernels=20
>> during
>> testing. That was before I started using the very nice udp_tunnel_*
>> library inside newer kernels, which then forced the dependency. But
>> when I did do this, it worked. Maybe I'll dust off the ugly old code
>> and hack something horrible together to work on old kernels. On the
>> other hand, this might not really be a great use of time... I don't
>> know.
>>=20
>> What system are you trying to run it on? Maybe if it's the same as=20
>> me,
>> we can both poke Ubnt together to upgrade their kernel...
>>=20
>> Jason
=

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

Fix this problem with your from address immediately.

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

>From your email:

From: "Jason A. Donenfeld" <artyom.h31@gmail.com>
X-Google-Original-From: "Jason A. Donenfeld" <Jason@zx2c4.com>
X-Mailer: geary/0.11.1

This seems like a very difficult "mistake" to make.

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Artyom Aleksyuk]

[-- Attachment #1: Type: text/plain, Size: 480 bytes --]

Looks like I've discovered a bug in the Geary client. I've already filled a
bug report, hope this issue will be fixed soon. Sorry again.
5 июля 2016 г. 2:42 AM пользователь "Jason A. Donenfeld" <Jason@zx2c4.com>
написал:

> From your email:
>
> From: "Jason A. Donenfeld" <artyom.h31@gmail.com>
> X-Google-Original-From: "Jason A. Donenfeld" <Jason@zx2c4.com>
> X-Mailer: geary/0.11.1
>
> This seems like a very difficult "mistake" to make.
>

[-- Attachment #2: Type: text/html, Size: 857 bytes --]

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

Please send me the bug report you've filed. If I'm able to reproduce
the behavior I'll unban you. Otherwise I'll assume you're another
malicious impersonator. (There are a lot of those...)

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Artyom Aleksyuk]

[-- Attachment #1: Type: text/plain, Size: 377 bytes --]

Here is it:
https://bugzilla.gnome.org/show_bug.cgi?id=768419
5 июля 2016 г. 2:52 AM пользователь "Jason A. Donenfeld" <Jason@zx2c4.com>
написал:

> Please send me the bug report you've filed. If I'm able to reproduce
> the behavior I'll unban you. Otherwise I'll assume you're another
> malicious impersonator. (There are a lot of those...)
>

[-- Attachment #2: Type: text/html, Size: 704 bytes --]

Re: [WireGuard] Can WireGuard be adopted for older kernels?

[Jason A. Donenfeld]

Hi Artyom,

It looks like you were the victim of a horrible bug after all:
https://bugzilla.gnome.org/review?bug=768419&attachment=330932

Welcome back to the list.

Jason