System crash consistently after transferring around 280 Mbyte over Wireguard

System crash consistently after transferring around 280 Mbyte over Wireguard

[Florian Franzeck]

[-- Attachment #1: Type: text/plain, Size: 1496 bytes --]

Hi,

I compiled version 20170421 for my turris omnia running a modified 
OpenWRT OS. I was really happy to see, how well wireguard performed 
between my 2 routers compared to OpenVPN. But i consistently could 
reproduce a relevant error: after transferring about 280Mbyte over the 
wireguard VPN over a samba share, both my routers crash and restart. As 
far as i can tell, there is no memory leak (1 router has 1Gbyte of RAM, 
the other 2 Gbyte and they don't fill up during the process).  Im not 
sure what information i could post, as no logs are written or how to 
debug this, but i'm  more than happy to help out. I opened the 
respective Ports on the firewall.
Configuration is pretty basic:
Router 1:
Interface]
ListenPort = 443
PrivateKey = xxx
[Peer]
PublicKey =  xxx
Endpoint = xxx:443
AllowedIPs = 0.0.0.0/0,::/0

Router 2:
[Interface]
ListenPort = 443
PrivateKey = xx

[Peer]PublicKey =  xxx
Endpoint = xxx:443
AllowedIPs = 0.0.0.0/0,::/0

I start the tunnel by these two scripts:

#!/bin/sh
ip link add dev tun1 type wireguard
ip addr add 10.1.3.1/32 peer 10.2.3.1 dev tun1
wg setconf tun1 /etc/wireguard/staticfmf/conf
ip link set tun1 up
route add -net 10.2.0.0 netmask 255.255.0.0 gw 10.1.3.1


#!/bin/sh
ip link add dev tun0 type wireguard
ip addr add 10.2.3.1/32 peer 10.1.3.1 dev tun0
wg setconf tun0 /etc/wireguard/staticheim/conf
ip link set tun0 up
route add -net 10.0.0.0 netmask 255.240.0.0 gw 10.2.3.1

Anyone able to help?

[-- Attachment #2: Type: text/html, Size: 1973 bytes --]

Re: System crash consistently after transferring around 280 Mbyte over Wireguard

[Jason A. Donenfeld]

Hey Florian,

So sorry to hear you're having this problem. That really really should
not be happening. I'm happy to help you debug it.

Could you enable netconsole on the box, so that you can send your
kernel log remotely over ethernet to a logging udp server in the event
of another crash? It's really easy to setup generally. Then you'll
have a good log for me to analyze.

Feel free to pop into IRC -- I'm zx2c4 -- and I can walk you through steps.

Jason