Skip to contentSkip to conversation

wireguard windows socket binds to device

DISCUSSION Andre Naujoks 2 participants 3 messages May 26, 2021 – May 27, 2021

lore
Share & tools
lore
#0 lore raw

Hi all.

I am currently trying to create a wireguard tunnel over another VPN under windows.

The setup would be a wirguard tunnel through an openvpn connection.

However the wireguard implementation on windows forcibly binds itself to the network device with the default route (which is fortunately visible in the wireguard log). The wireguard peer however resides inside the other VPN and is thus not reached by the windows client.

Is there a reason why the routing of the host machine is forcibly bypassed by binding to a specific network device or is this simply a bug in the windows implementation?

Best regards and thanks in advance
Andre

#2 lore raw

Am 26.05.21 um 16:37 schrieb Jason A. Donenfeld:

Quoted 3 lines

This might help:
https://lists.zx2c4.com/pipermail/wireguard/2021-March/006442.html

Hi.

Thanks for the answer.

The registry keys were not present, so I added them manually. That and the PostUp script seem to have done the trick.

For now, I see this as a viable workaround. Thank you for that.

I still see this as a bug though. I read the reasoning behind it in the other mail (https://lore.kernel.org/wireguard/CAHmME9rXV2_YG3fGMErDeTjfHeNKhDC2cCYA6Kw93n9A328QpQ@mail.gmail.com/).

Maybe only use the IP_UNICAST_IF to the default route if AllowedIPs is 0.0.0.0/0 and/or ::/0? For all other cases it seems unnecessary (errors in those cases would be configuration problems on the users side) and in some cases (i.e. mine, the non-default-route case) just wrong.

Thanks again and best Regards
Andre